CN116858943A - Hollow shaft intelligent preparation method and system for new energy automobile - Google Patents
Hollow shaft intelligent preparation method and system for new energy automobile Download PDFInfo
- Publication number
- CN116858943A CN116858943A CN202310054345.4A CN202310054345A CN116858943A CN 116858943 A CN116858943 A CN 116858943A CN 202310054345 A CN202310054345 A CN 202310054345A CN 116858943 A CN116858943 A CN 116858943A
- Authority
- CN
- China
- Prior art keywords
- sound
- detection signal
- feature
- feature map
- hollow shaft
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000002360 preparation method Methods 0.000 title claims abstract description 52
- 238000001514 detection method Methods 0.000 claims abstract description 338
- 238000000465 moulding Methods 0.000 claims abstract description 63
- 238000004519 manufacturing process Methods 0.000 claims abstract description 23
- 239000013598 vector Substances 0.000 claims description 133
- 238000012937 correction Methods 0.000 claims description 59
- 238000009826 distribution Methods 0.000 claims description 30
- 238000010586 diagram Methods 0.000 claims description 28
- 238000000034 method Methods 0.000 claims description 25
- 230000009467 reduction Effects 0.000 claims description 24
- 239000003638 chemical reducing agent Substances 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 15
- 230000009977 dual effect Effects 0.000 claims description 10
- 230000005236 sound signal Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 abstract description 12
- 238000005516 engineering process Methods 0.000 abstract description 6
- 238000013473 artificial intelligence Methods 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 9
- 238000005065 mining Methods 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 238000000605 extraction Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 238000003062 neural network model Methods 0.000 description 4
- 238000012546 transfer Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/14—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object using acoustic emission techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/4481—Neural networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2291/00—Indexing codes associated with group G01N29/00
- G01N2291/02—Indexing codes associated with the analysed material
- G01N2291/023—Solids
- G01N2291/0234—Metals, e.g. steel
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Acoustics & Sound (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Signal Processing (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The intelligent hollow shaft preparation process and system for new energy automobile includes acquiring detection sound detecting signal and reference sound detecting signal of the hollow shaft to be detected; and excavating differential correlation characteristic information between the detection sound detection signal of the hollow shaft and the standard reference sound detection signal by adopting an artificial intelligence technology based on deep learning, and detecting the molding quality of the hollow shaft based on the differential characteristic. Therefore, the forming quality of the hollow shaft can be accurately detected, so that the production quality of the new energy automobile is ensured.
Description
Technical Field
The application relates to the technical field of intelligent preparation, in particular to an intelligent preparation method and system of a hollow shaft for a new energy automobile.
Background
The hollow shaft is characterized in that a through hole is formed in the center of the shaft body, an inner key groove is formed in the through hole, a stepped cylinder is machined on the outer surface of the shaft body, an outer key groove is formed in the outer surface of the shaft body, the central through hole of the shaft is sleeved with a main shaft of the press chamber, and input power is directly transmitted to the main shaft of the press chamber through a transmission gear arranged on the cylinder on the outer surface of the shaft body to drive the shaft. Hollow shafts are important components of new energy automobiles.
In the production and preparation of the hollow shaft, the detection of the molding quality of the prepared hollow shaft is an important procedure. In the existing preparation scheme, the formed hollow shaft is required to be sent to various detection equipment to obtain a plurality of detection parameters, and whether the forming quality of the hollow shaft meets the requirement is judged based on the whole of the plurality of detection parameters. However, this molding quality detection scheme is time-consuming, and may also have potential safety hazards such as dropping during transfer of the hollow shaft.
Therefore, an optimized hollow shaft intelligent preparation scheme for new energy automobiles is expected.
Disclosure of Invention
The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides an intelligent preparation method and system of a hollow shaft for a new energy automobile, wherein the intelligent preparation method and system of the hollow shaft are used for acquiring a detection sound detection signal and a reference sound detection signal of the hollow shaft to be detected; and excavating differential correlation characteristic information between the detection sound detection signal of the hollow shaft and the standard reference sound detection signal by adopting an artificial intelligence technology based on deep learning, and detecting the molding quality of the hollow shaft based on the differential characteristic. Therefore, the forming quality of the hollow shaft can be accurately detected, so that the production quality of the new energy automobile is ensured.
According to one aspect of the application, there is provided an intelligent preparation method of a hollow shaft for a new energy automobile, comprising:
acquiring a detection sound detection signal of a hollow shaft to be detected;
acquiring a reference sound detection signal, wherein the reference sound detection signal is a sound detection signal of a hollow shaft with molding quality meeting preset requirements;
the detection sound detection signal and the reference sound detection signal are respectively passed through a sound noise reducer based on an automatic coder to obtain a noise-reduced detection sound detection signal and a noise-reduced reference sound detection signal;
passing the denoised detection sound detection signal and the denoised reference sound detection signal through a dual detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound feature map and a reference sound feature map, wherein the first sound waveform encoder and the second sound waveform encoder have the same network structure;
calculating a difference feature map between the detected sound feature map and the reference sound feature map;
performing feature distribution correction on the differential feature map to obtain a corrected differential feature map; and
and the corrected differential feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the molding quality of the hollow shaft to be detected meets the preset requirement.
In the hollow shaft intelligent preparation method for the new energy automobile, the automatic coder-decoder comprises a sound feature extractor and a sound feature decoder.
In the above method for intelligently preparing a hollow shaft for a new energy automobile, the steps of passing the detected sound detection signal and the reference sound detection signal through a sound noise reducer based on an automatic codec to obtain a post-noise-reduction detected sound detection signal and a post-noise-reduction reference sound detection signal respectively include: the sound feature extractor performs explicit spatial coding on the detected sound detection signal and the reference sound detection signal by using a convolution layer to obtain a detected sound feature and a reference sound feature; and the sound feature decoder uses a deconvolution layer to respectively carry out deconvolution processing on the detected sound feature and the reference sound feature so as to obtain the noise-reduced detected sound detection signal and the noise-reduced reference sound detection signal.
In the hollow shaft intelligent preparation method for the new energy automobile, the first sound waveform encoder and the second sound waveform encoder are depth residual error network models.
In the above method for intelligently preparing a hollow shaft for a new energy automobile, the calculating a difference feature map between the detected sound feature map and the reference sound feature map includes: calculating a difference feature map between the detected sound feature map and the reference sound feature map using the following formula; wherein, the formula is:
Wherein F is d Representing the differential feature map, F 1 Representing the detected sound feature map, F 2 The reference sound characteristic map is represented and,representing per-position subtraction.
In the above method for intelligently preparing a hollow shaft for a new energy automobile, the performing feature distribution correction on the differential feature map to obtain a corrected differential feature map includes: performing feature map expansion on the detected sound feature map and the reference sound feature map to obtain a detected sound feature vector and a reference sound feature vector; performing vector-based Hilbert spatial constraint on the detected sound feature vector and the reference sound feature vector to obtain a corrected feature vector; and correcting the difference feature map position-by-position feature values based on the correction feature vector to obtain the corrected difference feature map.
In the above hollow shaft intelligent preparation method for a new energy automobile, the performing a hilbert space constraint of a vector mode basis on the detected sound feature vector and the reference sound feature vector to obtain a corrected feature vector includes: performing vector-based Hilbert spatial constraint on the detected sound feature vector and the reference sound feature vector by the following formula to obtain the correction feature vector; wherein, the formula is:
Wherein V is 1 And V 2 Representing the detected sound feature vector and the reference sound feature vector, respectively, |·|| 2 Representing the two norms of the vector, cov 1D Expressed in terms of convolution operator (||V) 1 || 2 ,||V 2 || 2 ,V 1 V 2 T ) Vector pairOne-dimensional convolution is performed, alpha and beta being weightsHeavy super parameter, V c Representing the correction feature vector.
In the above method for intelligently preparing a hollow shaft for a new energy automobile, the correcting the position-by-position characteristic value of the differential characteristic map based on the correction characteristic vector to obtain the corrected differential characteristic map includes: performing dimension reconstruction on the correction feature vector to obtain a correction feature map; and multiplying the difference feature map by position points by taking the correction feature map as a weighted feature map to obtain the corrected difference feature map.
In the above intelligent preparation method for a hollow shaft of a new energy automobile, the step of passing the corrected differential feature map through a classifier to obtain a classification result, where the classification result is used to indicate whether the molding quality of the hollow shaft to be detected meets a predetermined requirement, and the method includes: expanding the corrected differential feature map into classification feature vectors according to row vectors or column vectors; performing full-connection coding on the classification feature vectors by using a plurality of full-connection layers of the classifier to obtain coded classification feature vectors; and passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.
According to another aspect of the present application, there is provided a hollow shaft intelligent preparation system for a new energy automobile, comprising:
the detection sound signal acquisition module is used for acquiring a detection sound detection signal of the hollow shaft to be detected;
the reference sound signal acquisition module is used for acquiring a reference sound detection signal, wherein the reference sound detection signal is a sound detection signal of a hollow shaft with molding quality meeting preset requirements;
the noise reduction module is used for respectively passing the detection sound detection signal and the reference sound detection signal through a sound noise reducer based on an automatic coder and decoder to obtain a post-noise reduction detection sound detection signal and a post-noise reduction reference sound detection signal;
the double detection module is used for enabling the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal to pass through a double detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound characteristic diagram and a reference sound characteristic diagram, and the first sound waveform encoder and the second sound waveform encoder have the same network structure;
the difference feature calculation module is used for calculating a difference feature map between the detected sound feature map and the reference sound feature map;
The characteristic distribution correction module is used for carrying out characteristic distribution correction on the differential characteristic map so as to obtain a corrected differential characteristic map; and
and the molding quality generation module is used for enabling the corrected differential feature map to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the molding quality of the hollow shaft to be detected meets the preset requirement.
Compared with the prior art, the intelligent preparation method and the intelligent preparation system for the hollow shaft of the new energy automobile are used for acquiring the detection sound detection signal and the reference sound detection signal of the hollow shaft to be detected; and excavating differential correlation characteristic information between the detection sound detection signal of the hollow shaft and the standard reference sound detection signal by adopting an artificial intelligence technology based on deep learning, and detecting the molding quality of the hollow shaft based on the differential characteristic. Therefore, the forming quality of the hollow shaft can be accurately detected, so that the production quality of the new energy automobile is ensured.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing embodiments of the present application in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.
Fig. 1 is a schematic view of a scenario of an intelligent preparation method of a hollow shaft for a new energy automobile according to an embodiment of the application.
Fig. 2 is a flowchart of an intelligent preparation method of a hollow shaft for a new energy automobile according to an embodiment of the application.
Fig. 3 is a schematic diagram of an architecture of an intelligent preparation method of a hollow shaft for a new energy automobile according to an embodiment of the application.
Fig. 4 is a flowchart of the sub-step of step S130 in the hollow shaft intelligent preparation method for a new energy automobile according to an embodiment of the present application.
Fig. 5 is a flowchart of the sub-step of step S160 in the hollow shaft intelligent preparation method for a new energy automobile according to an embodiment of the present application.
Fig. 6 is a flowchart of the sub-step of step S330 in the hollow shaft intelligent preparation method for a new energy automobile according to an embodiment of the present application.
Fig. 7 is a flowchart of the sub-step of step S170 in the hollow shaft intelligent preparation method for a new energy automobile according to an embodiment of the present application.
Fig. 8 is a block diagram of a hollow shaft intelligent preparation system for a new energy automobile according to an embodiment of the present application.
Detailed Description
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
Scene overview
As described above, in the existing manufacturing scheme, the formed hollow shaft is required to be sent to various detecting devices to obtain a plurality of detecting parameters, and whether the forming quality of the hollow shaft meets the requirement is judged based on the whole of the plurality of detecting parameters. However, this molding quality detection scheme is time-consuming, and may also have potential safety hazards such as dropping during transfer of the hollow shaft. Therefore, an optimized hollow shaft intelligent preparation scheme for new energy automobiles is expected.
Accordingly, it is considered that in the process of actually performing molding quality detection on the prepared hollow shaft, quality detection can be achieved by performing characteristic information comparison on the quality of the prepared hollow shaft and the quality of the hollow shaft which meets the preset requirement, and it is considered that since image data of the hollow shaft is difficult to acquire, the image data is difficult to capture hidden quality characteristics in the hollow shaft, and accuracy of molding quality detection is reduced. In particular, considering that the current sound detection technology has been gradually applied to quality detection of industrial products, in the technical solution of the present application, it is desirable to perform molding quality detection of a hollow shaft by comparing the difference characteristics of a detected sound detection signal of the hollow shaft to be detected and a sound detection signal of the hollow shaft whose molding quality meets a predetermined requirement in a high-dimensional space. In the process, the difficulty is how to dig out the differential correlation characteristic information between the detection sound detection signal of the hollow shaft and the standard reference sound detection signal, so as to realize the molding quality detection of the hollow shaft and ensure the production quality of the new energy automobile.
In recent years, deep learning and neural networks have been widely used in the fields of computer vision, natural language processing, text signal processing, and the like. In addition, deep learning and neural networks have also shown levels approaching and even exceeding humans in the fields of image classification, object detection, semantic segmentation, text translation, and the like.
Deep learning and development of neural networks provide new solutions and solutions for mining differential correlation characteristic information between the detected sound detection signal of the hollow shaft and the standard reference sound detection signal. Those of ordinary skill in the art will appreciate that the deep learning based deep neural network model may be adapted with appropriate training strategies, such as by a gradient descent back-propagation algorithm, to adjust parameters of the deep neural network model to enable modeling of complex nonlinear correlations between things, which is obviously suitable for modeling and mining differential correlation characteristic information between the detected sound detection signal of the hollow shaft and a standard reference sound detection signal.
Specifically, in the technical scheme of the application, firstly, a detection sound detection signal of a hollow shaft to be detected and a reference sound detection signal are acquired through a sound sensor, wherein the reference sound detection signal is a sound detection signal of the hollow shaft with molding quality meeting preset requirements. Then, in the process of actually collecting the sound detection signals, the hidden characteristic information of the detection sound detection signals and the reference sound detection signals is fuzzy due to the influence of external noise caused by the preparation environment of the hollow shaft, so that the accuracy of forming quality detection of the hollow shaft is lower. Therefore, in the technical scheme of the application, in order to accurately detect the molding quality of the hollow shaft, noise reduction processing is required to be performed on the collected sound detection signals. Specifically, in the technical scheme of the application, the detection sound detection signal and the reference sound detection signal are respectively passed through a sound noise reducer based on an automatic coder-decoder to obtain a post-noise-reduction detection sound detection signal and a post-noise-reduction reference sound detection signal. In particular, the automatic codec includes a sound feature extractor and a sound feature decoder, and the sound feature extractor explicitly spatially encodes the detected sound detection signal and the reference sound detection signal using a convolution layer to obtain a detected sound feature and a reference sound feature, respectively; and the sound feature decoder uses a deconvolution layer to respectively carry out deconvolution processing on the detected sound feature and the reference sound feature so as to obtain the noise-reduced detected sound detection signal and the noise-reduced reference sound detection signal.
Then, after the noise reduction processing is performed on the detected sound detection signal and the reference sound detection signal, the post-noise reduction detected sound detection signal and the post-noise reduction reference sound detection signal are subjected to feature mining by using a convolutional neural network model having excellent performance in terms of implicit feature extraction. In particular, in order to sufficiently and accurately extract the differential correlation feature distribution information of the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal in a high-dimensional space in the process of feature mining, so as to accurately detect the molding quality of the hollow shaft, in the technical scheme of the application, the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal are passed through a dual detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound feature map and a reference sound feature map. It should be noted that, here, the first acoustic waveform encoder and the second acoustic waveform encoder have the same network structure, and the first acoustic waveform encoder and the second acoustic waveform encoder are depth residual network models. It should be understood that the dual detection model including the acoustic waveform encoder with the same network structure is used to perform the feature extraction of the post-noise reduction detection acoustic detection signal and the post-noise reduction reference acoustic detection signal respectively, so as to extract the feature information that the difference between the acoustic detection signals at the acoustic source domain is not obvious, thereby more accurately performing the quality detection on whether the molding quality of the hollow shaft meets the predetermined requirement.
Further, a difference feature diagram between the detection sound feature diagram and the reference sound feature diagram is calculated, so that the difference correlation feature distribution information about the quality of the hollow shaft in a high-dimensional space of the detection sound detection signal of the hollow shaft to be detected and the sound detection signal of the hollow shaft with the molding quality meeting the preset requirement is represented, and the difference correlation feature distribution information is used as a classification feature diagram to be subjected to classification processing in a classifier, so that a classification result for representing whether the molding quality of the hollow shaft to be detected meets the preset requirement is obtained. That is, in the technical scheme of the application, the label of the classifier comprises that the molding quality of the hollow shaft to be detected meets the preset requirement, and the molding quality of the hollow shaft to be detected does not meet the preset requirement, wherein the classifier determines which classification label the classification characteristic map belongs to through a soft maximum function, so that the molding quality detection of the hollow shaft to be detected is realized, and the production quality of the energy automobile is further ensured.
In particular, in the technical solution of the present application, when calculating the difference feature map between the detected sound feature map and the reference sound feature map, since the detected sound feature map and the reference sound feature map are obtained by including independent branches of an automatic codec-based sound noise reducer and a sound waveform encoder, respectively, the feature distribution thereof may not be uniform in the feature extraction direction of the signal waveform semantics, which results in poor convergence of the overall feature distribution of the difference feature map, that is, the fitting effect thereof by a classifier may be deteriorated. On the other hand, if weights are directly set for the detected sound feature map and the reference sound feature map to fit the convergence direction thereof, the correlation between the feature values of the obtained differential feature map may be high, thereby reducing the classification accuracy of the differential feature map.
Thus, the detected sound feature map and the reference sound feature map are first expanded into detected sound feature vectors, e.g., denoted as V 1 And reference sound feature vectors, e.g. denoted V 2 And then to the detected sound feature vector V 1 And reference sound feature vector V 2 Performing Hilbert spatial constraint of vector modulus to obtain the correction feature vector V c Expressed as:
Cov 1D representing one-dimensional convolution operations, i.e. with convolution operators (||V) 1 || 2 ,||V 2 || 2 ,V 1 V 2 T ) Vector pairOne-dimensional convolution is performed where α and β are weight super-parameters.
Here, the correction feature vector V is obtained by convolving the vector V with a convolution operator in the hilbert space defining a vector sum modulo the vector inner product c Constraint is made to correct the feature vector V c Is defined in a finite closed domain in the Hilbert space based on the modulus of the vector and promotes the corrected feature vector V c Is a high-dimensional manifold of feature distributions of (a)Is described, thereby achieving sparse correlation between feature values while maintaining convergence of the feature distribution as a whole. Then, the correction feature vector V c And restoring the difference characteristic map to a correction characteristic map, and carrying out point multiplication on the difference characteristic map and the correction characteristic map, so that the fitting effect of the corrected difference characteristic map through a classifier and the accuracy of a classification result can be improved. Therefore, the forming quality of the hollow shaft can be accurately detected, and the production quality of the new energy automobile is ensured.
Based on the above, the application provides an intelligent preparation method of a hollow shaft for a new energy automobile, which comprises the following steps: acquiring a detection sound detection signal of a hollow shaft to be detected; acquiring a reference sound detection signal, wherein the reference sound detection signal is a sound detection signal of a hollow shaft with molding quality meeting preset requirements; the detection sound detection signal and the reference sound detection signal are respectively passed through a sound noise reducer based on an automatic coder to obtain a noise-reduced detection sound detection signal and a noise-reduced reference sound detection signal; passing the denoised detection sound detection signal and the denoised reference sound detection signal through a dual detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound feature map and a reference sound feature map, wherein the first sound waveform encoder and the second sound waveform encoder have the same network structure; calculating a difference feature map between the detected sound feature map and the reference sound feature map; performing feature distribution correction on the differential feature map to obtain a corrected differential feature map; and passing the corrected differential feature map through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the molding quality of the hollow shaft to be detected meets the preset requirement.
Fig. 1 is a schematic view of a scenario of an intelligent preparation method of a hollow shaft for a new energy automobile according to an embodiment of the application. As shown in fig. 1, in this application scenario, first, a detection sound detection signal (e.g., C1 as illustrated in fig. 1) of a hollow shaft to be detected is acquired, and a reference sound detection signal (e.g., C2 as illustrated in fig. 1) which is a sound detection signal of a hollow shaft whose molding quality meets a predetermined requirement is acquired; then, the acquired detection sound detection signal and reference sound detection signal are input into a server (e.g., S as illustrated in fig. 1) deployed with a hollow shaft intelligentized preparation algorithm for a new energy automobile, wherein the server is capable of processing the detection sound detection signal and the reference sound detection signal based on the hollow shaft intelligentized preparation algorithm for a new energy automobile to generate a classification result for indicating whether the molding quality of the hollow shaft to be detected meets a predetermined requirement.
Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.
Exemplary method
Fig. 2 is a flowchart of an intelligent preparation method of a hollow shaft for a new energy automobile according to an embodiment of the application. As shown in fig. 2, the intelligent preparation method of the hollow shaft for the new energy automobile according to the embodiment of the application comprises the following steps: s110, acquiring a detection sound detection signal of a hollow shaft to be detected; s120, acquiring a reference sound detection signal, wherein the reference sound detection signal is a sound detection signal of a hollow shaft with molding quality meeting preset requirements; s130, the detection sound detection signal and the reference sound detection signal are respectively passed through a sound noise reducer based on an automatic coder-decoder to obtain a noise-reduced detection sound detection signal and a noise-reduced reference sound detection signal; s140, the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal pass through a double detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound characteristic diagram and a reference sound characteristic diagram, wherein the first sound waveform encoder and the second sound waveform encoder have the same network structure; s150, calculating a difference characteristic diagram between the detected sound characteristic diagram and the reference sound characteristic diagram; s160, carrying out feature distribution correction on the differential feature map to obtain a corrected differential feature map; and S170, passing the corrected differential feature map through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the molding quality of the hollow shaft to be detected meets the preset requirement.
Fig. 3 is a schematic diagram of an architecture of an intelligent preparation method of a hollow shaft for a new energy automobile according to an embodiment of the application. As shown in fig. 3, in the network architecture, first, a detection sound detection signal of a hollow shaft to be detected is obtained; then, a reference sound detection signal is obtained, wherein the reference sound detection signal is a sound detection signal of a hollow shaft with molding quality meeting preset requirements; then, the detection sound detection signal and the reference sound detection signal are respectively passed through a sound noise reducer based on an automatic coder-decoder to obtain a noise-reduced detection sound detection signal and a noise-reduced reference sound detection signal; then, the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal are passed through a double detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound feature map and a reference sound feature map, wherein the first sound waveform encoder and the second sound waveform encoder have the same network structure; then, calculating a difference feature map between the detected sound feature map and the reference sound feature map; then, carrying out feature distribution correction on the differential feature map to obtain a corrected differential feature map; and finally, the corrected differential feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the molding quality of the hollow shaft to be detected meets the preset requirement.
Specifically, in step S110 and step S120, a detection sound detection signal of the hollow shaft to be detected is obtained; and acquiring a reference sound detection signal, wherein the reference sound detection signal is a sound detection signal of the hollow shaft with molding quality meeting preset requirements.
As described above, in the existing manufacturing scheme, the formed hollow shaft is required to be sent to various detecting devices to obtain a plurality of detecting parameters, and whether the forming quality of the hollow shaft meets the requirement is judged based on the whole of the plurality of detecting parameters. However, this molding quality detection scheme is time-consuming, and may also have potential safety hazards such as dropping during transfer of the hollow shaft. Therefore, an optimized hollow shaft intelligent preparation scheme for new energy automobiles is expected.
Accordingly, it is considered that in the process of actually performing molding quality detection on the prepared hollow shaft, quality detection can be achieved by performing characteristic information comparison on the quality of the prepared hollow shaft and the quality of the hollow shaft which meets the preset requirement, and it is considered that since image data of the hollow shaft is difficult to acquire, the image data is difficult to capture hidden quality characteristics in the hollow shaft, and accuracy of molding quality detection is reduced. In particular, considering that the current sound detection technology has been gradually applied to quality detection of industrial products, in the technical solution of the present application, it is desirable to perform molding quality detection of a hollow shaft by comparing the difference characteristics of a detected sound detection signal of the hollow shaft to be detected and a sound detection signal of the hollow shaft whose molding quality meets a predetermined requirement in a high-dimensional space. In the process, the difficulty is how to dig out the differential correlation characteristic information between the detection sound detection signal of the hollow shaft and the standard reference sound detection signal, so as to realize the molding quality detection of the hollow shaft and ensure the production quality of the new energy automobile.
In recent years, deep learning and neural networks have been widely used in the fields of computer vision, natural language processing, text signal processing, and the like. In addition, deep learning and neural networks have also shown levels approaching and even exceeding humans in the fields of image classification, object detection, semantic segmentation, text translation, and the like.
Deep learning and development of neural networks provide new solutions and solutions for mining differential correlation characteristic information between the detected sound detection signal of the hollow shaft and the standard reference sound detection signal. Those of ordinary skill in the art will appreciate that the deep learning based deep neural network model may be adapted with appropriate training strategies, such as by a gradient descent back-propagation algorithm, to adjust parameters of the deep neural network model to enable modeling of complex nonlinear correlations between things, which is obviously suitable for modeling and mining differential correlation characteristic information between the detected sound detection signal of the hollow shaft and a standard reference sound detection signal.
Specifically, in the technical scheme of the application, firstly, a detection sound detection signal of a hollow shaft to be detected and a reference sound detection signal are acquired through a sound sensor, wherein the reference sound detection signal is a sound detection signal of the hollow shaft with molding quality meeting preset requirements.
Specifically, in step S130, the detected sound detection signal and the reference sound detection signal are respectively passed through a sound noise reducer based on an automatic codec to obtain a post-noise-reduction detected sound detection signal and a post-noise-reduction reference sound detection signal.
Then, in the process of actually collecting the sound detection signals, the hidden characteristic information of the detection sound detection signals and the reference sound detection signals is fuzzy due to the influence of external noise caused by the preparation environment of the hollow shaft, so that the accuracy of forming quality detection of the hollow shaft is lower. Therefore, in the technical scheme of the application, in order to accurately detect the molding quality of the hollow shaft, noise reduction processing is required to be performed on the collected sound detection signals.
Specifically, in the technical scheme of the application, the detection sound detection signal and the reference sound detection signal are respectively passed through a sound noise reducer based on an automatic coder-decoder to obtain a post-noise-reduction detection sound detection signal and a post-noise-reduction reference sound detection signal. In particular, the automatic codec includes a sound feature extractor and a sound feature decoder, and the sound feature extractor explicitly spatially encodes the detected sound detection signal and the reference sound detection signal using a convolution layer to obtain a detected sound feature and a reference sound feature, respectively; and the sound feature decoder uses a deconvolution layer to respectively carry out deconvolution processing on the detected sound feature and the reference sound feature so as to obtain the noise-reduced detected sound detection signal and the noise-reduced reference sound detection signal.
That is, in an embodiment of the present application, fig. 4 is a flowchart showing a sub-step of step S130 in the hollow shaft intelligent manufacturing method for a new energy automobile according to an embodiment of the present application, and as shown in fig. 4, the step of passing the detected sound detection signal and the reference sound detection signal through a sound noise reducer based on an automatic codec to obtain a post-noise-reduction detected sound detection signal and a post-noise-reduction reference sound detection signal, respectively, includes: s210, the sound feature extractor uses a convolution layer to respectively perform explicit space coding on the detected sound detection signal and the reference sound detection signal so as to obtain a detected sound feature and a reference sound feature; and S220, the voice feature decoder uses a deconvolution layer to respectively carry out deconvolution processing on the detected voice feature and the reference voice feature so as to obtain the post-noise-reduction detected voice detection signal and the post-noise-reduction reference voice detection signal.
Specifically, in step S140, the post-noise reduction detection sound detection signal and the post-noise reduction reference sound detection signal are passed through a dual detection model including a first sound waveform encoder and a second sound waveform encoder, which have the same network structure, to obtain a detection sound feature map and a reference sound feature map.
Then, after the noise reduction processing is performed on the detected sound detection signal and the reference sound detection signal, the post-noise reduction detected sound detection signal and the post-noise reduction reference sound detection signal are subjected to feature mining by using a convolutional neural network model having excellent performance in terms of implicit feature extraction.
In particular, in order to sufficiently and accurately extract the differential correlation feature distribution information of the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal in a high-dimensional space in the process of feature mining, so as to accurately detect the molding quality of the hollow shaft, in the technical scheme of the application, the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal are passed through a dual detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound feature map and a reference sound feature map.
It should be noted that, here, the first acoustic waveform encoder and the second acoustic waveform encoder have the same network structure, and the first acoustic waveform encoder and the second acoustic waveform encoder are depth residual network models. It should be understood that the dual detection model including the acoustic waveform encoder with the same network structure is used to perform the feature extraction of the post-noise reduction detection acoustic detection signal and the post-noise reduction reference acoustic detection signal respectively, so as to extract the feature information that the difference between the acoustic detection signals at the acoustic source domain is not obvious, thereby more accurately performing the quality detection on whether the molding quality of the hollow shaft meets the predetermined requirement.
Specifically, in step S150, a difference feature map between the detected sound feature map and the reference sound feature map is calculated. Further, a difference feature diagram between the detection sound feature diagram and the reference sound feature diagram is calculated, so that the difference correlation feature distribution information about the quality of the hollow shaft in a high-dimensional space of the detection sound detection signal of the hollow shaft to be detected and the sound detection signal of the hollow shaft with the molding quality meeting the preset requirement is represented, and the difference correlation feature distribution information is used as a classification feature diagram to be subjected to classification processing in a classifier, so that a classification result for representing whether the molding quality of the hollow shaft to be detected meets the preset requirement is obtained.
Wherein said calculating a difference signature between said detected sound signature and said reference sound signature comprises: calculating a difference feature map between the detected sound feature map and the reference sound feature map using the following formula; wherein, the formula is:
wherein F is d Representing the differential feature map, F 1 Representing the detected sound feature map, F 2 Representing the reference soundThe characteristic diagram is shown in the drawing,representing per-position subtraction.
Specifically, in step S160, the characteristic distribution correction is performed on the differential characteristic map to obtain a corrected differential characteristic map. Fig. 5 is a flowchart of the substep of step S160 in the hollow shaft intelligent preparation method for a new energy automobile according to an embodiment of the present application, as shown in fig. 5, the performing feature distribution correction on the differential feature map to obtain a corrected differential feature map, including: s310, performing feature map expansion on the detected sound feature map and the reference sound feature map to obtain a detected sound feature vector and a reference sound feature vector; s320, performing Hilbert space constraint of vector mode basis on the detected sound feature vector and the reference sound feature vector to obtain a correction feature vector; and S330, correcting the difference feature map by position feature values based on the correction feature vector to obtain the corrected difference feature map.
In particular, in the technical solution of the present application, when calculating the difference feature map between the detected sound feature map and the reference sound feature map, since the detected sound feature map and the reference sound feature map are obtained by including independent branches of an automatic codec-based sound noise reducer and a sound waveform encoder, respectively, the feature distribution thereof may not be uniform in the feature extraction direction of the signal waveform semantics, which results in poor convergence of the overall feature distribution of the difference feature map, that is, the fitting effect thereof by a classifier may be deteriorated. On the other hand, if weights are directly set for the detected sound feature map and the reference sound feature map to fit the convergence direction thereof, the correlation between the feature values of the obtained differential feature map may be high, thereby reducing the classification accuracy of the differential feature map.
Thus, the detected sound feature map and the reference sound feature map are first expanded into detected sound feature vectors, for exampleSuch as denoted as V 1 And reference sound feature vectors, e.g. denoted V 2 And then to the detected sound feature vector V 1 And reference sound feature vector V 2 Performing Hilbert spatial constraint of vector modulus to obtain the correction feature vector V c . That is, the detection sound feature vector and the reference sound feature vector are subjected to hilbert space constraint of vector basis in the following formula to obtain the correction feature vector; wherein, the formula is:
wherein V is 1 And V 2 Representing the detected sound feature vector and the reference sound feature vector, respectively, |·|| 2 Representing the two norms of the vector, cov 1D Expressed in terms of convolution operator (||V) 1 || 2 ,||V 2 || 2 ,V 1 V 2 T ) Vector pairOne-dimensional convolution is carried out, alpha and beta are weight super parameters, V c Representing the correction feature vector.
Here, the correction feature vector V is obtained by convolving the vector V with a convolution operator in the hilbert space defining a vector sum modulo the vector inner product c Constraint is made to correct the feature vector V c Is defined in a finite closed domain in the Hilbert space based on the modulus of the vector and promotes the corrected feature vector V c Orthogonality between the base dimensions of the high-dimensional manifold of the feature distribution, thereby achieving sparse correlation between feature values while maintaining convergence of the feature distribution as a whole. Then, the correction feature vector V c And restoring the difference characteristic map to a correction characteristic map, and carrying out point multiplication on the difference characteristic map and the correction characteristic map, so that the fitting effect of the corrected difference characteristic map through a classifier and the accuracy of a classification result can be improved. In this way, the forming quality of the hollow shaft can be accurately improvedAnd detecting the rows, so as to ensure the production quality of the new energy automobile.
Further, in the embodiment of the present application, fig. 6 is a flowchart of a sub-step of step S330 in the hollow shaft intelligent preparation method for a new energy automobile according to the embodiment of the present application, as shown in fig. 6, the correcting the position-by-position eigenvalue correction of the differential eigenvector based on the correction eigenvector to obtain the corrected differential eigenvector includes: s410, carrying out dimension reconstruction on the correction feature vector to obtain a correction feature map; and S420, multiplying the difference feature map by position points by taking the correction feature map as a weighted feature map to obtain the corrected difference feature map.
Specifically, in step S170, the corrected differential feature map is passed through a classifier to obtain a classification result, where the classification result is used to indicate whether the molding quality of the hollow shaft to be detected meets a predetermined requirement. That is, in the technical scheme of the application, the label of the classifier comprises that the molding quality of the hollow shaft to be detected meets the preset requirement, and the molding quality of the hollow shaft to be detected does not meet the preset requirement, wherein the classifier determines which classification label the classification characteristic map belongs to through a soft maximum function, so that the molding quality detection of the hollow shaft to be detected is realized, and the production quality of the energy automobile is further ensured.
Fig. 7 is a flowchart of a sub-step of step S170 in the intelligent preparation method of a hollow shaft for a new energy automobile according to an embodiment of the present application, as shown in fig. 7, the step of passing the corrected differential feature map through a classifier to obtain a classification result, where the classification result is used to indicate whether the molding quality of the hollow shaft to be detected meets a predetermined requirement, and includes: s510, expanding the corrected differential feature map into classified feature vectors according to row vectors or column vectors; s520, performing full-connection coding on the classification feature vectors by using a plurality of full-connection layers of the classifier to obtain coded classification feature vectors; and S530, passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.
In a specific example of the present application, the classifier is used to process the corrected differential feature map in the following formula to obtain the classification result; wherein, the formula is:
O＝softmax{(W n ,B n ):…:(W 1 ,B 1 ) Project (F), where W 1 To W n Is a weight matrix, B 1 To B n As a bias vector, project (F) is to Project the corrected differential feature map as a vector.
In summary, according to the method for intelligently preparing the hollow shaft for the new energy automobile, which is disclosed by the embodiment of the application, a detection sound detection signal and a reference sound detection signal of the hollow shaft to be detected are obtained; and excavating differential correlation characteristic information between the detection sound detection signal of the hollow shaft and the standard reference sound detection signal by adopting an artificial intelligence technology based on deep learning, and detecting the molding quality of the hollow shaft based on the differential characteristic. Therefore, the forming quality of the hollow shaft can be accurately detected, so that the production quality of the new energy automobile is ensured.
Exemplary System
Fig. 8 is a block diagram of a hollow shaft intelligent preparation system for a new energy automobile according to an embodiment of the present application. As shown in fig. 8, a hollow shaft intelligent preparation system 100 for a new energy automobile according to an embodiment of the present application includes: a detection sound signal acquisition module 110, configured to acquire a detection sound detection signal of a hollow shaft to be detected; the reference sound signal obtaining module 120 is configured to obtain a reference sound detection signal, where the reference sound detection signal is a sound detection signal of a hollow shaft with molding quality meeting a predetermined requirement; the noise reduction module 130 is configured to pass the detected sound detection signal and the reference sound detection signal through a sound noise reducer based on an automatic codec, respectively, so as to obtain a post-noise reduction detected sound detection signal and a post-noise reduction reference sound detection signal; the dual detection module 140 is configured to pass the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal through a dual detection model including a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound feature map and a reference sound feature map, where the first sound waveform encoder and the second sound waveform encoder have the same network structure; a difference feature calculation module 150, configured to calculate a difference feature map between the detected sound feature map and the reference sound feature map; the feature distribution correction module 160 is configured to perform feature distribution correction on the differential feature map to obtain a corrected differential feature map; and a molding quality generating module 170, configured to pass the corrected differential feature map through a classifier to obtain a classification result, where the classification result is used to indicate whether the molding quality of the hollow shaft to be detected meets a predetermined requirement.
In one example, in the hollow shaft intelligent production system 100 for a new energy vehicle described above, the automatic codec includes a sound feature extractor and a sound feature decoder.
In one example, in the hollow shaft intelligent manufacturing system 100 for a new energy vehicle, the noise reduction module includes: an explicit spatial coding unit, configured to perform explicit spatial coding on the detected sound detection signal and the reference sound detection signal by using a convolution layer to obtain a detected sound feature and a reference sound feature; and a deconvolution processing unit, configured to deconvolute the detected sound feature and the reference sound feature by using deconvolution layers, so as to obtain the post-noise-reduction detected sound detection signal and the post-noise-reduction reference sound detection signal.
In one example, in the hollow shaft intelligent production system 100 for a new energy vehicle described above, the first acoustic waveform encoder and the second acoustic waveform encoder are depth residual network models.
In one example, in the hollow shaft intelligent manufacturing system 100 for a new energy automobile, the differential feature calculation module is further configured to: calculating a difference feature map between the detected sound feature map and the reference sound feature map using the following formula; wherein, the formula is:
Wherein F is d Representing the differential feature map, F 1 Representing the detected sound feature map, F 2 The reference sound characteristic map is represented and,representing per-position subtraction.
In one example, in the hollow shaft intelligent manufacturing system 100 for a new energy vehicle, the feature distribution correction module includes: the feature map unfolding unit is used for conducting feature map unfolding on the detected sound feature map and the reference sound feature map to obtain a detected sound feature vector and a reference sound feature vector; the space constraint unit is used for carrying out Hilbert space constraint of a vector mode base on the detected sound feature vector and the reference sound feature vector so as to obtain a correction feature vector; and a correction unit configured to perform position-by-position feature value correction on the differential feature map based on the correction feature vector to obtain the corrected differential feature map.
In one example, in the hollow shaft intelligent manufacturing system 100 for a new energy automobile described above, the space constraint unit is further configured to: performing vector-based Hilbert spatial constraint on the detected sound feature vector and the reference sound feature vector by the following formula to obtain the correction feature vector; wherein, the formula is:
Wherein V is 1 And V 2 Representing the detected sound feature vector and the reference sound feature vector, respectively, |·|| 2 Representing the two norms of the vector, cov 1D Expressed in terms of convolution operator (||V) 1 || 2 ,||V 2 || 2 ,V 1 V 2 T ) Vector pairOne-dimensional convolution is carried out, alpha and beta are weight super parameters, V c Representing the correction feature vector.
In one example, in the hollow shaft intelligent manufacturing system 100 for a new energy vehicle, the correction unit includes: the dimension reconstruction subunit is used for carrying out dimension reconstruction on the correction feature vector to obtain a correction feature map; and the calculating subunit is used for multiplying the difference characteristic diagram by position points by taking the correction characteristic diagram as a weighted characteristic diagram to obtain the corrected difference characteristic diagram.
In one example, in the hollow shaft intelligent manufacturing system 100 for a new energy automobile, the molding quality generating module includes: the unfolding unit is used for unfolding the corrected differential feature map into a classification feature vector according to a row vector or a column vector; the full-connection coding unit is used for carrying out full-connection coding on the classification characteristic vectors by using a plurality of full-connection layers of the classifier so as to obtain coded classification characteristic vectors; and the classification unit is used for passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.
Here, it will be understood by those skilled in the art that the specific functions and operations of the respective units and modules in the hollow shaft intelligent preparation system for a new energy vehicle 100 described above have been described in detail in the above description of the hollow shaft intelligent preparation system for a new energy vehicle with reference to fig. 1 to 7, and thus, repetitive descriptions thereof will be omitted.
The basic principles of the present application have been described above in connection with specific embodiments, however, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are merely examples and not intended to be limiting, and these advantages, benefits, effects, etc. are not to be considered as essential to the various embodiments of the present application. Furthermore, the specific details disclosed herein are for purposes of illustration and understanding only, and are not intended to be limiting, as the application is not necessarily limited to practice with the above described specific details.
The block diagrams of the devices, apparatuses, devices, systems referred to in the present application are only illustrative examples and are not intended to require or imply that the connections, arrangements, configurations must be made in the manner shown in the block diagrams. As will be appreciated by one of skill in the art, the devices, apparatuses, devices, systems may be connected, arranged, configured in any manner. Words such as "including," "comprising," "having," and the like are words of openness and mean "including but not limited to," and are used interchangeably therewith. The terms "or" and "as used herein refer to and are used interchangeably with the term" and/or "unless the context clearly indicates otherwise. The term "such as" as used herein refers to, and is used interchangeably with, the phrase "such as, but not limited to.
It is also noted that in the apparatus, devices and methods of the present application, the components or steps may be disassembled and/or assembled. Such decomposition and/or recombination should be considered as equivalent aspects of the present application.
The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
The foregoing description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit embodiments of the application to the form disclosed herein. Although a number of example aspects and embodiments have been discussed above, a person of ordinary skill in the art will recognize certain variations, modifications, alterations, additions, and subcombinations thereof.
Claims (10)
1. The intelligent hollow shaft preparation method for the new energy automobile is characterized by comprising the following steps of:
Acquiring a detection sound detection signal of a hollow shaft to be detected;
acquiring a reference sound detection signal, wherein the reference sound detection signal is a sound detection signal of a hollow shaft with molding quality meeting preset requirements;
the detection sound detection signal and the reference sound detection signal are respectively passed through a sound noise reducer based on an automatic coder to obtain a noise-reduced detection sound detection signal and a noise-reduced reference sound detection signal;
passing the denoised detection sound detection signal and the denoised reference sound detection signal through a dual detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound feature map and a reference sound feature map, wherein the first sound waveform encoder and the second sound waveform encoder have the same network structure;
calculating a difference feature map between the detected sound feature map and the reference sound feature map;
performing feature distribution correction on the differential feature map to obtain a corrected differential feature map; and
and the corrected differential feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the molding quality of the hollow shaft to be detected meets the preset requirement.
2. The method for intelligent production of hollow shafts for new energy vehicles according to claim 1, wherein the automatic codec comprises a sound feature extractor and a sound feature decoder.
3. The method for intelligent preparation of hollow axle for new energy automobile according to claim 2, wherein the passing the detection sound detection signal and the reference sound detection signal through the automatic codec-based sound noise reducer to obtain a post-noise-reduction detection sound detection signal and a post-noise-reduction reference sound detection signal, respectively, comprises:
the sound feature extractor performs explicit spatial coding on the detected sound detection signal and the reference sound detection signal by using a convolution layer to obtain a detected sound feature and a reference sound feature; and
and the sound feature decoder uses a deconvolution layer to respectively carry out deconvolution processing on the detected sound feature and the reference sound feature so as to obtain the noise-reduced detected sound detection signal and the noise-reduced reference sound detection signal.
4. The method for intelligent preparation of hollow shafts for new energy vehicles according to claim 3, wherein the first acoustic waveform encoder and the second acoustic waveform encoder are depth residual network models.
5. The method for intelligent preparation of a hollow shaft for a new energy automobile according to claim 4, wherein the calculating a difference feature map between the detected sound feature map and the reference sound feature map comprises: calculating a difference feature map between the detected sound feature map and the reference sound feature map using the following formula;
wherein, the formula is:
wherein F is d Representing the differential feature map, F 1 Representing the detected sound feature map, F 2 The reference sound characteristic map is represented and,representing per-position subtraction.
6. The intelligent preparation method of the hollow shaft for the new energy automobile according to claim 5, wherein the performing feature distribution correction on the differential feature map to obtain a corrected differential feature map comprises:
performing feature map expansion on the detected sound feature map and the reference sound feature map to obtain a detected sound feature vector and a reference sound feature vector;
performing vector-based Hilbert spatial constraint on the detected sound feature vector and the reference sound feature vector to obtain a corrected feature vector; and
and correcting the difference feature map position-by-position feature values based on the correction feature vector to obtain the corrected difference feature map.
7. The method for intelligent preparation of hollow shafts for new energy vehicles according to claim 6, wherein said performing a hilbert space constraint on the vector-based on the detected sound feature vector and the reference sound feature vector to obtain a corrected feature vector comprises:
performing vector-based Hilbert spatial constraint on the detected sound feature vector and the reference sound feature vector by the following formula to obtain the correction feature vector;
wherein, the formula is:
wherein V is 1 And V 2 Representing the detected sound feature vector and the reference sound feature vector, respectively, |·|| 2 Representing the two norms of the vector, cov 1D Expressed in terms of convolution operator (||V) 1 || 2 ,||V 2 || 2 ,V 1 V 2 T ) Vector pairOne-dimensional convolution is carried out, alpha and beta are weight super parameters, V c Representing the correction feature vector.
8. The method for intelligently preparing the hollow shaft for the new energy automobile according to claim 7, wherein the correcting the differential feature map based on the correction feature vector to obtain the corrected differential feature map comprises the following steps:
performing dimension reconstruction on the correction feature vector to obtain a correction feature map; and
And multiplying the difference feature map by position points by taking the correction feature map as a weighted feature map to obtain the corrected difference feature map.
9. The intelligent preparation method of the hollow shaft for the new energy automobile according to claim 8, wherein the step of passing the corrected differential feature map through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the molding quality of the hollow shaft to be detected meets a predetermined requirement or not, and the method comprises the following steps:
expanding the corrected differential feature map into classification feature vectors according to row vectors or column vectors;
performing full-connection coding on the classification feature vectors by using a plurality of full-connection layers of the classifier to obtain coded classification feature vectors; and
and the coding classification feature vector is passed through a Softmax classification function of the classifier to obtain the classification result.
10. A hollow shaft intelligent preparation system for new energy automobile, characterized by comprising:
the detection sound signal acquisition module is used for acquiring a detection sound detection signal of the hollow shaft to be detected;
the reference sound signal acquisition module is used for acquiring a reference sound detection signal, wherein the reference sound detection signal is a sound detection signal of a hollow shaft with molding quality meeting preset requirements;
The noise reduction module is used for respectively passing the detection sound detection signal and the reference sound detection signal through a sound noise reducer based on an automatic coder and decoder to obtain a post-noise reduction detection sound detection signal and a post-noise reduction reference sound detection signal;
the double detection module is used for enabling the noise-reduced detection sound detection signal and the noise-reduced reference sound detection signal to pass through a double detection model comprising a first sound waveform encoder and a second sound waveform encoder to obtain a detection sound characteristic diagram and a reference sound characteristic diagram, and the first sound waveform encoder and the second sound waveform encoder have the same network structure;
the difference feature calculation module is used for calculating a difference feature map between the detected sound feature map and the reference sound feature map;
the characteristic distribution correction module is used for carrying out characteristic distribution correction on the differential characteristic map so as to obtain a corrected differential characteristic map; and
and the molding quality generation module is used for enabling the corrected differential feature map to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the molding quality of the hollow shaft to be detected meets the preset requirement.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310054345.4A CN116858943A (en) | 2023-02-03 | 2023-02-03 | Hollow shaft intelligent preparation method and system for new energy automobile |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310054345.4A CN116858943A (en) | 2023-02-03 | 2023-02-03 | Hollow shaft intelligent preparation method and system for new energy automobile |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116858943A true CN116858943A (en) | 2023-10-10 |
Family
ID=88222213
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310054345.4A Pending CN116858943A (en) | 2023-02-03 | 2023-02-03 | Hollow shaft intelligent preparation method and system for new energy automobile |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116858943A (en) |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104937408A (en) * | 2013-02-01 | 2015-09-23 | 新日铁住金株式会社 | Flaw inspection method and flaw inspection device |
| CN105067704A (en) * | 2015-08-11 | 2015-11-18 | 北京新联铁科技股份有限公司 | Flaw detection system for hollow axle and flaw detection method |
| CN105911143A (en) * | 2016-07-08 | 2016-08-31 | 河南农业大学 | Wall hollowing detection method and device based on acoustic method |
| CN110031552A (en) * | 2019-05-27 | 2019-07-19 | 嘉兴博感科技有限公司 | A kind of monitoring structural health conditions damage characteristic value calculating method |
| CN111033223A (en) * | 2017-08-04 | 2020-04-17 | Bp北美公司 | Ultrasonic corrosion monitoring |
| CN111141823A (en) * | 2020-01-13 | 2020-05-12 | 石河子大学 | Hami melon maturity rapid detection method based on sound signals of smart phone |
| CN111223157A (en) * | 2019-12-27 | 2020-06-02 | 苏州二向箔科技有限公司 | Ultrasonic CT sound velocity imaging method based on depth residual error network |
| CN111931420A (en) * | 2020-08-07 | 2020-11-13 | 合肥工业大学 | Gas turbine fault prediction method based on nuclear regeneration Hilbert space |
| CN112858482A (en) * | 2021-01-14 | 2021-05-28 | 北京主导时代科技有限公司 | Automatic ultrasonic wound judging method and system for hollow shaft |
| KR20210064018A (en) * | 2019-11-25 | 2021-06-02 | 광주과학기술원 | Acoustic event detection method based on deep learning |
| CN113537145A (en) * | 2021-06-28 | 2021-10-22 | 青鸟消防股份有限公司 | Method, device and storage medium for rapidly solving false detection and missed detection in target detection |
| CN114002334A (en) * | 2021-09-29 | 2022-02-01 | 西安交通大学 | Structural damage acoustic emission signal identification method and device and storage medium |
| CN114324580A (en) * | 2021-12-03 | 2022-04-12 | 西安交通大学 | Intelligent knocking detection method and system for structural defects |
| CN114778706A (en) * | 2022-05-07 | 2022-07-22 | 厦门大学 | Indoor object echo characteristic processing method and system based on acoustic-electromagnetic intermodulation |
| CN115015394A (en) * | 2022-07-11 | 2022-09-06 | 西安交通大学 | Composite material defect ultrasonic detection method based on convolution network and trajectory tracking |
| WO2022234957A1 (en) * | 2021-05-03 | 2022-11-10 | 한국표준과학연구원 | Non-destructive ultrasonic testing method and system using deep learning, and autoencoder-based prediction model learning method used therein |
| CN115471498A (en) * | 2022-10-10 | 2022-12-13 | 温州市华炜鞋材科技有限公司 | Multi-angle waterproof monitoring shoe making machine and method for rain shoe production |
| CN115512166A (en) * | 2022-10-18 | 2022-12-23 | 湖北华鑫光电有限公司 | Intelligent preparation method and system of lens |
-
2023
- 2023-02-03 CN CN202310054345.4A patent/CN116858943A/en active Pending
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104937408A (en) * | 2013-02-01 | 2015-09-23 | 新日铁住金株式会社 | Flaw inspection method and flaw inspection device |
| CN105067704A (en) * | 2015-08-11 | 2015-11-18 | 北京新联铁科技股份有限公司 | Flaw detection system for hollow axle and flaw detection method |
| CN105911143A (en) * | 2016-07-08 | 2016-08-31 | 河南农业大学 | Wall hollowing detection method and device based on acoustic method |
| CN111033223A (en) * | 2017-08-04 | 2020-04-17 | Bp北美公司 | Ultrasonic corrosion monitoring |
| CN110031552A (en) * | 2019-05-27 | 2019-07-19 | 嘉兴博感科技有限公司 | A kind of monitoring structural health conditions damage characteristic value calculating method |
| KR20210064018A (en) * | 2019-11-25 | 2021-06-02 | 광주과학기술원 | Acoustic event detection method based on deep learning |
| CN111223157A (en) * | 2019-12-27 | 2020-06-02 | 苏州二向箔科技有限公司 | Ultrasonic CT sound velocity imaging method based on depth residual error network |
| CN111141823A (en) * | 2020-01-13 | 2020-05-12 | 石河子大学 | Hami melon maturity rapid detection method based on sound signals of smart phone |
| CN111931420A (en) * | 2020-08-07 | 2020-11-13 | 合肥工业大学 | Gas turbine fault prediction method based on nuclear regeneration Hilbert space |
| CN112858482A (en) * | 2021-01-14 | 2021-05-28 | 北京主导时代科技有限公司 | Automatic ultrasonic wound judging method and system for hollow shaft |
| WO2022234957A1 (en) * | 2021-05-03 | 2022-11-10 | 한국표준과학연구원 | Non-destructive ultrasonic testing method and system using deep learning, and autoencoder-based prediction model learning method used therein |
| CN113537145A (en) * | 2021-06-28 | 2021-10-22 | 青鸟消防股份有限公司 | Method, device and storage medium for rapidly solving false detection and missed detection in target detection |
| CN114002334A (en) * | 2021-09-29 | 2022-02-01 | 西安交通大学 | Structural damage acoustic emission signal identification method and device and storage medium |
| CN114324580A (en) * | 2021-12-03 | 2022-04-12 | 西安交通大学 | Intelligent knocking detection method and system for structural defects |
| CN114778706A (en) * | 2022-05-07 | 2022-07-22 | 厦门大学 | Indoor object echo characteristic processing method and system based on acoustic-electromagnetic intermodulation |
| CN115015394A (en) * | 2022-07-11 | 2022-09-06 | 西安交通大学 | Composite material defect ultrasonic detection method based on convolution network and trajectory tracking |
| CN115471498A (en) * | 2022-10-10 | 2022-12-13 | 温州市华炜鞋材科技有限公司 | Multi-angle waterproof monitoring shoe making machine and method for rain shoe production |
| CN115512166A (en) * | 2022-10-18 | 2022-12-23 | 湖北华鑫光电有限公司 | Intelligent preparation method and system of lens |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110322423B (en) | Multi-modal image target detection method based on image fusion | |
| CN109035142B (en) | Satellite image super-resolution method combining countermeasure network with aerial image prior | |
| CN113111760B (en) | Light-weight graph convolution human skeleton action recognition method based on channel attention | |
| CN113113030B (en) | High-dimensional damaged data wireless transmission method based on noise reduction self-encoder | |
| CN107633486A (en) | Structure Magnetic Resonance Image Denoising based on three-dimensional full convolutional neural networks | |
| CN109410917A (en) | Voice data classification method based on modified capsule network | |
| CN115235612B (en) | Intelligent fault diagnosis system and method for servo motor | |
| CN112084934B (en) | Behavior recognition method based on bone data double-channel depth separable convolution | |
| CN104899921A (en) | Single-view video human body posture recovery method based on multi-mode self-coding model | |
| CN112149645A (en) | Human body posture key point identification method based on generation of confrontation learning and graph neural network | |
| CN116151270A (en) | Parking test system and method | |
| CN116052254A (en) | Visual continuous emotion recognition method based on extended Kalman filtering neural network | |
| CN113269277B (en) | Continuous dimension emotion recognition method based on transducer encoder and multi-head multi-mode attention | |
| CN116858943A (en) | Hollow shaft intelligent preparation method and system for new energy automobile | |
| CN113705645B (en) | Self-adaptive joint model semi-supervised learning classification method for electroencephalogram signals | |
| CN112257817B (en) | Geological geology online semantic recognition method and device and electronic equipment | |
| CN106683049A (en) | Reconstruction method of the image super-resolution based on the saliency map and the sparse representation | |
| CN112232129A (en) | Electromagnetic information leakage signal simulation system and method based on generation countermeasure network | |
| CN116110422B (en) | Omnidirectional cascade microphone array noise reduction method and system | |
| CN115723280B (en) | Polyimide film production equipment with adjustable thickness | |
| CN117036901A (en) | Small sample fine adjustment method based on visual self-attention model | |
| CN117454102A (en) | Self-adaptive noise elimination method and device for river-crossing pipeline positioning detection system based on FPGA | |
| CN114630207B (en) | Multi-sensing-node sensing data collection method based on noise reduction self-encoder | |
| CN116418633A (en) | Depth expansion underwater sound channel estimation method based on sparse and low-rank characteristics | |
| CN116958701A (en) | Network abnormal flow detection method based on improved VGG16 and image enhancement |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |