JP5953779B2 - Failure prediction system, failure prediction device, and program - Google Patents

Description

  The present invention relates to a failure prediction system, a failure prediction device, and a program.

  For example, in an image forming apparatus that forms and outputs an image on a recording material such as paper, if a failure (including an abnormality, a failure, and a failure) that hinders the operation occurs, inconvenience is caused to the user of the image forming apparatus. It will be. Therefore, the occurrence of such a failure is predicted, and by making it possible to perform necessary measures such as parts replacement and repair immediately before the occurrence of the failure or after the failure has occurred, the use of the image forming apparatus can be improved. It would be desirable to reduce the time for a limited state.

Here, various inventions have been proposed so far regarding failure prediction for an apparatus such as an image forming apparatus.
For example, Patent Document 1 receives a plurality of types of state data from an image forming apparatus and accumulates them in a state database, and generates a plurality of types of target data for abnormality sign determination based on the plurality of types of state data. Determine whether multiple types of target data are less than or equal to each reference value set for each type of data, and add the weight set for each status data to the determination result for each type of status data, An invention for determining the presence or absence of an abnormality sign as a whole of several types of state data is disclosed.

  For example, in Patent Document 2, a causal relationship model that expresses a causal relationship between a maintenance work type and an alarm and an operation event related to the maintenance work type by a graph network structure based on the acquisition of the maintenance work data is created. An invention for updating an abnormality diagnosis model by synthesizing a relation model is disclosed.

JP 2009-037141 A JP 2010-122847 A

  The present invention relates to failure prediction for an apparatus such as an image forming apparatus, and an object thereof is to propose a technique capable of improving the failure prediction accuracy.

The present invention according to claim 1 uses an acquisition unit that acquires a parameter indicating an internal state of a target device, a specifying unit that specifies a usage status of the target device, and a parameter acquired by the acquisition unit. and predicting means for predicting the occurrence of a failure in the apparatus of the subject, on the basis of the usage specified by the specifying means, and adjusting means for adjusting the sensitivity of the parameters used to Oite prediction to the prediction means, a Failure prediction system characterized by having

The present invention according to claim 2 is acquired by an acquisition unit that acquires a parameter indicating an internal state of a target device, a specifying unit that specifies a usage status of the target device, a plurality of prediction algorithms, and the acquisition unit. Predicting means for predicting the occurrence of a failure in the target device using the parameters, and adjusting means for adjusting the degree of contribution of each prediction algorithm in the predicting means based on the usage status specified by the specifying means; A failure prediction system comprising:

The present invention according to claim 3 uses an acquisition unit that acquires a parameter indicating an internal state of the target device, a specifying unit that specifies a usage status of the target device, and a parameter acquired by the acquisition unit. A predicting unit that predicts the occurrence of a failure in the target device; and an adjusting unit that adjusts sensitivity of a parameter used for prediction in the predicting unit based on a use situation specified by the specifying unit. This is a failure prediction apparatus characterized by the above.

The present invention according to claim 4 is acquired by an acquisition unit that acquires a parameter indicating an internal state of the target device, a specifying unit that specifies a usage status of the target device, a plurality of prediction algorithms, and the acquisition unit. Predicting means for predicting the occurrence of a failure in the target device using the parameters, and adjusting means for adjusting the degree of contribution of each prediction algorithm in the predicting means based on the usage status specified by the specifying means; And a failure predicting device characterized by comprising:

The present invention according to claim 5 uses an acquisition function for acquiring a parameter indicating an internal state of a target device, a specific function for specifying a usage status of the target device, and a parameter acquired by the acquisition function. A prediction function for predicting the occurrence of a failure in the target device, and an adjustment function for adjusting sensitivity of a parameter used for prediction in the prediction function based on a use situation specified by the specific function. It is a program for realizing .

The present invention according to claim 6 is acquired by an acquisition function for acquiring a parameter indicating an internal state of a target apparatus, a specific function for specifying a usage status of the target apparatus, a plurality of prediction algorithms, and the acquisition function. A prediction function that predicts the occurrence of a failure in the target device using the parameters, and an adjustment function that adjusts the degree of contribution of each prediction algorithm in the prediction function based on the usage situation specified by the specific function; , Is a program for causing a computer to realize.

According to the first, third, and fifth aspects of the present invention, it is possible to adjust the sensitivity of a parameter used for predicting a failure according to a difference between devices due to a difference in device usage status (maintenance status or load status). This can improve the accuracy of predicting a failure.

According to the second, fourth, and sixth aspects of the present invention, the degree of contribution of each prediction algorithm used for predicting a failure is adjusted according to the difference between devices due to a difference in device usage status (maintenance status or load status). This can improve the accuracy of predicting a failure.

It is a figure which shows the example of the functional block of the server apparatus in the failure prediction system which concerns on one Embodiment of this invention. It is a figure which shows the example of the some linear regression (secondary) prediction curve obtained by sliding the area of object. It is a figure which shows the example of the change point detection by a data trend. It is a figure which shows the example of the change point detection by dispersion | distribution (standard deviation value). It is a figure which shows the example of the change point detection by the relationship (correlation coefficient parameter | index) of time series data. It is a figure which shows the example of the change point detection by Mahalanobis distance. It is a figure which shows the example of the functional block of a time series data change point detection part and a failure sign determination part. It is a figure which shows the example of the weighting value (coefficient) of each change point detection algorithm set for every kind of failure. FIG. 6 is a diagram illustrating an example of a correspondence table between a maintenance status of an image forming apparatus and an adjustment index. FIG. 6 is a diagram illustrating an example of a correspondence table in which parameter adjustment modes are set for each type of usage status of the image forming apparatus. FIG. 10 is a diagram illustrating another example of a correspondence table between the maintenance status of the image forming apparatus and the adjustment index. FIG. 4 is a diagram illustrating an example of a correspondence table between a load status of an image forming apparatus and an adjustment index. It is a figure which shows the example of the occurrence prediction of the failure using a causal network model. It is a figure which shows the example of the conditional probability table | surface which a monitoring parameter node has in a causal network model.

A failure prediction system according to an embodiment of the present invention will be described with reference to the drawings.
In the failure prediction system of this example, a server device (an example of a failure prediction device according to the present invention) connected to the image forming apparatus so as to be able to communicate with a wired or wireless communication is based on information collected from the image forming apparatus to be monitored. The occurrence of a failure in the image forming apparatus is predicted.

First, an image forming apparatus to be monitored will be described.
The image forming apparatus is an apparatus having an image forming function for forming and outputting an image on a recording material such as paper. Examples of the image forming apparatus include apparatuses such as a printer (document printing apparatus), a copier (document copying apparatus), a facsimile apparatus (document transfer apparatus), and a multi-function apparatus having a combination of functions of these apparatuses. included.

  Here, the image forming apparatus of the present example has a function of detecting values of a plurality of monitoring parameters indicating its internal state. This parameter is predetermined as a monitoring parameter that can contribute to the prediction of failure. For example, the photosensitive member potential, the photosensitive member charging current, the amount of semiconductor laser light, the developing device toner concentration, the primary transfer current, and the secondary transfer. There are various image forming parameters detected at any time during the operation of the image forming process such as current, fixing unit heat roll temperature, and procompat density. As the detection value of the image formation parameter, a measurement value measured at a part corresponding to the image formation parameter may be used, or a target value (set value) set as a target for controlling each part may be used. Other types of values such as the difference between the measured value and the target value may be used. Further, the image forming apparatus of this example also detects environmental information such as temperature and humidity as monitoring parameters.

  In the image forming apparatus of this example, when an execution command for a series of processes (jobs) for forming an image relating to one page or a plurality of pages on each recording material is received, an image is formed on the recording material according to the execution command and output. The value of each monitoring parameter is detected every time (every page), and after completion of all the image forming processes related to the job, operation state data storing the detected values of each monitoring parameter is transmitted to the server apparatus. As the operation state data, for example, data having a structure in which a device ID for identifying the device itself, a detection value of each monitoring parameter, a detection date and time thereof, and the like are stored is used.

Further, the image forming apparatus of this example has a function of acquiring usage status information indicating the usage status. The usage state information is used in the server device for adjusting the sensitivity and narrowing down of each monitoring parameter used for failure prediction, adjusting the contribution of each change point detection algorithm (an example of a prediction algorithm) in failure prediction, and the like.
As usage status information, for example, image formation such as the cleaning date / time and replacement date / time of each part related to the operation of the image forming process, the adjustment date / time and adjustment amount of each part, and the date / time of initial setup (for example, calibration) Information indicating the maintenance status of the apparatus can be given. Further, for example, information indicating the load status of the image forming apparatus, such as a print counter value after job execution (by color / monochrome), a job coverage average value, and a variance value, is also included in the usage status information.

  In the image forming apparatus of this example, when maintenance is performed, information on the maintenance status is received (or automatically detected) from the operator and acquired. In addition, load status information is automatically detected and acquired every time a job is executed. Along with the transmission of the operation status data, the device usage data storing the usage status information indicating the maintenance status and the load status is transmitted to the server device. As the device usage data, for example, data having a structure in which a device ID for identifying the device itself, information on maintenance status and usage status, acquisition date and time, and the like are stored is used.

  Here, instead of the configuration in which the operation state data and the device usage data are transmitted to the server device at the end of the image forming process based on the job command as described above, these data are temporarily stored in the memory. Alternatively, a configuration may be adopted in which the accumulated untransmitted data is transmitted when a predetermined transmission condition is satisfied. Specifically, for example, the passage of a predetermined time may be used as a transmission condition, and data may be transmitted every time the time passes (for example, every hour), or a request from a server device may be used as a transmission condition. Data may be sent in response to the request. Further, the operation state data and the device usage data may be transmitted based on different transmission conditions.

Next, the server device will be described.
In the server device of this example, the occurrence of a failure in the image forming apparatus is roughly predicted based on the operation state data and device usage data collected (received) from the image forming apparatus to be monitored. Here, the prediction of the failure occurrence is performed for each image forming apparatus by identifying each image forming apparatus by the apparatus ID included in the operation state data and the device usage data.
As shown in FIG. 1, the server apparatus of this example includes an image formation parameter collection unit 1, a time-series data storage unit 2, a device usage data collection unit 3, a device usage data storage unit 4, an adjustment index, as shown in FIG. A calculation unit 5, a time-series data change point detection unit 6, a failure sign determination unit 7, and a failure sign notification unit 8 are included.

The image forming parameter collection unit 1 acquires (receives) operation state data from the image forming apparatus to be monitored.
The time-series data storage unit 2 statistically processes the detection values in predetermined units (for example, job units or daily units) for each monitoring parameter based on the operation state data acquired by the image formation parameter collection unit 1. The results are time-series and accumulated (stored) as time-series data for a predetermined period (for example, the past three months). In the statistical processing, for example, an average value and a variance value are calculated.

The device usage data collection unit 3 acquires (receives) device usage data from the image forming apparatus to be monitored.
The device usage data storage unit 4 stores (saves) the device usage data acquired by the device usage data collection unit 3 for a predetermined period (for example, the past three months).
Of the usage status information included in the device usage data, information (such as parts cleaning date / time and replacement date / time) for receiving input from the operator is not received by the image forming apparatus, but is received by the server device (or server). The input may be received by another operation terminal connected to the apparatus via a network, and this may be collected by the device usage data collection unit 3 and stored in the device usage data storage unit 4.

The adjustment index calculation unit 5 changes the configuration and parameters of a time-series data change point detection unit 6 and a failure sign determination unit 7 described later based on the device usage data stored in the device usage data storage unit 4. An adjustment index is calculated.
In this example, for the image forming apparatus to be monitored, maintenance frequency (part cleaning frequency, part replacement frequency, etc.), page print amount, color / monochrome ratio, which is information necessary for calculating the adjustment index from the corresponding device usage data, Coverage statistics and the like are calculated, and an adjustment index is calculated from these calculated values.
A specific example of the adjustment index will be described later.

  The time series data change point detection unit 6 detects a change point at which the transition tendency of the time series data changes based on the time series data for each monitoring parameter accumulated in the time series data accumulation unit 2. In this example, two or more change point detection algorithms are used to detect change points for each monitoring parameter and each change point detection algorithm. At that time, as will be described later, based on the adjustment index calculated by the adjustment index calculation unit 5, the change point detection algorithm is selected, the parameter is changed, and the like.

  The failure sign determination unit 7 predicts the occurrence of a failure in the image forming apparatus to be monitored based on the change points for each monitoring parameter and each change point detection algorithm detected by the time-series data change point detection unit 6. In this example, the occurrence of a failure is predicted using two or more predictive judgment algorithms. At that time, as will be described later, based on the adjustment index calculated by the adjustment index calculation unit 5, the sign determination algorithm is selected, the weight (contribution) is changed, and the threshold is changed.

  The failure sign notification unit 8 notifies the user of the failure prediction system of the prediction result from the failure sign determination unit 7. The notification method is arbitrary. For example, the configuration is such that the information of the prediction result is transmitted to the image forming apparatus to be monitored and displayed on the operation panel provided in the image forming apparatus, or the terminal of the remote center is addressed. For example, the information of the prediction result is transmitted, and the image forming apparatus to be monitored is identified and displayed on the display panel provided in the terminal.

The time series data change point detector 6 will be further described.
The time series data change point detection unit 6 of this example detects a plurality of change points from various viewpoints such as data trend, variance (standard deviation value), time series data relationship (correlation coefficient index), Mahalanobis distance, and the like. The change point detection algorithm is used.

In the change point detection algorithm based on the data trend, processing is performed as illustrated in FIG. 2 based on the time series data of the target monitoring parameter (for example, the time series data of the average value of each job related to a certain monitoring parameter). A linear regression (secondary) prediction curve representing the transition of time series data is calculated for each section while sliding the section. Then, as illustrated in FIG. 3, a point at which the change tendency of the quadratic coefficient “a” in each linear regression (secondary) prediction curve “y = ax ² + bx + c” continues for a long period is detected as a change point. In the example of FIG. 3, the image quality trouble (an example of a failure) due to density unevenness has occurred a plurality of times, and for some image quality troubles, a change point is detected within 5 days from the occurrence. It can be seen that detection of this is useful for predicting failure.

  In the change point detection algorithm based on variance (standard deviation value), the standard deviation value is large based on the time series data of the target monitoring parameter (for example, the time series data of the standard deviation value for each job related to a certain monitoring parameter). A changing point is detected as a changing point. That is, as illustrated in FIGS. 4A to 4D, a point where the standard deviation value crosses the threshold is detected as a change point. As the threshold value, for example, a value optimized by matching with a case of a failure that has occurred in the past is used.

  The change point detection algorithm based on the relationship of time series data (correlation coefficient index) is a combination of two monitoring parameters that change in a relevant manner during normal times (no failure occurs). Every time, the transition of the relationship (correlation coefficient index) of each time series data is obtained, and the point at which the relationship is lost is detected as a change point. That is, as illustrated in FIGS. 5A to 5D, a point where the correlation coefficient index crosses the threshold is detected as a change point. As the threshold value, for example, a value optimized by matching with a case of a failure that has occurred in the past is used.

  In the change point detection algorithm based on Mahalanobis distance, based on the time series data of the target multivariate (for example, 4) monitoring parameters, the time series data at the normal time (the state where no failure has occurred) is used in advance. The Mahalanobis distance with respect to the created reference space is calculated for each time point, and a point deviating from the reference space is detected as a change point. That is, as illustrated in FIGS. 6A to 6D, a point where the Mahalanobis distance exceeds the threshold is detected as a change point.

Next, adjustment of the prediction mode (the processing contents of the time-series data change point detection unit 6 and the failure sign determination unit 7) based on the adjustment index calculated by the adjustment index calculation unit 5 will be described.
FIG. 7 shows an example of functional blocks of the time-series data change point detection unit 6 and the failure sign determination unit 7.

  Based on the image forming parameter group 11 (time-series data of each monitoring parameter) accumulated in the time-series data accumulation unit 2, the time-series data change point detection unit 6 uses a plurality of (in this example, ## The change points are detected by using n) change point detection algorithms 13 of 1 to #n. As the monitoring parameter to be a change point detection target, different parameters can be used according to the characteristics of each change point detection algorithm 13. In addition, a parameter weighting unit 12 for setting a weight for each monitoring parameter is provided in the preceding stage of the change point detection algorithm 13, and a weighting value corresponding to the adjustment index calculated by the adjustment index calculation unit 5 is provided. The sensitivity of each monitoring parameter in the change point detection algorithm 13 is adjusted by (coefficient).

  The failure sign determination unit 7 includes a plurality of total normalization units 14 corresponding to the respective change point detection algorithms 13, an algorithm weight setting unit 15 for setting a weight to each total normalization unit 14, and each total normalization unit 14 has a threshold majority decision unit 16 that performs a majority decision based on the result of No. 14 and performs failure sign judgment (occurrence prediction), and a sign decision threshold setting unit 17 that sets a majority threshold used in the majority decision of the majority decision unit 16.

  The tabulation standardization unit 14 tabulates the number of monitoring parameters whose change points are detected by the corresponding change point detection algorithm 13, normalizes the total number of monitoring parameters targeted by the change point detection algorithm 13, and then executes the algorithm. The weighting value (coefficient) set by the weight setting unit 15 is multiplied and output to the threshold majority decision determining unit 16. That is, the degree of contribution to the failure prediction is adjusted for each of the change point detection algorithms 13. As illustrated in FIG. 8, combinations of weight values (coefficients) for each algorithm are set in advance in a table for each type of failure, and are read out from the table and processed when performing predictive determination.

  The threshold majority decision unit 16 sums up the outputs from the respective change point detection algorithms 13, normalizes them by the total number of change point detection algorithms 13 (n in this example), and then sets them by the sign determination threshold setting unit 17. It is determined whether or not it is a sign of failure by comparing with a majority decision threshold. As the majority threshold, 0.5 is set if the majority decision is based on the majority, and a smaller value (for example, 0.3) is set if the decision criterion is relaxed.

  For example, there are three change point detection algorithms (# 1) to (# 3) as the change point detection algorithm 13, and the total standardization result of the change point detection algorithm (# 1) is 0.6, and the change point detection algorithm (# If the total standardization result of 2) is 0.4 and the total standardization result of the change point detection algorithm (# 3) is 0.2, the total sum of the total standardization results is 1.2, and the change point detection algorithm Normalized by 3 which is the number of 13, becomes 0.4. If the majority decision threshold is 0.3, this result is equal to or greater than the majority decision threshold, and therefore, it is determined as a sign of failure. In this way, in this example, the contribution of each change point detection algorithm in the failure prediction is adjusted in consideration of the maintenance frequency of the image forming apparatus.

  The determination result by the threshold majority determination unit 16 is output to the failure sign notification unit 8 by the determination result output unit 18. Specifically, for example, a date or the like determined (predicted) as a failure sign is output to the failure sign notification unit 8. Also, it is an output of another mode such as an output of a flag indicating whether or not it is a sign of failure (for example, 1 if it is a sign, 0 if it is not a sign), or a score obtained by quantifying the determination result. May be.

Next, the adjustment index calculation unit 5 will be described.
As an example, the adjustment index calculation unit 5 calculates an average value of the maintenance frequency per page printing based on the maintenance frequency and the page printing amount of the image forming apparatus every time the sign determination is performed for the image forming apparatus to be monitored. . That is, for example, based on the cumulative print counter number PV for the past month obtained from the history of print counter values and the part cleaning frequency Mc for the past month obtained from the history of parts cleaning date and time, ) To calculate the part cleaning frequency Nc per page printing.
Nc = Mc / PV (Formula 1)
In addition, an average value Nc (ave) of parts cleaning frequency per page printing is calculated in advance for the same model as the image forming apparatus, and Nc and Nc (ave) are compared to obtain NcRatio (= Nc / Nc (ave). )) Is calculated. NcRatio is an example of an index representing the maintenance status of the image forming apparatus, and an adjustment index is calculated by referring to the correspondence table illustrated in FIG. 9 using this as a key.

  The adjustment index calculated by the adjustment index calculation unit 5 is given to the time-series data change point detection unit 6 and used for weight setting of each monitoring parameter input to each change point detection algorithm 13. In this example, the larger the value of NcRatio, the more frequently the maintenance (part cleaning) is performed. As a result, the change in the detected value of the monitoring parameter corresponding to the part becomes smaller. Thus, the accuracy of the failure prediction is increased by adjusting the sensitivity of the monitoring parameter so as to increase the sensitivity.

In this example, the part cleaning frequency has been described as an example. However, for example, the adjustment index may be calculated from the part replacement frequency obtained using the part replacement date, and the part cleaning frequency and the part replacement frequency may be calculated. The adjustment index may be calculated in combination, and the adjustment index can be calculated using various types of maintenance frequencies calculated based on the device usage data.
In the above description, the adjustment index calculated from the maintenance frequency of the image forming apparatus to be monitored is applied to the parameter weighting unit 12 of the time-series data change point detection unit 6. The output of each change point detection algorithm 13 may be adjusted so as to be applied to the attachment setting unit 15 and to be an optimal combination of the change point detection algorithms 13.

It is also possible to calculate an adjustment index using information other than the maintenance frequency and adjust the sensitivity of each monitoring parameter.
FIG. 10 shows an example of a correspondence table in which parameter adjustment modes are set for each type of usage status of the image forming apparatus. In the example of FIG. 10, there are settings related to the maintenance status (calibration frequency, cleaning frequency, parts replacement frequency) and settings related to the load status (color ratio, coverage).

The setting part related to the maintenance status will be described.
In the sensitivity adjustment based on the calibration frequency, when the calibration frequency is high, the change in all the monitoring parameters becomes small. Therefore, the sensitivity is adjusted so as to increase the sensitivity of all the monitoring parameters.
In the sensitivity adjustment based on the cleaning frequency, when the cleaning frequency is high, the change in the monitoring parameter of the portion where the cleaning has been performed becomes small. Therefore, adjustment is performed so as to increase the sensitivity of the monitoring parameter only in the cleaning portion.
In the sensitivity adjustment based on the part replacement frequency, when the part replacement frequency is high, the change in the monitoring parameter at the part where the part is replaced becomes small. Therefore, the sensitivity is adjusted to increase the sensitivity of the monitoring parameter only in the replacement part.

The setting part related to the load situation will be described.
In sensitivity adjustment based on the color ratio (color: black and white), adjustment is performed in units of the ratio of the monitoring parameters related to color (M, Y, C) and the monitoring parameters related to black (K), and the color ratio is high. In this case, adjustment is made so that the sensitivity ratio of the color-related monitoring parameter is lowered (the sensitivity of the black-related monitoring parameter is increased).
In sensitivity adjustment based on coverage, when printing with high coverage is performed, toner-related monitoring parameters fluctuate greatly regardless of the signs of image quality problems. By adjusting to lower, the accuracy of failure prediction is improved as a whole.

Next, another aspect of adjustment related to the monitoring parameter will be described.
FIG. 11 shows another example of the correspondence table between the maintenance status of the image forming apparatus and the adjustment index.
In the example of FIG. 11, the adjustment index is set to 0 when the maintenance frequency of the image forming apparatus is high. This is because, in a device with a high maintenance frequency, although the fluctuation of the monitored parameter to be monitored is small, the monitoring parameter that hardly changes is operated so as to be excluded from the monitoring target by setting the adjustment index to 0. It is specified. That is, the amount of variance such as standard deviation is calculated for the time series data, and the adjustment index 0 is applied (excluded from the monitoring target) to the monitoring parameter, assuming that the change is small when the value is less than the specified value. In this way, only effective monitoring parameters with changes can be monitored, so that the accuracy of failure prediction can be improved.

Such exclusion of the monitoring parameter may be applied when the sensitivity is unnecessarily increased by adjusting the sensitivity of the monitoring parameter and the S / N ratio (signal-to-noise ratio) is decreased.
Here, the criterion for adjusting the adjustment index to 0 is used as the standard deviation threshold. However, when the ratio between the maximum value and the minimum value of the adjustment value is greater than or equal to the reference value (for example, 10 times), the monitoring parameter has high sensitivity The adjustment index may be set to 0.

Next, an example in which the adjustment index calculation unit 5 calculates the adjustment index from the viewpoint of the load state of the image forming apparatus will be described.
In the adjustment index calculation unit 5, for the monitoring target image forming apparatus, for example, based on the cumulative print counter number PVc for color printing in the past month and the cumulative print counter number PVbw for black and white print in the past month, The color / monochrome ratio CB is calculated by (Equation 2). The larger the value of CB, the higher the color printing ratio.
CB = PVc / PVbw (Formula 2)
In addition, an average value CB (ave) of the color / monochrome ratio is calculated in advance for the same model as the image forming apparatus, and CB and CB (ave) are compared with each other to obtain CB Ratio (= CB / CB (ave)). calculate. CBRatio is an example of an index representing the load status of the image forming apparatus, and by referring to the correspondence table illustrated in FIG. 12 using this as a key, an adjustment index for each color (by MYCK) is calculated.

The adjustment index for each color calculated by the adjustment index calculation unit 5 is given to the time-series data change point detection unit 6 and used for weight setting of each monitoring parameter for each color input to each change point detection algorithm 13. In this example, the larger the value of CBRatio, the more frequently color printing is performed. As a result, the change in the detected value of the color-related monitoring parameter becomes smaller. The sensitivity of the change point detection of the color-related monitoring parameter is increased by increasing the sensitivity of the color-related monitoring parameter and decreasing the sensitivity of the black-related monitoring parameter.
As described above, in this example, the sensitivity of the monitoring parameter is adjusted in consideration of the load state of the image forming apparatus. Note that the load status of the image forming apparatus may be taken into consideration in narrowing down the monitoring parameters and adjusting the contribution of each change point detection algorithm.

Next, another configuration example of the failure sign determination unit 7 will be described.
In this example, predictive failure determination (occurrence prediction) is performed using a causal network model such as a Bayesian network as illustrated in FIG.
In the example of FIG. 13, a monitoring parameter node (monitoring parameters 1 to 4) is associated with the sign determination node, and an adjustment index node (adjustment index 1) is further associated. The sign determination node has a probability table in which the failure probability density distribution is discretized. In this example, the degree of discreteness is 4, and there are four states of state0 to state3.

The monitoring parameter node corresponds to the monitoring parameter change point detection state in each change point detection algorithm 13 of the time-series data change point detection unit 6, and “Yes” or “No” is determined depending on whether or not the change point is detected. 2 states.
The adjustment index node is an evidence information setting node corresponding to the value calculated by the adjustment index calculation unit 5. For example, “Yes” is set when the maintenance frequency is high, and “No” is set when the maintenance frequency is low.
FIG. 14 shows an example of a conditional probability table possessed by the monitoring parameter node of the monitoring parameter 4. In this example, since the sign determination node having four states becomes a parent node, a four-state conditional probability table is used as shown in FIG.

An operation when a Bayesian network is used will be described.
The change point detection algorithm 13 of the time-series data change point detection unit 6 outputs the date of the detected change point of each monitoring parameter and the value of each monitoring parameter, which are input to the Bayesian network.

  Then, with respect to the monitoring parameter node corresponding to each monitoring parameter, the state (state 0 to state 3) into which the input monitoring parameter value is classified is examined, and the corresponding state is set as evidence information. This is performed for the monitoring parameter nodes of all the monitoring parameters. In addition, a state corresponding to the adjustment index calculated by the adjustment index calculation unit 5 is also set.

  After that, an operation for propagating evidence information on the network is performed, and a posteriori determination is executed by calculating a posteriori probability at the predictor determination node, and the result is used as a predictor determination result to predict a failure through the failure predictor notification unit 8. Notify system users. The calculated value of the posterior probability is compared with the threshold value, and if it is equal to or greater than the threshold value, “Trouble sign is detected” is displayed, and if it is less than the threshold value, “Trouble sign is not detected” is displayed. May be.

  In the above example, the adjustment index node corresponding to the adjustment index is provided in the Bayesian network. However, a plurality of Bayesian networks are prepared in advance as separate models for each adjustment index, and are calculated by the adjustment index calculation unit 5. The model to be used may be switched according to the adjustment index.

  As described above, in the server device of this example, the image forming parameter collection unit 1 collects a plurality of monitoring parameters indicating the internal state of the image forming device to be monitored and stores it in the time series data storage unit 2. The time series data change point detection unit 6 detects a change point at which the change tendency of the time series data of each monitoring parameter changes using a plurality of change point detection algorithms, and the failure sign determination unit 7 detects each change point. The occurrence of a failure is predicted based on the change point detected by the algorithm.

At that time, the device usage data collection unit 3 collects device usage data indicating the usage status (maintenance status and load status) of the image forming apparatus to be monitored, accumulates it in the device usage data storage unit 4, and adjusts the adjustment index. The calculation unit 5 calculates an adjustment index based on the device usage data, and outputs the adjustment index to the time-series data change point detection unit 6 and the failure sign determination unit 7, thereby adjusting the failure occurrence prediction mode.
More specifically, based on the adjustment index calculated by the adjustment index calculation unit 5, adjustment of the sensitivity of the monitoring parameter used by each change point detection algorithm for change point detection, and each change point detection algorithm for change inspection Narrow down the monitoring parameters used for output, and adjust the contribution of each change point detection algorithm in predicting the occurrence of failures.

  As described above, in the server device of this example, the failure occurrence prediction is performed in consideration of the usage status (maintenance status and load status) of the image forming apparatus to be monitored. Can be effectively increased.

  Here, the server device in the failure prediction system of the present example is a CPU (Central Processing Unit) that performs various arithmetic processing, a RAM (Random Access Memory) that is a work area of the CPU, a ROM that records a basic control program, and the like ( Main storage device such as Read Only Memory), auxiliary storage device such as HDD (Hard Disk Drive) that stores various programs and data, display device for displaying and outputting various information, and used for input operations by the operator Consists of a computer having hardware resources such as an input / output I / F that is an interface with an input device such as an operation button or a touch panel, and a communication I / F that is an interface for performing wired or wireless communication with other devices. Been Yes.

Then, the program according to the present invention is read from the auxiliary storage device or the like, loaded into the RAM, and executed by the CPU, thereby realizing the function of the failure prediction device according to the present invention on the computer of the server device.
In this example, the function of the acquisition unit according to the present invention is realized by the image forming parameter collection unit 1, and the function of the specification unit according to the present invention is realized by the device usage data collection unit 3 and the adjustment index calculation unit 5. The function of the prediction unit according to the present invention is realized by the time series data change point detection unit 6 and the failure sign determination unit 7, and the function of the adjustment unit according to the present invention is realized by the adjustment index calculation unit 5 and the time series data change point detection unit 6. This is realized by the failure sign determination unit 7.
Here, these functional units may be provided in the image forming apparatus, and each image forming apparatus may perform its own failure prediction.

  In addition, an apparatus other than the image forming apparatus may be the target of failure prediction. That is, in the configuration in which the occurrence prediction of the failure in the target device is performed using the parameter indicating the internal state of the device (a parameter that can contribute to the prediction of the failure), the usage status of the device is specified, and the usage status of the device By adjusting the mode of failure occurrence prediction in consideration of (maintenance status and load status), the accuracy of failure occurrence prediction in the device can be effectively increased.

Further, the program according to the present invention is set in the computer according to the present example by a format read from an external storage medium such as a CD-ROM storing the program or a format received via a communication network or the like. .
Note that the present invention is not limited to a mode in which each functional unit is realized by a software configuration as in this example, and each functional unit may be realized by a dedicated hardware module.

1: Image formation parameter collection unit, 2: Time series data storage unit, 3: Device usage data collection unit, 4: Device usage data storage unit, 5: Adjustment index calculation unit, 6: Time series data change point detection unit, 7 : Failure sign determination unit, failure sign notification unit 8,
11: Image formation parameter group, 12: Parameter weighting unit, 13: Change point detection algorithm, Totalization normalization unit 14, Algorithm weighting setting unit 15, Threshold majority decision determination unit 16, Predictive determination threshold setting unit 17, Determination result Output unit 18

Claims (6)

Acquisition means for acquiring a parameter indicating an internal state of the target device;
Identifying means for identifying the usage status of the target device;
Predicting means for predicting the occurrence of a failure in the target device, using the parameters acquired by the acquiring means;
Based on the usage specified by the specifying means, and adjusting means for adjusting the sensitivity of the parameters used to Oite prediction to the prediction means,
A failure prediction system comprising:   Acquisition means for acquiring a parameter indicating an internal state of the target device;
  Identifying means for identifying the usage status of the target device;
  Prediction means for predicting the occurrence of a failure in the target device using a plurality of prediction algorithms and parameters acquired by the acquisition means;
  An adjustment unit that adjusts the degree of contribution of each prediction algorithm in the prediction unit based on the usage situation specified by the specifying unit;
  A failure prediction system comprising:   Acquisition means for acquiring a parameter indicating an internal state of the target device;
  Identifying means for identifying the usage status of the target device;
  Predicting means for predicting the occurrence of a failure in the target device, using the parameters acquired by the acquiring means;
  Adjusting means for adjusting the sensitivity of parameters used for prediction in the prediction means based on the use situation specified by the specifying means;
  A failure prediction apparatus comprising:   Acquisition means for acquiring a parameter indicating an internal state of the target device;
  Identifying means for identifying the usage status of the target device;
  Prediction means for predicting the occurrence of a failure in the target device using a plurality of prediction algorithms and parameters acquired by the acquisition means;
  An adjustment unit that adjusts the degree of contribution of each prediction algorithm in the prediction unit based on the usage situation specified by the specifying unit;
  A failure prediction apparatus comprising:   An acquisition function for acquiring a parameter indicating the internal state of the target device;
  A specific function for identifying the usage status of the target device;
  A prediction function that predicts the occurrence of a failure in the target device using the parameters acquired by the acquisition function;
  An adjustment function for adjusting sensitivity of a parameter used for prediction in the prediction function based on the usage situation specified by the specific function;
  A program to make a computer realize. An acquisition function for acquiring a parameter indicating the internal state of the target device;
A specific function for identifying the usage status of the target device;
A prediction function that predicts the occurrence of a failure in the target device using a plurality of prediction algorithms and parameters acquired by the acquisition function;
An adjustment function that adjusts the degree of contribution of each prediction algorithm in the prediction function based on the usage situation specified by the specific function;
A program to make a computer realize.

Images