TWI841092B - Refrigerator monitoring system and monitoring host - Google Patents

Abstract

A refrigerator monitoring system includes at least one refrigerator device and a monitoring host. The refrigerator device includes a temperature sensor, a high pressure sensor and a low pressure sensor. The monitoring host stores a database and includes a monitoring analysis module and an abnormality judgment module. The database stores at least one sensing data, and the sensing data has a temperature value, a high pressure value and a low pressure value. The monitoring analysis module can generate an evaluation value based on at least one of the temperature value, the high pressure value and the low pressure value using a neural network algorithm, and compare the evaluation value with a comparison threshold value according to a comparison rule. If the comparison rule is not passed, the abnormality judgment module will analyze whether the sensing data meets an abnormality judgment condition to determine whether the refrigerator equipment is in an abnormal state.

Description

Refrigerator monitoring system and monitoring host

The present invention relates to a monitoring system and a monitoring host, and in particular to a refrigerator monitoring system and a monitoring host.

Currently, businesses that sell all kinds of fresh food usually install various types of refrigerator equipment to provide a proper refrigerated storage environment for different food characteristics and preservation methods. Since low temperature environment is the key factor in preserving fresh food, the internal temperature of various refrigerator equipment is usually continuously monitored to ensure the normal function of the refrigerator equipment. However, when adjusting the operating status of the refrigerator equipment (such as defrosting, strong refrigeration), directly monitoring the refrigerator equipment based on factors such as the internal temperature may lead to misjudgment of whether the function is normal, affecting the overall operation of the system.

Therefore, one of the objects of the present invention is to provide a refrigerator monitoring system that can solve the above-mentioned problems.

Therefore, in some embodiments, the refrigerator monitoring system of the present invention includes at least one refrigerator device and a monitoring host. The refrigerator device includes a temperature sensor for detecting the operating temperature, a high pressure sensor for detecting the condensation end pressure, and a low pressure sensor for detecting the evaporation end pressure. The monitoring host stores a database and includes a monitoring analysis module and an abnormality judgment module. The database stores at least one sensing data, and the sensing data has a temperature value generated by the temperature sensor, a high pressure value generated by the high pressure sensor, and a low pressure value generated by the low pressure sensor. The monitoring and analysis module can read the sensing data of the database, generate an evaluation value according to at least one of the temperature value, the high pressure value, and the low pressure value using a pre-trained neural network algorithm, and compare the evaluation value with a preset comparison threshold according to a preset comparison rule; if the comparison between the evaluation value and the comparison threshold is If the result does not pass the comparison rule, the abnormality determination module will analyze whether at least one of the temperature value, the high pressure value, and the low pressure value of the sensing data that does not pass the comparison rule meets a preset abnormality determination condition; if the abnormality determination module determines that the sensing data meets the abnormality determination condition, it will determine that the refrigerator equipment is in an abnormal state.

In some implementations, the abnormal judgment condition is that the temperature value of the sensing data is higher than a temperature threshold and lasts for more than a time threshold, the high pressure value of the sensing data is higher than a high pressure threshold and lasts for more than a time threshold, and/or the low pressure value of the sensing data is higher than a low pressure threshold and lasts for more than a time threshold.

In some implementations, the comparison rule is that the evaluation value is greater than the comparison threshold.

In some implementations, the neural network algorithm of the monitoring and analysis module includes an auto-encoder model, and the auto-encoder model is trained with a plurality of sensing data that are not determined to be abnormal.

In some implementations, the neural network algorithm of the monitoring and analysis module includes an automatic encoder model, which has an encoder that can receive the sensor data that has been preprocessed, and a decoder that generates the evaluation value based on the output result of the encoder. The encoder and the decoder each have 3 to 5 neural layer layers, each neural layer of the encoder has 64 to 128 neurons, and each neural layer of the decoder has 64 to 128 neurons.

In some implementations, the data preprocessing of the sensing data is to scale the temperature value, the high pressure value, and the low pressure value to a range of 0 to 1 and remove null values.

In some implementations, the neural network algorithm processes data with a batch size of 1 to 32 and an iteration number of 100 to 200.

Another object of the present invention is to provide a monitoring host.

Therefore, in some embodiments, the monitoring host of the present invention is suitable for working with at least one refrigerator device, the refrigerator device includes a temperature sensor for detecting operating temperature, a high pressure sensor for detecting condensation end pressure, and a low pressure sensor for detecting evaporation end pressure, and the monitoring host includes a database, a monitoring analysis module, and an abnormality determination module. The database stores at least one sensing data, the sensing data having a temperature value generated by the temperature sensor, a high pressure value generated by the high pressure sensor, and a low pressure value generated by the low pressure sensor. The monitoring and analysis module can read the sensing data of the database, and generate an evaluation value according to at least one of the temperature value, the high pressure value, and the low pressure value using a pre-trained neural network algorithm, and can compare the evaluation value with a preset comparison threshold according to a preset comparison rule. When the comparison result of the evaluation value and the comparison threshold value does not pass the comparison rule, the abnormality determination module will analyze whether at least one of the temperature value, the high pressure value, and the low pressure value of the sensing data that does not pass the comparison rule meets a preset abnormality determination condition; if it is determined that the sensing data meets the abnormality determination condition, the refrigerator equipment is determined to be in an abnormal state.

The present invention has at least the following effects: the refrigerator monitoring system can use the monitoring analysis module of the monitoring host to run the main monitoring program of the refrigerator equipment through the neural network algorithm. When the monitoring analysis module preliminarily determines that the refrigerator equipment is abnormal through the analysis of the evaluation value, the abnormal judgment module can further confirm whether the refrigerator equipment is indeed abnormal by setting the abnormal judgment condition. Through the cooperation of the monitoring analysis module and the abnormal judgment module, the real-time operation status monitoring of the refrigerator equipment can be achieved with excellent accuracy and operation efficiency.

1 to 4 are an embodiment of a refrigerator monitoring system 100 of the present invention. The refrigerator monitoring system 100 includes at least one refrigerator device 1 and a monitoring host 2. In this embodiment, two refrigerator devices 1 are used as an example, but the number of refrigerator devices 1 is not limited to a specific number in practice.

The refrigerator equipment 1 is suitable for being installed in a store where fresh food needs to be preserved, and includes a temperature sensor 11 for detecting an operating temperature, a high-pressure sensor 12 for detecting a condensation end pressure, and a low-pressure sensor 13 for detecting an evaporation end pressure, as well as components such as a cabinet body, a cabinet door, a compressor, a condenser, a capillary tube, an evaporator, and an air supply fan that are not specifically shown in the figure. During operation, the data generated by the temperature sensors 11, the high pressure sensors 12, and the low pressure sensors 13 are collected and aggregated by a management host (not shown) installed in a merchant's store, and then the management host transmits the various data to the monitoring host 2.

The monitoring host 2 is, for example, a server system constructed by multiple server hosts, and can perform real-time equipment status monitoring on the refrigerator equipment 1 according to the data generated by the temperature sensors 11, the high pressure sensors 12, and the low pressure sensors 13. The monitoring host 2 includes a database 21 stored in a storage unit such as a hard disk (not shown), a monitoring analysis module 22, and an abnormality determination module 23.

The database 21 stores at least one (usually multiple) sensing data, each of which has a temperature value generated by the temperature sensor 11, a high pressure value generated by the high pressure sensor, and a low pressure value generated by the low pressure sensor 13. According to the continuous monitoring process of the temperature sensor 11, the high pressure sensor 12, and the low pressure sensor 13, the temperature value, the high pressure value, and the low pressure value can form a data type with a specific time sequence relationship, such as FIG. 4.

The monitoring and analysis module 22 is operated in the form of a software module, and can read the sensing data of the database 21, and analyze the sensing data by a pre-trained neural network algorithm. The abnormality determination module 23 is also operated in the form of a software module, and can further determine whether the refrigerator equipment 1 is in a normal state or an abnormal state according to the analysis result of the monitoring and analysis module 22. This part of the content will be explained later.

Referring to the flow chart and related diagrams of FIG. 2 , the following describes the process of the refrigerator monitoring system 100 performing real-time equipment status monitoring on the refrigerator equipment 1 by the monitoring host 2 .

In steps S01 and S02, the monitoring and analysis module 22 can read the sensing data of the database 21, and generate an evaluation value according to at least one of the temperature values, the high pressure values, and the low pressure values using a pre-trained neural network algorithm, and compare the evaluation value with a preset comparison threshold according to a preset comparison rule. If the comparison result passes the comparison rule, it means that the refrigerator equipment 1 is currently in a normal state, and the process of generating the evaluation value and comparing it with the comparison rule in step S01 will be continuously and regularly executed as shown in the A1 interval of Figure 3. If the comparison result does not pass the comparison rule, it means that the refrigerator equipment 1 is likely to be in an abnormal state, and the subsequent procedures of S03 need to be executed.

Table 1: Example of sensing data Temperature value ( ℃) High pressure value (MPa) Low pressure value (MPa) -5 0.375 0.083 -5 0.371 0.086

In this embodiment, the neural network algorithm of the monitoring and analysis module 22 includes an autoencoder model, and the autoencoder model (autoencoder) is trained in an unsupervised manner using a plurality of sensing data that are not judged as abnormal. In terms of model structure, the autoencoder model has an encoder that can receive the sensing data that has been pre-processed, and a decoder that generates the evaluation value based on the output result of the encoder. After the neural network algorithm is trained, the output result of the encoder (the evaluation value) will be close to the data received by the encoder (the sensing data). Since the present embodiment uses the sensing data in the normal state as shown in Table 1 to train the automatic encoder model, if the sensing data received by the encoder of the monitoring and analysis module 22 in steps S01 and S02 are generated by the refrigerator equipment 1 in the abnormal state, the evaluation value output by the decoder will fail to pass the comparison rule.

For example, in this embodiment, the evaluation value generated by the sensing data in an abnormal state is shown in the A2 section of FIG. 3, which is smaller than the evaluation value generated by the sensing data in a normal state in the A1 section. Therefore, in this embodiment, the comparison rule used in steps S01 and S02 can be set to, for example, "the evaluation value is greater than the comparison threshold." In this way, if the comparison rule is not passed, it means that the evaluation value generated by the current sensing data is too small, and there is a greater possibility that the sensing data received by the monitoring and analysis module 22 is generated by the refrigerator equipment 1 in an abnormal state, and thus subsequent analysis and confirmation are required.

Furthermore, regarding the encoder and the decoder of the aforementioned automatic encoder model, the present embodiment is configured such that each preferably has 3 to 5 neural layers, each neural layer of the encoder has 64 to 128 neurons, and each neural layer of the decoder has 64 to 128 neurons. By training the neural network algorithm and performing actual monitoring and analysis according to this specification, good analysis accuracy can be obtained with the cooperation of a sufficient number of neural layers and neurons, and the overfitting problem of high accuracy for training data but poor accuracy for new data will not occur due to excessive numbers of neural layers and neurons. In addition, before the sensing data is input into the monitoring and analysis module 22, it is preferred to perform data preprocessing by scaling the temperature value, the high pressure value, and the low pressure value to a range of 0 to 1 and removing null values, so as to accelerate the convergence speed of the neural network algorithm and improve the analysis efficiency. Furthermore, the neural network algorithm preferably processes data with a batch size of 1 to 32 and an iteration number of 100 to 200 times, so as to achieve a better balance between analysis and processing efficiency and hardware performance load.

In steps S03-S06, if the comparison result of the evaluation value and the comparison threshold value by the monitoring and analysis module 22 does not pass the comparison rule, the abnormality determination module 23 will further analyze whether at least one of the temperature value, the high pressure value, and the low pressure value of the sensing data that does not pass the comparison rule meets a preset abnormality determination condition in step S03. If the abnormality determination module determines that the sensing data meets the abnormality determination condition, it will determine that the refrigerator device 1 is in an abnormal state as in step S06. Otherwise, it will determine that the refrigerator device 1 is in an abnormal state as in step S05. Subsequently, the above results can be presented to the management personnel for reference in the form of text, images, sounds, etc., so as to facilitate early maintenance and detection of the abnormal refrigerator equipment 1.

For example, the abnormal judgment condition may be that the temperature value of the sensing data is higher than a temperature threshold and lasts for more than a time threshold, the high pressure value of the sensing data is higher than a high pressure threshold and lasts for more than a time threshold, and/or the low pressure value of the sensing data is higher than a low pressure threshold and lasts for more than a time threshold. Taking the operation of the refrigerator equipment 1 in the defrosting state as an example, the temperature threshold in this state may be set to 12°C and the time threshold may be set to 50 minutes. Therefore, taking Figures 3 and 4 as examples, in the A1 section of Figure 3 and the B1 section of Figure 4, the evaluation value calculated by the monitoring and analysis module 22 is greater than the comparison threshold, and the temperature value is not higher than the temperature threshold. At this time, it can be determined that the refrigerator equipment 1 is in a normal state. On the other hand, in the A2 section of Figure 3 and the B2 section of Figure 4, the evaluation value calculated by the monitoring and analysis module 22 is less than the comparison threshold, and the subsequent analysis by the abnormality determination module 23 shows that the temperature value has exceeded 12°C (the temperature threshold) for more than 50 minutes (the time threshold), then the dual calculation and analysis of the monitoring and analysis module 22 and the abnormality determination module 23 can be used to determine that the refrigerator equipment 1 has an abnormality.

In summary of the above description, the refrigerator monitoring system 100 of the present invention can run the main monitoring program of the refrigerator devices 1 by the monitoring analysis module 22 of the monitoring host 2 through the neural network algorithm. When the monitoring analysis module 22 preliminarily determines that the refrigerator devices 1 are abnormal through the analysis of the evaluation value, the abnormality judgment module 23 can further confirm whether the refrigerator devices 1 are indeed abnormal by setting the abnormality judgment condition. By the cooperation of the monitoring and analysis module 22 and the abnormality determination module 23, the real-time operation status monitoring of the refrigerator equipment 1 can be realized with extremely good accuracy and operation efficiency, so the purpose of the present invention can be achieved.

However, the above is only an example of the implementation of the present invention, and it cannot be used to limit the scope of the implementation of the present invention. All simple equivalent changes and modifications made according to the scope of the patent application of the present invention and the content of the patent specification are still within the scope of the patent of the present invention.

100: Refrigerator Monitoring System

1: Refrigerator equipment

11: Temperature sensor

12: High pressure sensor

13: Low pressure sensor

2: Monitor host

21: Database

22: Monitoring and analysis module

23: Abnormal Judgment Module

S01~S06: Steps

Other features and effects of the present invention will be clearly presented in the implementation method of the reference drawings, in which: Figure 1 is a system diagram illustrating an embodiment of the refrigerator monitoring system of the present invention; Figure 2 is an operation flow chart of the refrigerator monitoring system; Figures 3 and 4 are curve diagrams illustrating the state changes of the evaluation value and temperature value during the operation of the refrigerator monitoring system.

100: Refrigerator monitoring system

1: Refrigerator equipment

11: Temperature sensor

12: High pressure sensor

13: Low pressure sensor

2: Monitor host

21: Database

22: Monitoring and analysis module

23: Abnormal determination module

Claims (7)

A refrigerator monitoring system comprises: at least one refrigerator device, including a temperature sensor for detecting operating temperature, a high pressure sensor for detecting condensation end pressure, and a low pressure sensor for detecting evaporation end pressure; and a monitoring host computer, which stores a database and includes a monitoring analysis module and an abnormality judgment module. The database stores at least one sensing data. The sensing data The temperature sensor has a temperature value, the high pressure value, and the low pressure value, wherein the monitoring and analysis module can read the sensing data of the database, generate an evaluation value according to at least one of the temperature value, the high pressure value, and the low pressure value using a pre-trained neural network algorithm, and generate an evaluation value according to a pre-trained neural network algorithm. The evaluation value is compared with a preset comparison threshold according to a comparison rule set; if the comparison result of the evaluation value and the comparison threshold does not pass the comparison rule, the abnormal judgment module analyzes whether at least one of the temperature value, the high pressure value, and the low pressure value of the sensing data that does not pass the comparison rule meets a preset abnormal judgment condition, and the abnormal judgment condition is that the temperature value of the sensing data The temperature is higher than a temperature threshold and lasts longer than a time threshold, the high pressure value of the sensing data is higher than a high pressure threshold and lasts longer than a time threshold, and/or the low pressure value of the sensing data is higher than a low pressure threshold and lasts longer than a time threshold; if the abnormality determination module determines that the sensing data meets the abnormality determination condition, it will determine that the refrigerator equipment is in an abnormal state. A refrigerator monitoring system as described in claim 1, wherein the comparison rule is that the evaluation value is greater than the comparison threshold. A refrigerator monitoring system as described in claim 1, wherein the neural network algorithm of the monitoring and analysis module includes an auto-encoder model, and the auto-encoder model is trained with a plurality of sensing data that are not judged as abnormal. The refrigerator monitoring system as described in claim 1, wherein the neural network algorithm of the monitoring analysis module includes an automatic encoder model, the automatic encoder model has an encoder capable of receiving the sensor data that has been pre-processed, and a decoder that generates the evaluation value based on the output result of the encoder, the encoder and the decoder each have 3 to 5 neural layers, each neural layer of the encoder has 64 to 128 neurons, and each neural layer of the decoder has 64 to 128 neurons. A refrigerator monitoring system as described in claim 4, wherein the data preprocessing of the sensing data is to scale the temperature value, the high pressure value, and the low pressure value to a range of 0 to 1 and remove null values. A refrigerator monitoring system as described in claim 4, wherein the neural network algorithm processes data with a batch size of 1 to 32 and an iteration number of 100 to 200. A monitoring host is suitable for working with at least one refrigerator device. The refrigerator device includes a temperature sensor for detecting an operating temperature, a high pressure sensor for detecting a condensation end pressure, and a low pressure sensor for detecting an evaporation end pressure. The monitoring host includes: a database for storing at least one sensing data having a sensed data generated by the temperature sensor; a temperature value, a high pressure value generated by the high pressure sensor, and a low pressure value generated by the low pressure sensor; a monitoring and analysis module, which can read the sensing data of the database, and generate an evaluation value according to at least one of the temperature value, the high pressure value, and the low pressure value with a pre-trained neural network algorithm, and can also generate an evaluation value according to a preset comparison rule. The evaluation value is compared with a preset comparison threshold; and an abnormal judgment module, when the comparison result of the evaluation value and the comparison threshold does not pass the comparison rule, analyzes whether at least one of the temperature value, the high pressure value, and the low pressure value of the sensing data that does not pass the comparison rule meets a preset abnormal judgment condition, and the abnormal judgment condition is that the temperature of the sensing data The temperature value is higher than a temperature threshold and lasts for more than a time threshold, the high pressure value of the sensing data is higher than a high pressure threshold and lasts for more than a time threshold, and/or the low pressure value of the sensing data is higher than a low pressure threshold and lasts for more than a time threshold; if it is determined that the sensing data meets the abnormal judgment condition, the refrigerator equipment is determined to be in an abnormal state.

Images