State determination device and state determination method
The state determination device improves the accuracy of predicting processing equipment failures by generating a unit space and adjusting reference data with Mahalanobis distances, reducing unnecessary stoppages and operational losses.
Patent Information
- Application Number
- JP2021075832
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2021-04-28
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2041-04-28
AI Technical Summary
Existing state determination devices for processing equipment suffer from inaccurate predictions of device abnormalities, leading to unnecessary stoppages and increased operational losses due to incorrect failure determinations.
A state determination device that utilizes a control unit to generate a unit space based on feature quantities of successfully processed lots, calculates Mahalanobis distances, and adjusts reference data by deleting or adding feature quantities based on thresholds to improve accuracy in determining the state of the processing device.
Enhances the accuracy of determining the state of processing equipment, reducing unnecessary stoppages and minimizing operational losses by effectively predicting potential failures.
Smart Images

Figure 0007593218000001 
Figure 0007593218000002 
Figure 0007593218000003
Abstract
Description
[Technical field]
[0001] The present disclosure relates to a state determination device and a state determination method. [Background technology]
[0002] 2. Description of the Related Art Conventionally, there is known a condition determination device that provides an operator of a work machine with useful information for identifying a fault location (see, for example, Patent Document 1). [Prior art documents] [Patent documents]
[0003] [Patent Document 1] JP 2013-199804 A Summary of the Invention [Problem to be solved by the invention]
[0004] When a failure is predicted to occur in the processing device, the state of the processing device is determined to be abnormal. When the state of the processing device is determined to be abnormal, the processing device is stopped and inspected. When the determination is incorrect, an opportunity loss occurs due to the unnecessary stop of the processing device. It is desired to improve the accuracy of determining the state of the processing device so that the frequency of unnecessary stop of the processing device can be reduced.
[0005] Therefore, an object of the present disclosure is to propose a state determination device and a state determination method that can improve the accuracy of determining the state of a processing device. [Means for solving the problem]
[0006] One embodiment of the present disclosure that solves the above problem is as follows. [1] A control unit is provided for determining a state of the processing device, The control unit is acquiring, as reference data, feature quantities of a plurality of lots normally processed by the processing device; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data; calculating a Mahalanobis distance of the feature amount of each lot included in the reference data in the unit space, and deleting the feature amount of the lot whose Mahalanobis distance is equal to or greater than a deletion threshold value from the reference data; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which the state of the processing device is to be judged; Calculating a Mahalanobis distance of the determination target data in the update unit space; a state determination device that outputs a result that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value that is set based on the update unit space; [2] A control unit is provided for determining a state of the processing device, The control unit is acquiring, as reference data, feature quantities of a plurality of lots normally processed by the processing device; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data; calculating, in the unit space, a Mahalanobis distance of a feature quantity of a newly acquired lot to be added to the reference data, and adding, to the reference data, the feature quantity of the lot whose Mahalanobis distance is less than an addition threshold value; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which the state of the processing device is to be judged; Calculating a Mahalanobis distance of the determination target data in the update unit space; a state determination device that outputs a result that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value that is set based on the update unit space; [3] A control unit is provided for determining a state of the processing device, The control unit is acquiring, as reference data, feature quantities of a plurality of lots normally processed by the processing device; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data; calculating a Mahalanobis distance of the feature amount of each lot included in the reference data in the unit space, and deleting the feature amount of the lot whose Mahalanobis distance is equal to or greater than a deletion threshold value from the reference data; calculating, in the unit space, a Mahalanobis distance of a feature quantity of a newly acquired lot to be added to the reference data, and adding, to the reference data, the feature quantity of the lot whose Mahalanobis distance is less than an addition threshold value; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which the state of the processing device is to be judged; Calculating a Mahalanobis distance of the determination target data in the update unit space; a state determination device that outputs a result that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value that is set based on the update unit space; [4] The control unit sets the deletion threshold and the addition threshold so that the number of lots to be excluded from the reference data is equal to the number of lots to be added to the reference data. [5] The control unit adds the feature quantities of a predetermined number or more lots newly acquired to be added to the reference data to the reference data if the Mahalanobis distance of the feature quantities of the predetermined number or more lots is continuously greater than or equal to the addition threshold. [6] A state determination method executed by a state determination device that determines a state of a processing device, comprising: acquiring, as reference data, feature quantities of a plurality of lots normally processed by the processing device; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data; calculating a Mahalanobis distance of feature quantities of each lot included in the reference data in the unit space, and deleting feature quantities of lots whose Mahalanobis distances are equal to or greater than a deletion threshold value from the reference data; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which a state of the processing device is to be judged; calculating a Mahalanobis distance of the data to be determined in the update unit space; a step of outputting a result of determining that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value set based on the update unit space; A state determination method comprising: [7] A state determination method executed by a state determination device that determines a state of a processing device, comprising: acquiring, as reference data, feature quantities of a plurality of lots normally processed by the processing device; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data; a step of calculating, in the unit space, a Mahalanobis distance of a feature quantity of a newly acquired lot to be added to the reference data, and adding, to the reference data, the feature quantity of the lot whose Mahalanobis distance is less than an addition threshold value; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which a state of the processing device is to be judged; calculating a Mahalanobis distance of the data to be determined in the update unit space; a step of outputting a result of determining that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value set based on the update unit space; A state determination method comprising: [8] A state determination method executed by a state determination device that determines a state of a processing device, comprising: acquiring, as reference data, feature quantities of a plurality of lots normally processed by the processing device; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data; calculating a Mahalanobis distance of feature quantities of each lot included in the reference data in the unit space, and deleting feature quantities of lots whose Mahalanobis distances are equal to or greater than a deletion threshold value from the reference data; a step of calculating, in the unit space, a Mahalanobis distance of a feature quantity of a newly acquired lot to be added to the reference data, and adding, to the reference data, the feature quantity of the lot whose Mahalanobis distance is less than an addition threshold value; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which a state of the processing device is to be judged; calculating a Mahalanobis distance of the data to be determined in the update unit space; a step of outputting a result of determining that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value set based on the update unit space; A state determination method comprising: Effect of the Invention
[0007] According to an embodiment of the present disclosure, the accuracy of determining the state of a processing device can be improved. [Brief description of the drawings]
[0008] [Figure 1] 1 is a block diagram showing an example of the configuration of a state determination system according to an embodiment; [Diagram 2] FIG. 2 is a diagram showing a configuration example of a wire saw device as a processing device. [Diagram 3]1 is a table showing an example of data items obtained from a wire saw device. [Figure 4] 13 is a graph showing an example of MD of each lot calculated in a unit space based on cause data items. [Diagram 5] 13 is a graph showing an example of MD of each lot calculated in a unit space according to a comparative example. [Figure 6] 13 is a graph showing an example of MD of feature quantities of lots that are candidates for exclusion from unit space data. [Figure 7] 13 is a graph showing an example of MD of feature quantities of lots that are candidates for addition to unit space data. [Figure 8] 13 is a graph showing an example of an MD value of determination target data calculated based on a unit space before updating. [Figure 9] 13 is a graph showing an example of an MD value of determination target data calculated based on a unit space updated by a method of the comparative example. [Figure 10] 13 is a graph showing an example of an MD value of determination target data calculated based on a unit space updated by the method of this embodiment. [Figure 11] 13 is a graph showing an example of the MD value of the determination target data calculated based on a unit space updated by a method in which exception processing is added to the method of the present embodiment. [Figure 12] 1 is a flowchart illustrating an example of a procedure of a state determination method according to an embodiment. [Figure 13] 13 is a flowchart showing an example of a subroutine process for generating a unit space. [Figure 14] 13 is a flowchart showing an example of a subroutine process for updating a unit space. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0009] (Overview of state determination system 1 according to one embodiment) A status determination system 1 (see FIG. 1) or a status determination device 50 (see FIG. 1) according to an embodiment of the present disclosure determines the status of a processing device 10 (see FIG. 1 or FIG. 2) installed in a factory or the like that manufactures products. The processing device 10 processes received members in at least some of the processes for manufacturing products and delivers them as processed products.
[0010] The processing device 10 may break down and stop during processing. In this case, the processed product that was being processed becomes defective. The occurrence of defective products results in losses for the factory. Furthermore, the man-hours required for recovery work of the processing device 10 are greater than the man-hours required for maintenance work on the processing device 10. Therefore, a breakdown of the processing device 10 during processing increases losses for the factory.
[0011] The state determination system 1 or the state determination device 50 according to an embodiment of the present disclosure determines the possibility of failure of the processing device 10 before the processing device 10 fails during processing, and prompts the stopping or maintenance work of the processing device 10. In this way, losses caused by the processing device 10 can be reduced.
[0012] An example of the configuration of the state determination system 1 and the state determination device 50 according to an embodiment will be described below.
[0013] (System configuration example) 1, a condition determination system 1 according to one embodiment includes a condition determination device 50 and a sensor 60. The sensor 60 measures and outputs various data relating to the condition of the processing device 10.
[0014] The state determination device 50 includes a control unit 51, an output unit 53, and an input unit 54. The control unit 51 provides control and processing capabilities for executing various functions of the state determination device 50. As will be described later, the control unit 51 determines whether the state of the processing device 10 is an abnormal state.
[0015] The control unit 51 may include at least one processor. The processor may execute a program that realizes various functions of the control unit 51. The processor may be realized as a single integrated circuit. The integrated circuit is also called an IC (Integrated Circuit). The processor may be realized as a plurality of communicatively connected integrated circuits and discrete circuits. The processor may be realized based on various other known technologies.
[0016] The control unit 51 may include a storage unit. The storage unit may include an electromagnetic storage medium such as a magnetic disk, or may include a memory such as a semiconductor memory or a magnetic memory. The storage unit may include a non-transitory computer-readable medium. The storage unit stores various information and programs executed by the control unit 51. The storage unit may function as a work memory for the control unit 51. At least a part of the storage unit may be configured as a separate entity from the control unit 51.
[0017] The control unit 51 acquires measurement data from the sensor 60. The control unit 51 may acquire the measurement data from the sensor 60 based on a standard such as RS485. The control unit 51 may be connected to the sensor 60 via a communication interface such as a LAN (Local Area Network). The control unit 51 is not limited to the above example and may be connected to the sensor 60 in various ways.
[0018] The output unit 53 outputs information acquired from the control unit 51. The output unit 53 may output information to notify an operator or a maintenance person of the processing apparatus 10. The output unit 53 may include a display device. The display device may include, for example, a liquid crystal display, an organic EL (Electroluminescence) display, an inorganic EL display, or the like, but is not limited to these, and may include other devices. The output unit 53 may display the information acquired from the control unit 51 on the display device as characters or images, etc., and notify the information to the surroundings.
[0019] The output unit 53 may include a light source such as an LED (Light Emission Diode) or a halogen lamp. The output unit 53 may notify an operator, maintenance personnel, or the like in the vicinity of information by turning on or blinking the light source based on information acquired from the control unit 51. The output unit 53 may include a buzzer such as a piezoelectric buzzer or an electromagnetic buzzer, or a speaker that emits a predetermined sound. The output unit 53 may notify an operator, maintenance personnel, or the like in the vicinity of information by sounding a buzzer or generating a sound from a speaker based on information acquired from the control unit 51.
[0020] The output unit 53 may output information to the processing apparatus 10. The output unit 53 may include a communication module so as to be communicatively connected to the processing apparatus 10.
[0021] The input unit 54 includes an input device for accepting operations or inputs by an operator or a maintenance technician managing the status determination device 50. The input device may include, for example, a keyboard or physical keys, or may include a touch panel or a touch sensor, or a pointing device such as a mouse. When the input device is a touch panel or a touch sensor, it may be integrated with the display of the output unit 53. The input device may include, for example, a microphone that accepts voice input. The input device of the input unit 54 is not limited to these examples, and may include various other devices.
[0022] (Specific example of processing device 10: wire saw device) In this embodiment, it is assumed that the processing device 10 is a wire saw device. As shown in FIG. 2, the processing device 10 as a wire saw device includes a wire group 16 in which a wire 12 is stretched in parallel between a plurality of rollers 14 so as to be capable of reciprocating movement. The processing device 10 includes a workpiece holding mechanism 18 that holds the workpiece W and moves the workpiece W in a direction in which the workpiece W is pushed against the wire group 16. The processing device 10 includes a pair of nozzles 20 that supply slurry to an area of the wire group 16 in which the workpiece W is pushed. The processing device 10 cuts the workpiece W with the wire group 16. It is assumed that the workpiece W is a block of silicon or the like (a single crystal ingot cut into a block shape). The processing device 10 cuts the workpiece W and delivers sliced wafers of silicon or the like as processed products.
[0023] The wire 12 is wound around a pair of wire reels 38A and 38B. The wire 12 is tensioned from one wire reel 38A through the guide roller 32, the roller 14, etc. to the other wire reel 38B.
[0024] The wire reels 38A and 38B are each rotated by a drive motor 36. When the drive motor 36 is driven to rotate the wire reels 38A and 38B, the wire 12 is unwound from one wire reel 38A and can run through the guide roller 32, roller 14, etc. to the other wire reel 38B. The wire 12 runs through a tension applying means including a dancer arm 33, dancer roller 34, etc. Tension is applied to the wire 12 as the wire 12 runs through the tension applying means. The wire 12 runs through a touch roller 35. The touch roller 35 follows the position of the wire 12 as it moves when it is unwound from the wire reels 38A and 38B or wound onto the wire reels 38A and 38B.
[0025] The wire 12 is wound helically multiple times across multiple rollers 14. The helically wound wire 12 constitutes a wire group 16 arranged in parallel between the rollers 14 in a direction perpendicular to the roller axial direction X. The rollers 14 are configured such that polyurethane resin is pressed around a steel cylinder, and grooves are cut into the surface at a fixed pitch. The wire 12 fits into the grooves cut into the surface of the rollers 14, allowing the wire group 16 to run stably.
[0026] The running direction of the wire 12 is controlled by the rotation direction of the drive motor 36. The wire 12 can be controlled to run in one direction, or can be controlled to run back and forth as necessary. The magnitude of tension applied to the wire 12 may be set as appropriate. The running speed of the wire 12 may be set as appropriate.
[0027] The slurry supplied from the nozzle 20 to the wire group 16 is stored in a slurry tank 40, and is sent from the slurry tank 40 to the nozzle 20 via a slurry chiller 42 that adjusts the temperature of the slurry.
[0028] The sliced wafers as processed products discharged from the wire saw device are further processed in steps such as polishing, and are shipped as final wafer products.
[0029] (Example of configuration of state determination device 50) Hereinafter, a configuration example will be described in which the state determination device 50 determines whether the state of the processing device 10 is an abnormal state. In the following description, it is assumed that the processing device 10 is a wire saw device.
[0030] <Abnormal condition of wire saw equipment> The abnormal state of the wire saw device corresponds to a state in which there is a high possibility that the wire saw device will stop during cutting of a block. The wire saw device may stop due to a break in the wire 12. The wire saw device may also stop without the wire 12 being broken.
[0031] Various factors are considered as the cause of the wire saw device coming to a stop. The following factors are considered as known factors. For example, if an abnormality occurs in the supply of slurry, the wire 12 may become more susceptible to breakage. For example, if slurry adheres to a sensor that detects breakage of the wire 12, the wire saw device may erroneously detect a breakage of the wire 12 and stop. For example, the wire saw device may stop due to an abnormality in the amount or temperature of the slurry in the slurry tank 40. On the other hand, the wire saw device may stop due to an unknown factor that does not fit the known factors.
[0032] The control unit 51 of the state determination device 50 acquires measurement values of various data items obtained from the wire saw device. The data items obtained from the wire saw device include, for example, a slurry flow rate, a wire tension, or a cooling water flow rate. The data items obtained from the wire saw device are not limited to these examples and may include various other items.
[0033] The wire saw device executes the process of cutting one block into sliced wafers as one lot process. The control unit 51 can acquire a plurality of measurement values for each data item during one lot process. The wire saw device cuts one block into a plurality of sliced wafers over a predetermined time. That is, a predetermined time is required for one lot process. The control unit 51 may acquire the measurement values of each data item, for example, every 15 seconds during one lot process. Then, the control unit 51 can acquire a plurality of measurement values for each data item during one lot process. The control unit 51 calculates the average value and the standard deviation of the plurality of measurement values acquired during each lot process. The control unit 51 may divide the period from the start to the end of each lot process into the first process period to the nth process period, and calculate the average value and the standard deviation of the plurality of measurement values acquired during each process period. For example, the control unit 51 may set the period during which the cutting distance of the block is from 0 mm to 10 mm during each lot process as the first process period, and calculate the average value and the standard deviation of the plurality of measurement values acquired during the first process period. The average value and the standard deviation of the plurality of measurement values calculated for the first to nth process periods of each lot are also referred to as the first to nth characteristic amounts of the lot, respectively. The first to nth characteristic amounts of each lot are also collectively referred to as the characteristic amounts of each lot. The characteristic amount of the lot calculated based on the measurement value of each data item is simply also referred to as the characteristic amount of the lot of each data item. The control unit 51 may calculate a plurality of characteristic amounts for each data item of each lot as described above, or may calculate one characteristic amount.
[0034] The control unit 51 determines whether the wire saw device is in a state where the probability of stopping is low or in a state where the possibility of stopping is high based on the characteristic amount of the lot of each data item of the wire saw device. The state where the probability of the wire saw device stopping is low is also referred to as the normal state. The state where the possibility of the wire saw device stopping is high is also referred to as the abnormal state.
[0035] <Regarding the MT method (Mahalanobis-Taguchi method)> The state of the wire saw device may be determined each time processing of one lot is performed. The state of the wire saw device may be determined by focusing on only one data item of the wire saw device. In this case, under the assumption that the distribution of the feature amount of the lot of one data item follows a normal distribution, it may be determined whether the feature amount of the lot obtained when processing the lot with the wire saw device is an abnormal value. If the feature amount of a certain lot is determined to be an abnormal value, it may be determined that there is a high possibility that the wire saw device will stop during processing of the next lot. In other words, the state of the wire saw device may be determined to be an abnormal state.
[0036] When the number of data items obtained from the wire saw device is large, the state of the wire saw device may be determined based on the multiple data items. In this case, the distribution of the feature values of the multiple data items for each lot may be considered to follow a multivariate normal distribution. The control unit 51 may determine whether the feature value of the lot for each data item is an abnormal value based on the Mahalanobis-Taguchi method (MT method), which is one of the multivariate analysis techniques, and if the feature value of the lot is an abnormal value, determine that the state of the wire saw device when the lot is processed is in an abnormal state. Specifically, when multiple feature values are calculated for one lot, the control unit 51 may determine that the state of the wire saw device when the lot is processed is in an abnormal state if at least one feature value is an abnormal value.
[0037] The control unit 51 generates a unit space based on the feature quantities of each lot of a plurality of data items obtained when the wire saw device is in a normal state. The unit space is expressed as a mathematical formula that specifies the distribution of the feature quantities of the lot of each data item. Specifically, the control unit 51 normalizes the feature quantities of the lot of each data item, generates a correlation coefficient matrix including correlation coefficients between data items as elements, and calculates the inverse matrix of the correlation coefficient matrix, thereby generating a mathematical formula that expresses the unit space. In other words, the unit space is specified by the inverse matrix of the correlation coefficient matrix.
[0038] The unit space represents the characteristics of the distribution of the feature quantities of a lot processed when the wire saw device is in a normal state. The unit space can also be said to represent the characteristics of a group of lots normally cut by the wire saw device. The unit space can also be said to represent the characteristics of a group of lots normally processed by the processing device 10. The control unit 51 can calculate the Mahalanobis Distance (MD) of data consisting of a set of feature quantities of each of a plurality of data items obtained from the wire saw device, based on a formula representing the unit space. MD is a concept commonly used in the field of quality engineering, and is described, for example, in the following literature. Reference: Makoto Kunigo, "Introduction to Mahalanobis Distance", Journal of the Society of Quality Engineering, Vol. 9, No. 1, pp. 13-19
[0039] MD is an index that indicates how close the distribution of the feature values of a lot processed when the state of the wire saw device is unknown is to the distribution of the feature values of a lot processed when the state of the wire saw device is normal. The smaller the MD calculated from the feature values of a lot, the more likely it is that the feature values of that lot are the feature values of a lot processed when the state of the wire saw device is normal. Conversely, the larger the MD calculated from the feature values of a lot, the more likely it is that the feature values of that lot are the feature values of a lot processed when the state of the wire saw device is abnormal.
[0040] Therefore, the control unit 51 can determine whether the state of the wire saw device when the feature value of a newly processed lot is obtained is normal or abnormal by calculating the MD of the feature value of the newly processed lot based on the unit space. Hereinafter, the data of the wire saw device including the feature value of a lot processed in a predetermined period in the past, which is used to generate the unit space, is also referred to as reference data. Moreover, the data obtained when the wire saw device processes a new lot and used to determine the state of the wire saw device when the lot is being processed is also referred to as determination target data. The control unit 51 calculates the MD of the determination target data in the unit space generated based on the reference data, and determines the state of the wire saw device. The control unit 51 may generate the unit space based on the reference data, or may obtain information about the unit space from an external device. When the control unit 51 calculates multiple feature values for one data item based on data obtained from the wire saw device during the processing of one lot, it calculates the MD of data that has multiple feature values calculated for each of the multiple data items as one set. In other words, the control unit 51 regards the data that has multiple feature values calculated for each of the multiple data items as one set as the reference data and the determination target data. For example, if 30 features are calculated for each of 40 data items in processing one lot, one set of data will contain 1,200 features.
[0041] <Creating unit space based on causal data items> As described above, the wire saw device stops due to various factors. The factors that cause the wire saw device to stop may include factors that are known to a user of the wire saw device and factors that are unknown to the user of the wire saw device. The known factors are classified as known failure modes. The known failure modes are also referred to as known modes.
[0042] When the probability of the wire saw device stopping is high, the feature value of the lot of the data item corresponding to the factor that causes the stoppage may be an abnormal value. The control unit 51 can determine whether the feature value of the lot of the data item corresponding to the factor that causes the wire saw device to stop is an abnormal value, thereby determining whether the possibility of the wire saw device stopping is high. The factor that causes the wire saw device to stop is also referred to as a stoppage factor. The data item corresponding to the stoppage factor is also referred to as a cause data item. There may be multiple factors that cause the wire saw device to stop. Therefore, there is a cause data item corresponding to each of the multiple stoppage factors.
[0043] The control unit 51 may generate a unit space based on the feature amount of the lot of the cause data item in order to determine whether the feature amount of the lot of the cause data item is an abnormal value. The feature amount of the lot of the cause data item includes a plurality of feature amounts of the cause data item acquired from the wire saw device during the processing of the lot. On the other hand, the feature amount of the lot of the cause data item does not include any feature amount of items other than the cause data item acquired from the wire saw device during the processing of the lot. The unit space may be generated by an external device. Specifically, the control unit 51 acquires, as reference data, the measured values of the plurality of data items for each lot processed from the start of operation of the wire saw device until the stop of operation, and calculates the feature amount of the lot based on the measured values. The control unit 51 generates the unit space based on the feature amount of the lot processed when the state of the wire saw device was found to be normal, among the acquired reference data.
[0044] In order to determine the state of the wire saw device, the control unit 51 acquires, as the data to be determined, the measured values of a plurality of data items of a lot newly processed by the wire saw device, and calculates the feature amount of the lot based on the measured values. The control unit 51 calculates, for each stop cause, the MD of the feature amount of the lot of the cause data item among the data acquired as the data to be determined, in a unit space generated in advance based on the feature amount of the lot of the cause data item. The control unit 51 can determine whether the feature amount of the lot of the cause data item is an abnormal value based on the calculated MD. If the feature amount of the lot of the cause data item is an abnormal value, it can be determined that the wire saw device is highly likely to stop during the processing of the next lot. In other words, it can be determined that the state of the wire saw device when the target lot from which the data to be determined was acquired was processed is an abnormal state.
[0045] Here, it is assumed that the unit space is generated based on the feature quantities of the lots of data items other than the cause data items and not corresponding to the stop cause. The data items not corresponding to the stop cause are also called non-cause data items. When the unit space is generated based on not only the feature quantities of the lots of the cause data items but also the feature quantities of the lots of the non-cause data items, the unit space also represents trends that are not related to the stop of the wire saw device. As described above, the unit space is generated based on the correlation of the feature quantities of the lots of each data item. In that case, the unit space generated based on the feature quantities of the lots of the non-cause data items includes relationships with small correlations and is redundant compared to the unit space generated based only on the feature quantities of the lots of the cause data items. The MD calculated in the redundant unit space is insensitive to changes in the feature quantities of the lots of the cause data items due to the influence of the feature quantities of the lots of the non-cause data items. In other words, by generating the unit space based only on the cause data items, the control unit 51 can detect changes in the feature quantities of the lots of the cause data items with high sensitivity by calculating the MD in the unit space.
[0046] The control unit 51 acquires, as the data to be judged, the measured values of each data item of the lot when the wire saw device processes one lot, and calculates the feature amount of the lot based on the measured values. The control unit 51 calculates the MD of the feature amount of the lot of the cause data item included in the data to be judged in the generated unit space. If the calculated MD is less than the threshold, the control unit 51 judges that the state when the wire saw device processed one lot was normal. If the calculated MD is equal to or greater than the threshold, the control unit 51 judges that the state when the wire saw device processed one lot was abnormal.
[0047] <Creating unit space corresponding to known modes> The failure mode that causes the wire saw device to stop may include a known failure mode or an unknown failure mode. Hereinafter, it is assumed that the failure mode that causes the wire saw device to stop includes at least two known modes. The control unit 51 generates a unit space based on the cause data items corresponding to each known mode. The unit space may be generated for each known mode. When the unit space is generated for each known mode, it is associated with each known mode.
[0048] The control unit 51 calculates the MD of the lot feature amount of the cause data item corresponding to each known mode, which is included in the determination target data obtained when the wire saw device processes one lot, in the unit space corresponding to each known mode. When the MD calculated for a certain known mode is equal to or greater than a predetermined threshold, the control unit 51 determines that there is a high possibility that the wire saw device will stop in that known mode when processing the next lot. The predetermined threshold may be set to the same value for each known mode, or may be set to a different value.
[0049] The at least two known modes that cause the wire saw apparatus to stop include a first known mode and a second known mode. The known modes are not limited to the first known mode and the second known mode, and may include, for example, a third known mode. It should be noted that identifiers such as "first" and "second" do not indicate the superiority or inferiority of the known modes. A data item corresponding to a factor that causes the wire saw apparatus to stop in the first known mode is also referred to as a first cause data item. The control unit 51 generates a unit space based on the feature amount of the lot of the first cause data item. The unit space generated based on the feature amount of the lot of the first cause data item is also referred to as a first unit space. A data item corresponding to a factor that causes the wire saw apparatus to stop in the second known mode is also referred to as a second cause data item. The control unit 51 generates a unit space based on the feature amount of the lot of the second cause data item. The unit space generated based on the feature amount of the lot of the second cause data item is also referred to as a second unit space. The cause data items are not limited to the first cause data item and the second cause data item, and may include, for example, a third cause data item corresponding to a cause of stopping in the third known mode. The unit space is not limited to the first unit space and the second unit space, and may include, for example, a third unit space generated based on the feature quantity of the lot of the third cause data item. Furthermore, identifiers such as "first" and "second" do not indicate the superiority or inferiority of the cause data items or unit spaces.
[0050] The control unit 51 calculates, in the first unit space, the MD of the feature amount of the lot of the first cause data item included in the judgment target data obtained when the wire saw device processes one lot. The control unit 51 calculates, in the second unit space, the MD of the feature amount of the lot of the second cause data item included in the judgment target data. The control unit 51 may determine that the wire saw device is more likely to stop in the first known mode when the MD calculated in the first unit space is equal to or greater than a predetermined threshold. The control unit 51 may determine that the wire saw device is more likely to stop in the second known mode when the MD calculated in the second unit space is equal to or greater than a predetermined threshold. When the control unit 51 determines that the wire saw device is more likely to stop in one of the first known mode and the second known mode, it may not be necessary to determine the other mode. Even if it is determined that the wire saw device is more likely to stop in at least one mode, the control unit 51 may determine whether the wire saw device is more likely to stop in another mode. If it is determined that the wire saw device is more likely to stop in a plurality of failure modes, the user can carry out maintenance work collectively for the target failure modes.
[0051] The first cause data item and the second cause data item each include a plurality of data items. Both the first cause data item and the second cause data item include a data item that is not included in the other cause data item. In other words, the first cause data item and the second cause data item are not in an inclusive relationship with each other. Also, the data items included in the first cause data item and the data items included in the second cause data item are not completely identical. If the first cause data item and the second cause data item are in an inclusive relationship or completely identical, the first known mode and the second known mode are not independent failure modes, but are related to each other and of the same type. By both the first cause data item and the second cause data item including a data item that is not included in the other cause data item, the control unit 51 can detect at least two known modes that occur independently of each other with high accuracy.
[0052] <Data used to generate unit space> As described above, the control unit 51 generates a unit space based on the feature quantity of the lot of the cause data item of the lot processed during the period when the state of the wire saw device was normal in the reference data. An example of data included in the reference data will be described below.
[0053] The state of the wire saw device may become unstable in the processing of several to several dozen lots after the start of operation after part replacement work, disassembly, assembly work, cleaning work, etc. of the wire saw device, in which the replaced or assembled parts are not yet operating smoothly. After that, the state of the wire saw device may become stable in the processing of several tens to several hundreds of lots as the parts start to operate smoothly. Furthermore, the state of the wire saw device may become unstable in the processing of several to several tens of lots until the failure, as the probability of failure increases due to deterioration of the cleaning state due to deterioration of the parts over time or accumulation of slurry, etc.
[0054] The control unit 51 acquires the measurement values of each data item in each lot processed from when the wire saw device starts to when it stops operating, and calculates the feature amount of each lot based on the measurement values as reference data. The feature amount of each lot is calculated based on the measurement values of N data items from D1 to DN. In this embodiment, the control unit 51 calculates multiple feature amounts for each data item. The control unit 51 may calculate one feature amount for each data item.
[0055] As illustrated in the table of FIG. 3, the feature quantity of a lot calculated as reference data is associated with a lot and a data item. The columns of the table of FIG. 3 represent N data items from D1 to DN. The rows of the table of FIG. 3 represent lots. The lots are further divided into three periods, P1, P2, and P3, based on the time when they were processed. P1 corresponds to M1 lots processed after the wire saw device resumed operation. P3 corresponds to M3 lots processed immediately before the lot when the wire saw device stopped. P2 corresponds to M2 lots not included in P1 and P3. It can also be said that P2 corresponds to a lot processed while the wire saw device was in a stable state.
[0056] The control unit 51 calculates a plurality of feature amounts for each of the M1 lots from D1P1_1 to D1P1_M1 based on the measurement value of the data item D1 of each of the M1 lots during a period P1. The control unit 51 calculates a plurality of feature amounts for each of the M1 lots from DNP1_1 to DNP1_M1 based on the measurement value of the data item DN of each of the M1 lots during a period P1. The control unit 51 calculates a plurality of feature amounts for each of the M2 lots from D1P2_1 to D1P2_M2 based on the measurement value of the data item D1 of each of the M2 lots during a period P2. The control unit 51 calculates a plurality of feature amounts for each of the M2 lots from DNP2_1 to DNP2_M2 based on the measurement value of the data item DN of each of the M2 lots during a period P2. The control unit 51 calculates a plurality of feature amounts for each of the M3 lots from D1P3_1 to D1P3_M3 based on the measurement value of the data item D1 of each of the M3 lots during a period P3. During a period P3, the control unit 51 calculates a plurality of feature amounts for each of the M3 lots from DNP3_1 to DNP3_M3 based on the measurement values of the data items DN for each of the M3 lots.
[0057] Assume that the wire saw apparatus targeted for acquiring the reference data exemplified in the table of FIG. 3 stopped in a known mode during the processing of the lots after period P3. In other words, the lots in period P3 correspond to the last M3 lots processed by the wire saw apparatus before the wire saw apparatus stops in the known mode.
[0058] The control unit 51 generates a unit space based on the feature amounts of the lots in period P2 during which the state of the wire saw apparatus is stable, among the feature amounts of the lots included in the acquired reference data. Here, assume that the feature amounts of some lots are excluded from the feature amounts of each lot in period P2 used to generate the unit space. If a plurality of feature amounts are calculated for the lot to be excluded, assume that all the feature amounts of that lot are excluded. The reason for excluding the feature amounts of some lots will be described later. The generated unit space is associated with the known mode.
[0059] <State determination of wire saw apparatus based on MD> The control unit 51 acquires the feature amounts of the lots of each data item during the lot processing of the wire saw apparatus as determination target data, and calculates the MD of the feature amounts of the lots acquired as the determination target data in the unit space generated in advance based on the reference data.
[0060] <<Verification of validity of unit space>> Here, the unit space needs to be generated so that the state determination device 50 can determine whether the wire saw apparatus is in a stable state or a state with a high probability of failure based on the MD of the feature amounts of the lots. That is, the unit space needs to be generated such that the MD of the feature amounts of the lots in a state where the wire saw apparatus is likely to stop is equal to or greater than a predetermined threshold, and the MD of the feature amounts of the lots in a stable state of the wire saw apparatus is less than the predetermined threshold.
[0061] In order to confirm whether the generated unit space satisfies the above-mentioned conditions, the MD was calculated for the feature amount of each lot included in the reference data exemplified in FIG. 3. When multiple feature amounts are calculated for each lot, the MD is calculated based on the data in which the multiple feature amounts of each lot are combined into one set. FIG. 4 shows a bar graph in which the MD values calculated for the feature amount of each lot included in the reference data are arranged in the order in which they were processed by the wire saw device. The MD calculated for the feature amount of each lot is also referred to as the MD of each lot. The horizontal axis represents the lot. The vertical axis represents the MD value of each lot. The range of the lots included in each of the periods P1, P2, and P3, which are divided based on the state of the wire saw device, is displayed along the horizontal axis. The number of lots included in the period P1 (M1), the number of lots included in the period P2 (M2), and the number of lots included in the period P3 (M3) are 10 lots, 143 lots, and 40 lots, respectively. The ratio of the number of lots included in the periods P1, P2, and P3, respectively, to the total number of lots is about 5%, about 75%, and about 20%. The ratio of the number of lots included in each period is not limited to this example and may be changed as appropriate.
[0062] The bar graph shown on the far right of FIG. 4 represents the MD of the stopped lot 71 that was being processed when the wire saw device was stopped. In the graph of FIG. 4, the predetermined threshold is represented by MD_T. When the vertical bar representing the MD value of each lot extends above the dashed line representing the value of MD_T, the MD is greater than the predetermined threshold. The control unit 51 sets the value of MD_T so that the MD of each lot included in the period P2 is less than MD_T and the MD of at least one lot included in the period P3 is equal to or greater than MD_T. A lot whose MD is equal to or greater than MD_T is represented as an abnormality detection lot 72. By setting MD_T in this way, the control unit 51 can determine that the state of the wire saw device is abnormal before the wire saw device stops during lot processing and causes a loss.
[0063] The MD of the lot included in the period P2 tends to be smaller than the MD of the lot included in the periods P1 and P3. This is because the unit space is generated based on the feature of the lot included in the period P2. The feature of each lot used to generate the unit space acts to expand the unit space according to the feature of each lot itself. Therefore, when the MD of the feature of the lot used to generate the unit space is calculated in a certain unit space, the calculated MD naturally becomes a small value. Conversely, when the MD of the feature of the lot not used to generate the unit space is calculated in a certain unit space, the calculated MD is likely to become a large value. In other words, the wider the distribution of the feature of the lot used to generate the unit space, the smaller the MD calculated in the generated unit space.
[0064] As described above, some of the feature quantities of the lot are excluded from the feature quantities of the lot in the period P2 used to generate the unit space. The lot from which the measurement values used to calculate the feature quantities used to generate the unit space are obtained is referred to as the lot 73 within the unit space. The lot from which the measurement values used to calculate the feature quantities not used to generate the unit space are obtained is referred to as the lot 74 outside the unit space. In other words, the unit space is generated based on the feature quantities of the lot 73 within the unit space and not based on the feature quantities of the lot 74 outside the unit space. As described above, the MD of the lot used to generate the unit space is naturally calculated to be a small value. Therefore, the MD of the lot 74 outside the unit space tends to be a larger value than the MD of the lot 73 within the unit space.
[0065] Here, the lot outside the unit space 74 is a lot whose feature value calculated based on the measurement value is not used to generate the unit space, but which was processed during the period P2 in which the state of the wire saw device is normal. Therefore, a predetermined threshold is set so that the MD of the lot outside the unit space 74 is less than a predetermined threshold. If the MD of the lot outside the unit space 74 is set to be equal to or greater than the predetermined threshold, the state of the wire saw device is likely to be erroneously determined to be abnormal even if it is normal. Therefore, by setting the predetermined threshold to a value greater than the MD of the lot outside the unit space 74, the control unit 51 can easily avoid determining that the state of the wire saw device is abnormal even if it is normal, in the state determination of the wire saw device described later. In addition, if a threshold has already been set, it is possible to verify whether the unit space is appropriately set based on the value of the MD of the lot (lot outside the unit space 74) removed from the unit space.
[0066] As described above, the control unit 51 generates a unit space based on the feature amount of each lot processed during a predetermined period in the past from the start to the stop of the wire saw device. That is, the control unit 51 acquires the feature amount of a past processed lot, for which the timing at which the wire saw device was stopped is known, as reference data, and generates a unit space based on the reference data. In the unit space generated in this manner, the control unit 51 acquires the feature amount of a lot newly processed by the wire saw device as data to be determined, calculates the MD of the lot, and can determine the state of the wire saw device.
[0067] <<Example of comparison of data item selection>> The unit space that was the subject of the above-mentioned validity verification was generated based on the feature quantities of the lot of the cause data items that are known to have a large contribution to the known mode. Here, as a comparative example, an example of the MD value calculated in the unit space generated based on the non-cause data items that are known to have little or small contribution to the known mode among the data items of the lot processed until it was stopped in the known mode will be described. FIG. 5 shows a bar graph in which the MD values of each lot are arranged in the order of processing by the wire saw device in the unit space generated based on the non-cause data items. The meanings of the horizontal and vertical axes and the meanings of the periods P1, P2, and P3 are the same as those in FIG. 4 and will be omitted.
[0068] When comparing the MD values of the lots exemplified in FIG. 5, the difference between the MD of the lot 74 outside the unit space in the period P2 and the MD of each lot in the period P3 is small. When MD_T, which represents a predetermined threshold, is set low, the possibility of erroneously determining that the state of the wire saw device is normal but abnormal increases. When MD_T is set high, the possibility of not determining that the state of the wire saw device is abnormal even though the state is abnormal and there is a high possibility that the wire saw device will stop increases. When the state of the wire saw device is abnormal but not determined to be abnormal, the wire saw device loses the opportunity to be stopped preventively, and the possibility of it breaking down and stopping during lot processing increases. In the example of FIG. 5, no matter how MD_T is set, the possibility of erroneously determining the state of the wire saw device increases.
[0069] In view of the examples of FIG. 4 and FIG. 5, the accuracy of judging the state of the wire saw device can be improved by subtracting the MD of the lot 74 outside the unit space in the period P2 from the MD of at least some of the lots in the period P3 to a predetermined value or more. The predetermined value may be appropriately determined. The control unit 51 can appropriately select a cause data item from a plurality of data items so that the MD of the lots included in the periods P2 and P3 satisfies the above-mentioned condition, and generate a unit space based on the cause data item. The control unit 51 sets the predetermined threshold value to a value smaller than the MD of at least some of the lots in the period P3 and larger than the maximum value of the MD of the lot 74 outside the unit space in the period P2. This makes it easier for the control unit 51 to avoid judging the state of the wire saw device as abnormal even though it is normal.
[0070] If the state determination device 50 generates a unit space based on all data items, the unit space will also be generated based on non-cause data items that contribute little or only slightly to the cause of the wire saw device stopping. In that case, the MD value of the lot calculated in the unit space generated based on all data items will not easily reflect the change in the state of the wire saw device. Therefore, when the state of the wire saw device is determined based on the MD value of the lot calculated in the unit space generated based on all data items, the determination accuracy may be low, as in the example of FIG. 5.
[0071] <<Update unit space>> As described above, the control unit 51 of the state determination device 50 generates the unit space based on the reference data acquired from the wire saw device. However, the feature amount of each lot of the data items of the processing device 10 such as the wire saw device changes with an increase in the operating time or with the passage of time. For example, when cooling water is used in the processing device 10, the cooling water temperature may vary depending on the season. Also, for example, the feature amount of each lot may change due to the change over time of the parts of the processing device 10.
[0072] The criteria for judging the state of the wire saw device may change depending on the change in the feature amount of each lot of the data items of the wire saw device. The state judging device 50 according to the present embodiment can update the unit space by updating the reference data for generating the unit space so that the judging criteria match the change in the feature amount of each lot of the data items of the wire saw device. An example of a configuration for updating the reference data will be described below.
[0073] <<<Removal of reference data>>> The control unit 51 calculates the MD value of the feature amount of each lot included in the reference data in the unit space generated based on the reference data. The control unit 51 regards the feature amounts calculated for each of the multiple data items of one lot as a set of feature amounts, and calculates one MD value for one lot based on the set of feature amounts. If the MD value is large, the feature amount of that lot causes the unit space to become wider. If the unit space is too wide, it becomes difficult to detect that the state of the wire saw device is abnormal.
[0074] Therefore, when the MD value of the feature amount of each lot is equal to or greater than a predetermined threshold, the control unit 51 deletes the feature amount of the lot from the reference data. When the feature amount of the lot includes feature amounts of multiple data items, the control unit 51 deletes the feature amounts of all the data items of the lot from the reference data collectively. When the feature amount of the lot includes multiple feature amounts for each data item, the control unit 51 deletes all the feature amounts of each data item collectively from the reference data. When the MD value of the feature amount of each lot is less than a predetermined threshold, the control unit 51 leaves the feature amount of the lot in the reference data. The control unit 51 determines whether to delete all the multiple feature amounts of the multiple data items of the lot from the reference data or to leave them in the reference data. The predetermined threshold is also called a deletion threshold. The control unit 51 may determine whether to delete the feature amounts of all the lots included in the reference data from the reference data, or may determine whether to delete the feature amounts of some of the lots from the reference data.
[0075] FIG. 6 illustrates the MD values of the features of the 20 oldest lots obtained from the wire saw device. The horizontal axis of FIG. 6 represents the numbers assigned to each lot in order of age. The vertical axis represents the MD value of the feature of each lot. On the vertical axis, the deletion threshold is indicated as MD_D. The control unit 51 determines that the MD values of the features of the first to tenth lots and the twelfth lot in order of age are less than the deletion threshold, and leaves them in the reference data. Conversely, the control unit 51 determines that the MD values of the features of the eleventh lot and the thirteenth to twentieth lots in order of age are greater than or equal to the deletion threshold, and deletes them from the reference data.
[0076] The control unit 51 updates the existing unit space by generating a unit space based on the reference data from which the feature quantities of some lots have been deleted. The updated unit space is also referred to as an updated unit space. The updated unit space generated by deleting the feature quantities of lots with a large MD value may be narrower than a unit space generated by simply deleting the feature quantities of the oldest 20 lots from the reference data. In other words, the control unit 51 can control the unit space not to be too wide by appropriately deleting the feature quantities of some lots from the reference data, thereby suppressing a decrease in sensitivity due to seasonal fluctuations or changes over time. As a result, the abnormality detection performance of the state determination device 50 is improved or is less likely to decrease.
[0077] When generating an updated unit space based on the reference data, the control unit 51 may apply various methods described as a method for generating a unit space. For example, the control unit 51 may generate a unit space by dividing the reference data from which the feature quantity of the lot has been deleted or added into data within the unit space and data outside the unit space.
[0078] <<<Adding reference data>>> The control unit 51 acquires the feature quantities of a newly processed lot by operating the wire saw device. The control unit 51 may add the newly acquired feature quantities of the lot to the reference data. In this way, the feature quantities of the lot that have changed depending on the state of the wire saw device can be reflected in the unit space. As a result, the criteria for determining the state of the wire saw device can be changed depending on the change in the state of the wire saw device.
[0079] The control unit 51 calculates the MD value of the feature of the newly acquired lot in the existing unit space. The control unit 51 regards the feature calculated for each of the multiple data items of one lot as a set of feature, and calculates one MD value for one lot based on the set of feature. If the MD value is large, the feature of that lot causes the unit space to become wider. If the unit space is too wide, it becomes difficult to detect that the state of the wire saw device is abnormal. Therefore, the control unit 51 may add only the feature of the lot with a small MD value to the reference data.
[0080] Therefore, when the MD value of the newly acquired feature of a lot is less than a predetermined threshold, the control unit 51 adds the feature of the lot to the reference data. When the feature of a lot includes feature of multiple data items, the control unit 51 adds the feature of all the data items of the lot together to the reference data. When the feature of a lot includes multiple feature for each data item, the control unit 51 adds all the feature of each data item together to the reference data. On the other hand, when the MD value of the newly acquired feature of a lot is equal to or greater than a predetermined threshold, the control unit 51 does not add the feature of the lot to the reference data. The control unit 51 judges whether to add all of the multiple feature of the multiple data items of the lot together to the reference data. The predetermined threshold for judging whether to add a new feature to the reference data is also called an addition threshold. The addition threshold is, delete It may be set to the same value as the threshold value, or may be set to a different value.
[0081] The control unit 51 may acquire the feature amounts of a plurality of lots at once, calculate the MD value of the feature amount of each lot, and determine whether to add it to the reference data.
[0082] FIG. 7 shows an example of the MD values of the features of 20 lots newly acquired from the wire saw device. The horizontal axis of FIG. 7 represents the numbers assigned to each lot in the order in which they were processed by the wire saw device. The vertical axis represents the MD values of the features of each lot. addition The threshold value is indicated as MD_A. The control unit 51 determines that the MD values of the feature amounts of the 11th to 19th lots in the order of processing are less than the addition threshold value, and adds them to the reference data. Conversely, the control unit 51 determines that the MD values of the feature amounts of the 1st to 10th lots and the 20th lot in the order of processing are equal to or greater than the addition threshold value, and does not add them to the reference data.
[0083] The control unit 51 updates the existing unit space by generating a unit space based on the reference data to which the feature quantities of the new lot have been added. The updated unit space generated by adding only the feature quantities of lots with small MD values is less likely to expand than a unit space generated by simply adding the feature quantities of 20 lots to the reference data. In other words, the control unit 51 can control the unit space not to expand too much by appropriately adding the feature quantities of the new lot to the reference data, thereby suppressing a decrease in sensitivity due to seasonal fluctuations or changes over time. As a result, the anomaly detection performance of the status determination device 50 is improved or is less likely to decrease.
[0084] <<<Additional Exception Handling for Reference Data>>> The feature amount of a lot may shift due to a change in the state of the wire saw device. When adding only the feature amount of a lot whose MD value is less than the addition threshold to the reference data, it may not be possible to update the unit space to shift according to the shift in the feature amount of the lot. Therefore, the control unit 51 may add the feature amount of a new lot to the reference data based on the following conditions as an exception processing for adding the reference data.
[0085] The control unit 51 calculates the MD value of the feature amount of each lot newly processed by the wire saw device. The control unit 51 determines whether the MD value of the feature amount of each lot is less than the addition threshold. In principle, the control unit 51 adds only the feature amount of the lot whose MD value is less than the addition threshold to the reference data.
[0086] Here, as an exception process, when a predetermined number or more of lots appear consecutively in which the MD value is equal to or greater than the additional threshold, the control unit 51 may add the feature quantities of those lots to the reference data. This is because it is statistically unlikely that data for a predetermined number or more of lots consecutively deviates from a predetermined range (a range determined based on the unit space in this embodiment). For example, it is statistically unlikely that a process capability index (Cp or Cpk, etc.) used in quality control deviates from the predetermined number or more of lots consecutively. Here, the predetermined number may be set to a value of 2 or more. For example, the predetermined number may be set so that the probability that the MD value continues to be equal to or greater than the additional threshold for a predetermined number or more consecutively is 0.3% or less (so-called 3σ probability).
[0087] From the above, it is considered that the phenomenon in which the MD value is equal to or greater than the additional threshold value for a predetermined number of consecutive times is not a sudden change in the feature amount of the lot processed by the wire saw device, but occurs as a reflection of a change in the trend of the feature amount of each lot. Therefore, by the control unit 51 executing the above-mentioned exception processing, the change in the trend of the feature amount of the lot processed by the wire saw device can be reflected in the unit space. As a result, the accuracy of the abnormality detection by the state determination device 50 can be improved.
[0088] <<<Deleting and adding reference data>>> The control unit 51 may perform an operation to delete features of some lots from the reference data and an operation to add features of a new lot to the reference data, and then generate a unit space based on the newly obtained reference data by performing these operations.
[0089] In the example of Fig. 6, feature amounts of 11 lots are deleted from the reference data. In the example of Fig. 7, feature amounts of 9 lots are added to the reference data. As a result, the number of lots constituting the reference data is reduced by 2. In order to keep the number of lots constituting the reference data constant, the control unit 51 may set a deletion threshold and an addition threshold so that the number of lots to be deleted and the number of lots to be added match.
[0090] For example, in the example of FIG. 6, the control unit 51 determines whether the MD value of the feature amount of the fifth lot is equal to the deletion threshold. End In this way, the number of lots deleted from the reference data becomes 10. On the other hand, in the example of FIG. 7, the control unit 51 may increase the addition threshold so that the MD value of the feature amount of the first lot becomes less than the addition threshold. In this way, the number of lots added to the reference data becomes 10. As a result, the control unit 51 can match the number of lots to be deleted with the number of lots to be added.
[0091] The control unit 51 delete The number of lots to be deleted and the number of lots to be added may be made equal by increasing both the threshold value and the addition threshold value. delete The number of lots removed may be made equal to the number of lots added by reducing both the threshold and the addition threshold. delete By increasing both the threshold value and the additional threshold value or decreasing both the threshold value and the additional threshold value, the state of the wire saw device is more easily reflected in the unit space. As a result, the abnormality detection performance of the state determination device 50 is improved or is less likely to deteriorate.
[0092] <<Determination of the state of the wire saw device based on the generated or updated unit space and the determination threshold value>> For a wire saw device whose stopping timing is unknown, the control unit 51 acquires, as data to be judged, the characteristic quantities of a lot calculated based on measurement values obtained when the device processes the lot, and judges whether there is an increased possibility that the wire saw device will stop based on the data to be judged.
[0093] Specifically, the control unit 51 calculates the MD of a newly processed lot in a unit space generated or updated based on the feature amount of a previously processed lot. The control unit 51 determines, based on the calculated MD value, whether or not there is a high possibility that the wire saw device will stop in processing the next lot after that lot.
[0094] When MD is less than a predetermined threshold, the control unit 51 determines that there is a low possibility that the wire saw device will stop, that is, the state of the wire saw device when processing that lot is normal. When MD is equal to or greater than a predetermined threshold, the control unit 51 determines that there is a high possibility that the wire saw device will stop, that is, the state of the wire saw device when processing that lot is abnormal. The predetermined threshold for determining the state of the wire saw device is also referred to as a determination threshold.
[0095] The control unit 51 may notify the user that the state of the wire saw device is abnormal by having the output unit 53 display or output by voice. Based on the notification, the user may not have the wire saw device process the next lot, but may preventively stop the wire saw device and perform maintenance work. The control unit 51 may output to the wire saw device that the state of the wire saw device is abnormal, and preventively stop the wire saw device.
[0096] <<Judging abnormal conditions based on known modes>> When it is known that the wire saw device has stopped in a known mode, the control unit 51 may acquire, as reference data, the feature quantities of the lot processed from the start of operation to the stop of operation. Data including the feature quantities of the lot processed until the device stopped in the known mode is also referred to as known mode data. The control unit 51 may generate or update the unit space based on the known mode data. The unit space based on the known mode data is also referred to as a known mode unit space.
[0097] The control unit 51 calculates the MD of a lot processed by the wire saw device whose state is unknown in the known mode unit space. When the MD of the lot calculated in the known mode unit space is equal to or greater than a predetermined threshold, the control unit 51 may determine that the wire saw device is in an abnormal state where there is a high possibility that the wire saw device will stop in the known mode. The abnormal state where there is a high possibility that the wire saw device will stop in the known mode is also referred to as an abnormal state based on the known mode.
[0098] When the known mode includes a plurality of failure modes such as a first known mode or a second known mode, the control unit 51 may generate or update a unit space corresponding to each failure mode based on reference data including data of a lot processed until the wire saw device stops in each failure mode. For example, the control unit 51 may generate or update a first unit space corresponding to the first known mode or a second unit space corresponding to the second known mode. The control unit 51 may calculate the MD of a lot processed by the wire saw device in a unit space corresponding to each failure mode. When the MD of a lot calculated in a unit space corresponding to each failure mode is equal to or greater than a predetermined threshold, the control unit 51 may determine that the wire saw device is in an abnormal state in which the possibility of stopping in a failure mode in which the MD is equal to or greater than the predetermined threshold is high. The thresholds for determining the values of MD of the first unit space and the second unit space are also referred to as a first threshold and a second threshold, respectively. The first threshold and the second threshold may be different values from each other, or may be the same value. The states in which the wire saw device is likely to stop in the first known mode and the second known mode are also referred to as the first abnormal state and the second abnormal state, respectively. The thresholds are not limited to the first threshold and the second threshold, and may include, for example, a third threshold set to determine the MD of the third unit space. The abnormal states are not limited to the first abnormal state and the second abnormal state, and may include, for example, a third abnormal state corresponding to a state in which the wire saw device is likely to stop in the third known mode. It should be noted that identifiers such as "first" and "second" do not indicate the superiority or inferiority of the thresholds or abnormal states.
[0099] The control unit 51 may determine whether the state of the wire saw device is an abnormal state based on a known mode or a first or second abnormal state, and may cause the output unit 53 to output the determination result. The control unit 51 may notify a user, such as an operator or maintenance personnel of the wire saw device, of the determination result. Based on the notified determination result, the user performs various tasks, such as inspection work, part replacement work, or repair work of the wire saw device. This allows the user to properly maintain the wire saw device so that it does not break down during lot processing.
[0100] <<Determining the possibility of an unknown cause stopping>> The control unit 51 may acquire, as reference data, the characteristic quantities of the lot processed from the start of operation until the wire saw device is stopped for an unknown reason. Data including the characteristic quantities of the lot processed until the wire saw device is stopped for an unknown reason is also referred to as unknown cause data. The control unit 51 may generate or update the unit space based on the unknown cause data. The unit space based on the unknown cause data is also referred to as an unknown cause unit space. The control unit 51 may generate or update the unknown cause unit space by selecting all data items obtained from the wire saw device, or may generate or update the unknown cause unit space by selecting some of the data items.
[0101] The control unit 51 calculates the MD of a lot processed by the wire saw device in an unknown state in the unknown cause unit space. When the MD of a lot calculated in the unknown cause unit space is equal to or greater than a predetermined threshold, the control unit 51 may determine that the wire saw device is in an abnormal state in which the possibility of stopping due to an unknown cause is high. The threshold for determining the value of the MD of a lot calculated in the unknown cause unit space is also referred to as the unknown cause threshold. The unknown cause threshold is set to the same value as the predetermined threshold, the first threshold, the second threshold, or the like used to determine the value of the MD of a lot calculated in the known mode unit space, but may be set to a different value.
[0102] When the control unit 51 determines that the wire saw device is likely to stop due to an unknown cause, the control unit 51 may further narrow down the causes of the wire saw device stopping by performing a validity analysis. The validity analysis corresponds to analyzing which data item among the data items used to generate or update the cause unit space has a large effect on the value of MD. The control unit 51 may perform the validity analysis by various methods. The control unit 51 can determine a data item that increases the value of MD of a lot calculated in the cause unknown unit space by the validity analysis. The data item that increases the value of MD of a lot calculated in the cause unknown unit space is also referred to as a cause candidate item. A specific example of a method of determining a cause candidate item by the validity analysis will be described later.
[0103] The control unit 51 may determine the candidate cause items and output them to the output unit 53, thereby notifying a user, such as an operator or maintenance staff, of the wire saw device. Based on the data items determined as the candidate cause items, the user performs various tasks, such as inspection work, part replacement work, or repair work, of the wire saw device. This allows the wire saw device to be properly maintained before it breaks down during lot processing.
[0104] <Changes in MD value of data to be judged due to updating of unit space> As described above, the control unit 51 may update the unit space by adding the feature amount of a new lot processed by the wire saw device to the reference data or deleting the feature amount of an old lot from the reference data. In order to explain the effect of updating the unit space by the method of this embodiment, the MD values of the data to be determined are compared in the following cases: when the control unit 51 does not update the unit space, when the unit space is updated by the method of the comparative example, and when the unit space is updated by the method of this embodiment.
[0105] In the following description, the unit space is generated or updated based on the reference data of the first wire saw device. The data to be judged is the feature of the lot processed by the second wire saw device. That is, for convenience, the device that acquires the reference data is different from the device that acquires the data to be judged. By differentiating the device that acquires the reference data from the device that acquires the data to be judged, the value of MD is calculated at a large value. By calculating the value of MD at a large value, it becomes easier to compare the values of MD in each case.
[0106] First, when the unit space is not updated, the MD value of the judgment target data calculated in the unit space is illustrated in Fig. 8. The horizontal axis of Fig. 8 represents the lot, and the vertical axis represents the MD value of each lot.
[0107] Next, as a method of the comparative example, the control unit 51 deletes the feature amounts of the oldest 20 lots from the reference data and adds the feature amounts of the newly processed 20 lots without considering the MD values of the feature amounts of each lot. The MD values of the data to be judged calculated in the unit space updated by the method of the comparative example are illustrated in Fig. 9. The horizontal axis of Fig. 9 represents the lot, and the vertical axis represents the MD value of each lot.
[0108] Next, in the method of this embodiment, the control unit 51 deletes the feature quantities of 11 lots whose MD values are equal to or greater than the deletion threshold among the feature quantities of the 20 lots starting from the oldest one in the reference data, and adds the feature quantities of 9 lots whose MD values are less than the addition threshold among the feature quantities of the newly processed 20 lots. The MD values of the data to be determined calculated in the unit space updated by the method of this embodiment are illustrated in Fig. 10. The horizontal axis of Fig. 10 represents the lot, and the vertical axis represents the MD value of each lot.
[0109] Comparing Fig. 8 with Fig. 9, the MD value of each lot is smaller in Fig. 9. The smaller MD value of each lot means that the sensitivity of detecting whether the state of the wire saw device is abnormal is lower when the determination is made using the unit space updated by the method of the comparative example than when the determination is made using the unit space that has not been updated.
[0110] On the other hand, when comparing Fig. 8 with Fig. 10, the MD value of each lot is larger in Fig. 10. The fact that the MD value of each lot is larger means that when using a unit space updated by the method of this embodiment, the sensitivity of detecting whether the state of the wire saw device is abnormal is improved compared to when a unit space that has not been updated is used for judgment.
[0111] As described above, the control unit 51 of this embodiment can respond to data that fluctuates due to seasonal variations or changes over time by deleting old reference data and adding new reference data while taking into account the MD value of the lot's characteristics, thereby improving the sensitivity of anomaly detection in the wire saw device using the unit space.
[0112] Furthermore, as an exception process in the method of this embodiment, when a predetermined number or more of lots continue with the MD value being equal to or greater than the addition threshold, the control unit 51 adds the lot meeting that condition to the reference data. The MD value of the judgment target data calculated in the unit space updated based on the reference data to which the exception process is applied in the method of this embodiment is illustrated in Fig. 11. The horizontal axis in Fig. 11 represents the lot, and the vertical axis represents the MD value of each lot.
[0113] Comparing FIG. 8 with FIG. 11, the MD value of each lot is larger in FIG. 11. That is, the MD value of each lot is not smaller. Also, comparing FIG. 10 with FIG. 11, the MD value of each lot is smaller in FIG. 11. From these comparison results, it means that by applying exception processing for adding the feature amount of a new lot to the reference data, the unit space does not become too narrow, and the sensitivity of detecting whether the state of the wire saw device is abnormal does not decrease. In other words, the unit space is updated to shift in accordance with the shift in the trend of the feature amount of the lot newly processed by the wire saw device. As a result, the accuracy of detecting whether the state of the wire saw device is abnormal can be improved.
[0114] <Summary> As described above, the state determination system 1 and state determination device 50 according to this embodiment can determine whether the processing device 10, such as a wire saw device, is in an abnormal state in which the possibility of the processing device 10 breaking down is increased before the processing device 10 actually breaks down. In this way, the processing device 10 is less likely to break down during processing. If the processing device 10 breaks down during processing, a large loss will occur. By making the processing device 10 less likely to break down during processing, the loss can be reduced.
[0115] (Example of procedure for determining status) The control unit 51 of the state determination device 50 may execute a state determination method including the steps of the flow charts illustrated in Figures 12, 13, and 14, for example. The control unit 51 can determine whether the state of the wire saw device is abnormal by executing the illustrated state determination method. The state determination method may be realized as a state determination program executed by the control unit 51. The steps illustrated in Figures 12 and 13 are examples and may be changed as appropriate.
[0116] The control unit 51 generates or updates a unit space (step S1). The control unit 51 generates or updates the unit space based on reference data of the wire saw device. When the wire saw device stops in a known mode, the control unit 51 may generate or update a unit space corresponding to the known mode based on reference data obtained as data of the lot processed until the device stopped. When the wire saw device stops for an unknown reason, the control unit 51 may generate or update a unit space corresponding to the case where the cause is unknown based on reference data obtained as data of the lot processed until the device stopped.
[0117] An example of a procedure for generating a unit space is shown as a subroutine in FIG. 13. The control unit 51 acquires reference data (step S11 in FIG. 13). The reference data includes a feature amount of a lot calculated based on the measurement values of each lot processed when the wire saw device operated in the past until it stopped. The control unit 51 may acquire, as the reference data, a feature amount of a lot processed before the wire saw device stopped in a known mode such as the first known mode or the second known mode. The control unit 51 may acquire, as the reference data, a feature amount of a lot processed before the wire saw device stopped for an unknown reason.
[0118] The control unit 51 selects a cause data item from the reference data (step S12). The reference data includes a plurality of data items that can be acquired from the wire saw device. The control unit 51 may select a data item corresponding to a predetermined failure mode that is the cause of the wire saw device stopping as the cause data item. The control unit 51 may select, for example, data items corresponding to a first known mode and a second known mode as the first cause data item and the second cause data item, respectively. When the control unit 51 acquires, as reference data, the feature amounts of the lot that was processed before the wire saw device stopped due to an unknown cause, the control unit 51 may select all the cause data items.
[0119] The control unit 51 classifies the reference data into data within the unit space, which is data on the lot 73 within the unit space, and data outside the unit space, which is data on the lot 74 outside the unit space (step S13).
[0120] The control unit 51 generates a unit space based on the data within the unit space (step S14). The control unit 51 may generate a first unit space, a second unit space, or a unit space of unknown cause as the unit space.
[0121] The control unit 51 determines whether the generated unit space is appropriate (step S15). Specifically, the control unit 51 calculates the MD of the lot 73 in the unit space and the MD of the lot 74 outside the unit space in the generated unit space. For example, the control unit 51 may determine that the generated unit space is appropriate when the difference between the MD of the lot 73 in the unit space and the MD of the lot 74 outside the unit space is equal to or less than a predetermined value. When the control unit 51 determines that the generated unit space is not appropriate (step S15: NO), it returns to the procedure of step S13. The control unit 51 may return to the procedure of step S12. When the control unit 51 determines that the generated unit space is appropriate (step S15: YES), it ends the subroutine processing exemplified in FIG. 13 and returns to the procedure of step S2 in FIG. 12.
[0122] The control unit 51 may generate unit spaces corresponding to a plurality of known modes in a subroutine process of a unit space generation procedure exemplified in Fig. 13. The control unit 51 may generate a unit space of unknown cause corresponding to a stop of unknown cause.
[0123] FIG. 14 shows an example of a procedure for updating the unit space as a subroutine.
[0124] The control unit 51 calculates the value of MD of the feature amount of each lot included in the reference data used to generate the unit space (step S21). The control unit 51 determines whether the value of MD of the feature amount of each lot is equal to or greater than the deletion threshold (step S22). If the value of MD is equal to or greater than the deletion threshold (step S22: YES), the control unit 51 deletes the feature amount of the lot from the reference data (step S23). If the value of MD is not equal to or greater than the deletion threshold (step S22: NO), that is, if the value of MD is less than the deletion threshold, the control unit 51 skips step S23 and proceeds to the procedure of step S24 so as not to delete the feature amount of the lot from the reference data. The control unit 51 may execute the procedures of steps S22 and S23 for the feature amounts of some lots included in the reference data, or may execute the procedures of all lots. The control unit 51 can delete the feature amount of the lot from the reference data as appropriate by executing the procedures from steps S21 to S23.
[0125] The control unit 51 acquires the feature amount of a lot newly processed by the wire saw device, and calculates the MD value of the feature amount of the acquired lot (step S24). The control unit 51 determines whether the MD value of the feature amount of the lot is less than the addition threshold value (step S25). If the MD value is less than the addition threshold value (step S25: YES), the control unit 51 adds the feature amount of the lot to the reference data (step S26).
[0126] If the value of MD is not less than the addition threshold (step S25: NO), that is, if the value of MD is equal to or greater than the addition threshold, the control unit 51 determines whether the MD value of the feature of the lot is equal to or greater than the addition threshold for a predetermined number of consecutive times or more (step S27). If the MD value of the feature of the lot is not equal to or greater than the addition threshold for a predetermined number of consecutive times or more (step S27: NO), the control unit 51 skips the procedure of step S26 so as not to add a feature of a new lot to the reference data and proceeds to the procedure of step S28. If the MD value of the feature of the lot is equal to or greater than the addition threshold for a predetermined number of consecutive times or more (step S27: YES), the control unit 51 proceeds to the procedure of step S26 and adds to the reference data the feature of the predetermined number or more of lots whose MD values are equal to or greater than the addition threshold for consecutive times. The control unit 51 can appropriately add the feature of the lot to the reference data by executing the procedures from step S24 to S27.
[0127] The control unit 51 generates an update unit space based on the reference data (step S28). After executing the procedure of step S28, the control unit 51 ends the subroutine processing exemplified in Fig. 14, and returns to the procedure of step S2 in Fig. 12.
[0128] Returning from the subroutine process of FIG. 13 or FIG. 14 executed as the procedure of step S1 of FIG. 12, the control unit 51 acquires the determination target data (step S2 of FIG. 12).
[0129] The control unit 51 calculates the MD of the lot included in the judgment target data (step S3). The control unit 51 may calculate the MD of the lot included in the judgment target data for each of the multiple unit spaces generated or updated in the procedure of step S1. The control unit 51 may calculate the MD of the lot included in the judgment target data in the first unit space. The MD of the lot included in the judgment target data calculated in the first unit space is also referred to as the MD of the first unit space. The control unit 51 may calculate the MD of the lot included in the judgment target data in the second unit space. The MD of the lot included in the judgment target data calculated in the second unit space is also referred to as the MD of the second unit space. The control unit 51 may calculate the MD of the lot included in the judgment target data in the unknown cause unit space. The MD of the lot included in the judgment target data calculated in the unknown cause unit space is also referred to as the MD of the unknown cause unit space.
[0130] The control unit 51 determines whether the MD of the first unit space is equal to or greater than the first threshold value (step S4). The control unit 51 may set the first threshold value when generating the first unit space in the procedure of step S1, for example. The first threshold value may be set so as to detect that the state of the wire saw device is in an abnormal state based on the first known mode, while preventing the wire saw device from being erroneously detected as abnormal when the state of the wire saw device is normal.
[0131] If the MD of the first unit space is equal to or greater than the first threshold (step S4: YES), the control unit 51 proceeds to the procedure of step S8. If the MD of the first unit space is not equal to or greater than the first threshold (step S4: NO), that is, if the MD of the first unit space is less than the first threshold, the control unit 51 determines whether the MD of the second unit space is equal to or greater than the second threshold (step S5). The control unit 51 may set the second threshold when generating the second unit space in the procedure of step S1, for example. The second threshold can be set so as to detect that the state of the wire saw device is in an abnormal state based on the second known mode, while not erroneously detecting that the state of the wire saw device is abnormal when the state of the wire saw device is normal.
[0132] If the MD of the second unit space is equal to or greater than the second threshold (step S5: YES), the control unit 51 proceeds to the procedure of step S8. If the MD of the second unit space is not equal to or greater than the second threshold (step S5: NO), that is, if the MD of the second unit space is less than the second threshold, the control unit 51 determines whether the MD of the unknown cause unit space is equal to or greater than the unknown cause threshold (step S6). For example, the control unit 51 may set the unknown cause threshold when generating the unknown cause unit space in the procedure of step S1. The unknown cause threshold can be set so as to detect that the state of the wire saw device is in an unknown abnormal state, while not erroneously detecting that the wire saw device is in an abnormal state when the state of the wire saw device is normal.
[0133] If the MD of the unknown cause unit space is not equal to or greater than the unknown cause threshold (step S6: NO), that is, if the MD of the unknown cause unit space is less than the unknown cause threshold, the control unit 51 ends the execution of the procedure of the flowchart in FIG.
[0134] When the MD of the unknown cause unit space is equal to or greater than the unknown cause threshold (step S6: YES), the control unit 51 determines a cause candidate item that is a data item corresponding to a factor that increases the possibility of the wire saw device stopping (step S7). The control unit 51 may determine the cause candidate item by validity analysis. A specific example of the validity analysis will be described later.
[0135] After the procedure of step S7, or when it is determined in the procedure of step S4 or S5 that MD is equal to or greater than the threshold value, the control unit 51 preventively stops the wire saw device (step S8). In this way, the wire saw device as the processing device 10 can be maintained by inspection work, part replacement work, repair work, or the like before it breaks down and stops.
[0136] According to the above-described state determination method, it is possible to determine whether the state of the wire saw device as the processing device 10 is abnormal before the wire saw device breaks down. In this way, the wire saw device as the processing device 10 is less likely to break down and stop during lot processing. If the processing device 10 breaks down during lot processing, a large loss will occur, such as the lot being processed having to be discarded. By making it less likely that the processing device 10 will break down and stop during lot processing, losses can be reduced.
[0137] (Other embodiments) In the above-described embodiment, a configuration has been described in which the state determination device 50 determines the state of a wire saw device as the processing device 10. The processing device 10 that the state determination device 50 is not limited to a wire saw device, and may determine the state of other devices such as a polishing device.
[0138] Although the embodiments according to the present disclosure have been described based on the drawings and examples, it should be noted that a person skilled in the art can make various modifications or alterations based on the present disclosure. Therefore, it should be noted that these modifications or alterations are included in the scope of the present disclosure. For example, the functions included in each component or each step can be rearranged so as not to be logically inconsistent, and multiple components or steps can be combined into one or divided. Although the embodiments according to the present disclosure have been described mainly with respect to the device, the embodiments according to the present disclosure can also be realized as a method including steps executed by each component of the device. The embodiments according to the present disclosure can also be realized as a method, a program, or a storage medium on which a program is recorded, executed by a processor included in the device. It should be understood that these are also included in the scope of the present disclosure.
[0139] The graphs included in this disclosure are schematic and do not necessarily correspond to the actual scale. [Industrial Applicability]
[0140] According to the embodiment of the present disclosure, the accuracy of determining the state of the processing device 10 can be improved. [Explanation of symbols]
[0141] 1. Status Judgment System 10 Processing device (12: wire, 14: roller, 16: wire group, 18: work holding mechanism, 20: nozzle, 32: guide roller, 33: dancer arm, 34: dancer roller, 35: touch roller, 36: drive motor, 38A, 38B: wire reel, 40: slurry tank, 42: slurry chiller, W: work (block), X: roller axis direction) 50 state determination device (51: control unit, 53: output unit, 54: input unit) 60 Sensors 71 Stop Lot 72 Abnormal detection lot 73 Lots in a Unit Space 74 Unit Space Outer Lot
Claims
1. A control unit is provided for determining a state of the processing device, The control unit is acquiring, as reference data, feature quantities of a plurality of lots processed when the state of the processing device is known to be normal, the feature quantities being measurement values of at least one data item of the processing device acquired during a period in which the processing device processed each of the plurality of lots; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data and using data as the reference data, the data being a set of a plurality of feature amounts calculated for each of a plurality of periods during the processing of one lot; calculating a Mahalanobis distance of the feature amount of each lot included in the reference data in the unit space, and deleting the feature amount of the lot whose Mahalanobis distance is equal to or greater than a deletion threshold value from the reference data; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which the state of the processing device is to be judged; Calculating a Mahalanobis distance of the determination target data in the update unit space; a state determination device that outputs a result that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value that is set based on the update unit space;
2. A control unit is provided for determining a state of the processing device, The control unit is acquiring, as reference data, feature quantities of a plurality of lots processed when the state of the processing device is known to be normal, the feature quantities being measurement values of at least one data item of the processing device acquired during a period in which the processing device processed each of the plurality of lots; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data and using data as the reference data, the data being a set of a plurality of feature amounts calculated for each of a plurality of periods during the processing of one lot; calculating, in the unit space, a Mahalanobis distance of a feature quantity of a newly acquired lot to be added to the reference data, and adding, to the reference data, the feature quantity of the lot whose Mahalanobis distance is less than an addition threshold value; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which the state of the processing device is to be judged; Calculating a Mahalanobis distance of the determination target data in the update unit space; a state determination device that outputs a result that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value that is set based on the update unit space;
3. A control unit is provided for determining a state of the processing device, The control unit is acquiring, as reference data, feature quantities of a plurality of lots processed when the state of the processing device is known to be normal, the feature quantities being measurement values of at least one data item of the processing device acquired during a period in which the processing device processed each of the plurality of lots; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data and using data as the reference data, the data being a set of a plurality of feature amounts calculated for each of a plurality of periods during the processing of one lot; calculating a Mahalanobis distance of the feature amount of each lot included in the reference data in the unit space, and deleting the feature amount of the lot whose Mahalanobis distance is equal to or greater than a deletion threshold value from the reference data; calculating, in the unit space, a Mahalanobis distance of a feature quantity of a newly acquired lot to be added to the reference data, and adding, to the reference data, the feature quantity of the lot whose Mahalanobis distance is less than an addition threshold value; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which the state of the processing device is to be judged; Calculating a Mahalanobis distance of the determination target data in the update unit space; a state determination device that outputs a result that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value that is set based on the update unit space;
4. The condition determination device according to claim 3 , wherein the control unit sets the deletion threshold and the addition threshold such that a number of lots to be removed from the reference data is equal to a number of lots to be added to the reference data.
5. The condition determination device according to any one of claims 2 to 4, wherein the control unit adds the feature quantities of a predetermined number or more lots newly acquired to be added to the reference data to the reference data when the Mahalanobis distance of the feature quantities of the predetermined number or more lots is consecutively greater than or equal to the addition threshold value.
6. A state determination method executed by a state determination device that determines a state of a processing device, comprising: acquiring, as reference data, feature quantities of a plurality of lots processed when the state of the processing device is known to be normal, the feature quantities being measurement values of at least one data item of the processing device acquired during a period in which the processing device processed each of the plurality of lots; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data and using data as the reference data, the data including a set of a plurality of feature amounts calculated for each of a plurality of periods during the processing of one lot; calculating a Mahalanobis distance of feature quantities of each lot included in the reference data in the unit space, and deleting feature quantities of lots whose Mahalanobis distances are equal to or greater than a deletion threshold value from the reference data; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which a state of the processing device is to be judged; calculating a Mahalanobis distance of the data to be determined in the update unit space; a step of outputting a result of determining that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value set based on the update unit space; A state determination method comprising:
7. A state determination method executed by a state determination device that determines a state of a processing device, comprising: acquiring, as reference data, feature quantities of a plurality of lots processed when the state of the processing device is known to be normal, the feature quantities being measurement values of at least one data item of the processing device acquired during a period in which the processing device processed each of the plurality of lots; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data and using data as the reference data, the data including a set of a plurality of feature amounts calculated for each of a plurality of periods during the processing of one lot; a step of calculating, in the unit space, a Mahalanobis distance of a feature quantity of a newly acquired lot to be added to the reference data, and adding, to the reference data, the feature quantity of the lot whose Mahalanobis distance is less than an addition threshold value; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which a state of the processing device is to be judged; calculating a Mahalanobis distance of the data to be determined in the update unit space; a step of outputting a result of determining that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value set based on the update unit space; A state determination method comprising:
8. A state determination method executed by a state determination device that determines a state of a processing device, comprising: acquiring, as reference data, feature quantities of a plurality of lots processed when the state of the processing device is known to be normal, the feature quantities being measurement values of at least one data item of the processing device acquired during a period in which the processing device processed each of the plurality of lots; generating a unit space representing characteristics of a group of lots normally processed by the processing device based on the reference data and using data as the reference data, the data including a set of a plurality of feature amounts calculated for each of a plurality of periods during the processing of one lot; calculating a Mahalanobis distance of feature quantities of each lot included in the reference data in the unit space, and deleting feature quantities of lots whose Mahalanobis distances are equal to or greater than a deletion threshold value from the reference data; a step of calculating, in the unit space, a Mahalanobis distance of a feature quantity of a newly acquired lot to be added to the reference data, and adding, to the reference data, the feature quantity of the lot whose Mahalanobis distance is less than an addition threshold value; generating the unit space as an updated unit space based on the updated reference data; acquiring, as judgment target data, a feature amount of a lot for which a state of the processing device is to be judged; calculating a Mahalanobis distance of the data to be determined in the update unit space; a step of outputting a result of determining that the state of the processing device is abnormal when the Mahalanobis distance is equal to or greater than a determination threshold value set based on the update unit space; A state determination method comprising:
Citation Information
Patent Citations
Abnormality diagnostic device of work machine
JP2013199804A
Production facility abnormality diagnosis system
JP2018092437A
Abnormality diagnosis method, abnormality diagnosis device, and abnormality diagnosis program
WO2021019857A1
