Enterprise production big data mining method based on erp system

Abstract

The present application relates to big data mining technical field, specifically relates to a kind of enterprise production big data mining method based on erp system, determine the associated time period of multiple dimensions in production quality abnormal moment, determine the time overlap degree between any production abnormal event in each two dimensions, determine suspect factor combination, suspect factor combination is composed of one of each dimension production abnormal event, and wherein the time overlap degree of each two production abnormal events all satisfy preset condition;By the abnormal situation of preset key process feature in time sequence, determine abnormal occurrence moment and abnormal determination moment, thus determine the causal priority of each dimension in suspect factor combination, extract the production abnormal event of the dimension corresponding to the maximum causal priority, improve data mining accuracy and reliability, so as to accurately obtain the real reason of production process abnormality.

Description

Technical Field

[0001] This invention relates to the field of big data mining technology, specifically to a method for mining enterprise production big data based on an ERP system. Background Technology

[0002] In modern manufacturing environments driven by ERP (Enterprise Resource Planning) systems, effectively attributing production process anomalies (such as sudden increases in batch defect rates or deviations in process cycle time) is crucial for achieving lean manufacturing and quality control. Multi-source, multi-dimensional production event data (such as personnel changes, material switching, equipment status, and tooling replacements) exist as isolated records in ERP systems. Traditionally, monitoring production process anomalies relies on manual experience to investigate single-dimensional production event data (such as operator or equipment) when product quality is abnormal, aiming to uncover related production event data that caused the anomaly. However, because multi-dimensional production event data are often closely correlated, relying solely on single-dimensional production event data for data mining can compromise accuracy and reliability, failing to reveal the true cause of production process anomalies. Summary of the Invention

[0003] To address the low accuracy problem of existing data mining methods, this invention provides a method for mining enterprise production big data based on an ERP system, comprising: Identify production anomaly events across multiple dimensions within the associated timeframe at the moment of production quality anomaly. Determine the temporal overlap between any production anomaly events in each pair of dimensions; Determine a combination of suspicious factors, wherein the combination of suspicious factors consists of one of the production anomalies in each dimension, and the time overlap of every two production anomalies in the combination of suspicious factors meets a preset condition. The timing of anomalies and the timing of anomaly determination are determined by the abnormal situations in the time sequence of pre-defined key process characteristics; Based on the time of occurrence of the anomaly and the time of determination of the anomaly, the causal priority of each dimension in the combination of suspect factors is determined. Extract production anomaly events from the production anomaly dimension, where the production anomaly dimension corresponds to the dimension with the highest causal priority.

[0004] In one specific embodiment, the process of obtaining the time overlap includes: The intersection and union durations of the time intervals of two candidate production anomaly events are determined respectively; the two candidate production anomaly events are any production anomaly events in any two dimensions. The time overlap of the two candidate production anomalies is obtained by calculating the ratio of the intersection duration to the union duration.

[0005] In one specific embodiment, the time overlap degree satisfies a preset condition, including: the time overlap degree is greater than or equal to a preset time overlap degree threshold.

[0006] In one specific embodiment, the process of obtaining the preset key process features includes: Determine the baseline normal production batch from all production batches prior to the time of the aforementioned production quality anomaly. Preset key process characteristics are determined based on the benchmark normal production batch.

[0007] In one specific embodiment, the time of the anomaly occurrence is the time when a production anomaly event occurs; The process of obtaining the anomaly determination time includes: Determine the baseline range of the preset key process features; After the time when the anomaly occurs, the time when the anomaly is determined is determined by comparing each preset key process feature in chronological order to see if it is outside the baseline interval.

[0008] In one specific embodiment, determining the anomaly determination time by comparing each preset key process feature in chronological order to see if it is outside the baseline interval includes: When a predetermined number of consecutive key process features are outside the baseline range, the first moment in the time sequence of the predetermined number of key process features is determined as the anomaly determination moment.

[0009] In one specific embodiment, the multiple dimensions are: personnel changes, material switching, production equipment changes, and tooling replacement; The causal priorities of each dimension include the causal priorities of personnel changes, material switching, production equipment changes, and tooling replacements.

[0010] In one specific embodiment, the process of obtaining the causal priority of the personnel change includes: Determine the time interval between the time the anomaly occurred and the time the anomaly was determined; When the abnormal time interval is greater than a preset interval threshold, the percentage of preset key process features that are within the baseline interval is determined during the time period between the time of the abnormality occurrence and the time of the abnormality determination; the causal priority of the personnel change is obtained based on the abnormal time interval and the percentage of the number; the causal priority of the personnel change is negatively correlated with both the abnormal time interval and the percentage of the number. When the abnormal time interval is less than or equal to the preset interval threshold, the causal priority of the personnel change is obtained based on the abnormal time interval; the causal priority of the personnel change is negatively correlated with the abnormal time interval.

[0011] In one specific embodiment, the process of obtaining the causal priority of material switching, production equipment anomaly, and tooling replacement includes: The causal priority of the material switching, production equipment anomaly, and tooling replacement is obtained based on the abnormal time interval between the time of the anomaly occurrence and the time of the anomaly determination; the causal priority of the material switching, production equipment anomaly, and tooling replacement is negatively correlated with the abnormal time interval.

[0012] In one specific embodiment, the associated time period is a preset duration period with the time of the production quality abnormality as the cutoff time.

[0013] The present invention also provides an enterprise production big data mining system based on an ERP system, comprising: a memory and a processor; the memory is connected to the processor; the memory is used to store program instructions; the processor is used to implement the above-mentioned enterprise production big data mining method based on an ERP system when the program instructions are executed.

[0014] This invention offers the following technical advantages: It analyzes production anomaly events across multiple dimensions within a correlated timeframe during the occurrence of production quality anomalies. Compared to a one-sided analysis of a single dimension, this approach improves the reliability and comprehensiveness of data analysis, enabling the discovery of data closely related to the occurrence of production quality anomalies. This enhances the accuracy and reliability of data mining, allowing for the accurate identification of the true causes of production process anomalies. By identifying a strong candidate set of suspected factor combinations based on the temporal overlap between production anomaly events of different dimensions, irrelevant production anomaly events can be effectively eliminated, making data mining more targeted. Furthermore, based on the determined anomaly occurrence time and anomaly determination time, the causal priority of each dimension is obtained, thereby identifying the production anomaly events of the dimension that are the cause of the production quality anomaly. Ultimately, this improves the accuracy and reliability of data mining, accurately identifying the true causes of production process anomalies. Attached Figure Description

[0015] Figure 1 This is a flowchart of a method for mining enterprise production big data based on an ERP system, provided by the present invention. Detailed Implementation

[0016] This embodiment provides a big data mining method for enterprise production based on an ERP system. It is applicable to the product manufacturing process where, when production quality abnormalities occur, big data mining of production-related data can reveal the true cause of the abnormalities. In this embodiment, the specific product manufacturing scenario is an SMT production line in an electronic assembly workshop, producing main control boards.

[0017] The production quality of the main control boards on the production line is monitored in real time, and an anomaly alarm signal is output when an anomaly occurs. Specifically, an anomaly can occur when the defect rate of the main control boards exceeds a preset defect rate threshold (e.g., 3%). The defect rate can be either the batch defect rate or the sampling defect rate. The batch defect rate is calculated by dividing the number of defective main control boards in the current production batch by the total number of defective main control boards in the current production batch. The sampling defect rate is calculated by dividing the number of defective main control boards sampled from the current production batch by the total number of defective main control boards sampled. The sampling defect rate can be calculated before the current production batch is completed; therefore, to improve the timeliness of production quality anomaly monitoring, the sampling defect rate is used as an example here.

[0018] The moment when the defect rate of the main control board exceeds a preset defect rate threshold is defined as the moment of production quality anomaly, and the associated time period of the production quality anomaly is determined. The associated time period is a time window for tracing product quality, and is a period of time before the production quality anomaly moment. Specifically, the associated time period is a preset duration with the production quality anomaly moment as the cutoff time. The preset duration is set according to actual monitoring needs; this embodiment uses 4 hours as an example.

[0019] like Figure 1 As shown in the figure, the enterprise production big data mining method based on an ERP system provided in this embodiment includes the following steps: Step S1: Identify production anomaly events across multiple dimensions within the associated time period at the moment of production quality anomaly; Step S2: Determine the temporal overlap between any production anomaly events in each pair of dimensions; Step S3: Determine the suspected factor combination. The suspected factor combination consists of one of the production anomaly events in each dimension, and the time overlap of every two production anomaly events in the suspected factor combination meets the preset conditions. Step S4: Determine the time of occurrence and the time of determination of the anomaly based on the abnormal situation of the preset key process characteristics in the time sequence; Step S5: Based on the time of anomaly occurrence and the time of anomaly determination, determine the causal priority of each dimension in the combination of suspected factors; Step S6: Extract production anomaly events from the production anomaly dimension. The production anomaly dimension is the dimension corresponding to the highest causal priority.

[0020] Based on the overall approach of the above steps, the specific implementation process of each step is described below.

[0021] Step S1: Identify production anomaly events across multiple dimensions within the associated time period at the moment of production quality anomaly.

[0022] This system retrieves production-related data from multiple dimensions within the relevant time period at which production quality anomalies occur, based on the ERP system. In one specific embodiment, these dimensions include: personnel changes, material switching, production equipment anomalies, and tooling changes. Personnel changes refer to changes in personnel during the production process, such as operator onboarding, replacements, and shift changes. This retrieves at least one production anomaly event related to personnel changes within the relevant time period. Each anomaly event represents a change record for each personnel change, including: workstation ID, operator ID, and change time. Material switching refers to batch switching records for work orders during the production process. This retrieves at least one production anomaly event related to material switching within the relevant time period. Each anomaly event represents a switching record for each material switching, including: work order ID, process ID, old batch number, new batch number, and switching time. Production equipment anomalies refer to production equipment being in an abnormal state, particularly abnormal situations such as inspection anomalies, fault reports, and downtime records. This retrieves at least one production anomaly event related to production equipment anomalies within the relevant time period. Each anomaly event includes: production equipment ID and production equipment anomaly time. Tooling change refers to the replacement of tooling / molds, thereby obtaining at least one production anomaly event related to tooling change within the associated time period. Each production anomaly event includes: production equipment ID or workstation ID, old tooling ID, new tooling ID, and change time. The change time, switchover time, production equipment anomaly time, and change time mentioned above all refer to time periods, that is, the time period between the start and end time of the event, which is essentially the time period during which the event occurs.

[0023] Each production anomaly event involving personnel changes within the associated time period of a production quality anomaly is constructed into a personnel change event set; each production anomaly event involving material switching within the associated time period of a production quality anomaly is constructed into a material switching event set; each production anomaly event involving production equipment changes within the associated time period of a production quality anomaly is constructed into a production equipment change event set; and each production anomaly event involving tooling changes within the associated time period of a production quality anomaly is constructed into a tooling change event set.

[0024] It should be understood that if only some of the above-mentioned dimensions (and at least two dimensions) of production anomalies exist within the relevant time period, then the production anomalies of those dimensions will be included in the subsequent data analysis; if there are no production anomalies in all of the above-mentioned dimensions within the relevant time period, then no further data analysis will be performed, and a no-data-collection alarm signal will be output to facilitate staff in investigating the cause; if there is a production anomaly in one dimension within the relevant time period, then no further data analysis will be performed, and that dimension will be directly used as the production anomaly dimension, and the production anomalies of that dimension will be extracted.

[0025] Step S2: Determine the temporal overlap between any production anomaly events in each of the two dimensions.

[0026] When quality anomalies occur on the production floor, they are often the result of multiple adverse factors coupled over time. For example, if the production equipment malfunctions, experienced operators and stable material batches may prevent defective products from being produced. However, if personnel replacement and material batch switching occur simultaneously, the same equipment malfunction may exceed normal boundaries, leading to anomalies. Existing traceability methods typically only investigate along a single dimension, ignoring the cumulative effect between multiple dimensions and easily missing the true cause of the anomaly. Therefore, it is necessary to use the moment of the production quality anomaly as an anchor point to analyze the correlation between production anomaly events in various dimensions, thereby accurately identifying the cause of the anomaly.

[0027] For any two dimensions, one production anomaly event is randomly selected from each dimension, i.e., one production anomaly event is randomly selected from the event sets of each dimension, defined as the i-th production anomaly event and the j-th production anomaly event. The time overlap between the i-th and j-th production anomaly events is determined; the time overlap indicates the degree of overlap between the time periods of the events occurring in these two production anomaly events. In a specific embodiment, the time intervals (i.e., the event occurrence time periods) of the i-th and j-th production anomaly events are determined, and then the intersection of the event occurrence time periods of the i-th and j-th production anomaly events is determined, and the intersection duration is determined. Similarly, the union of the event occurrence time periods of the i-th and j-th production anomaly events is determined, and the union duration is determined. The intersection duration is divided by the union duration; the resulting ratio is the intersection-union ratio, which is used as the time overlap between the i-th and j-th production anomaly events. The larger the value of the time overlap, the closer the time overlap between the i-th and j-th production anomalies, and the stronger the temporal correlation between them. This yields the time overlap between any two production anomalies in each dimension.

[0028] Step S3: Determine the suspected factor combination. The suspected factor combination consists of one of the production anomaly events in each dimension, and the time overlap of every two production anomaly events in the suspected factor combination meets the preset conditions.

[0029] This embodiment sets a preset condition to determine whether the time overlap between any two production anomalies meets this preset condition. Meeting the preset condition indicates that the time overlap between the two production anomalies is at a high level. In a specific embodiment, this embodiment sets a preset time overlap threshold, the value of which ranges from 0 to 1. The specific value needs to be set based on actual judgment and experience. For example, if a stricter filtering logic is required, the preset time overlap threshold can be set slightly higher. As an example, this embodiment uses 0.5. Accordingly, meeting the preset condition specifically means that the time overlap is greater than or equal to the preset time overlap threshold.

[0030] If a production anomaly is found in each of the four dimensions, resulting in four production anomalies, and the temporal overlap between any two of these four anomalies satisfies a preset condition, then these four anomalies constitute a suspected factor combination. Therefore, a suspected factor combination consists of one production anomaly from each dimension, and the temporal overlap between any two production anomalies in the suspected factor combination satisfies the preset condition. This yields at least one suspected factor combination.

[0031] It should be noted that if the suspected factor combination cannot be obtained through the above processing, no further data processing will be performed, and an alarm signal indicating no suspected factor combination will be directly output to facilitate staff in investigating the cause.

[0032] Step S4: Determine the time of occurrence and the time of determination of the anomaly based on the abnormal situation of the preset key process characteristics in the time sequence.

[0033] By analyzing each suspected factor combination, the dimensions of production anomalies associated with production anomalies in each suspected factor combination are determined. The following analysis uses any one suspected factor combination as an example.

[0034] This embodiment defines a preset key process feature, which is a relatively critical production process parameter that affects production quality. In one specific embodiment, the production batches prior to the time of the production quality anomaly are determined from the ERP system. Specifically, the production batches prior to the time of the production quality anomaly are production batches that have the same product code, the same production line, and the same process route as the current production batch.

[0035] The combination of suspect factors can only describe the simultaneous existence of multiple production anomalies in time, but correlation does not equal causation. To determine which dimension is most likely to lead to the anomaly, a steady-state reference system for production must be introduced. Accordingly, a baseline normal production batch is determined from all production batches prior to the time of the production quality anomaly, serving as the steady-state reference system. The baseline normal production batch is a production batch with a very low defect rate, such as a batch defect rate below 1.5%. It should be understood that if there are at least two production batches with a defect rate below 1.5%, the production batch with the lowest defect rate is selected as the baseline normal production batch; conversely, if there are at least two production batches with the lowest defect rates, either one is randomly selected as the baseline normal production batch.

[0036] Analyze the production process parameters corresponding to the baseline normal production batch to determine the preset critical process characteristics. In one specific embodiment, since the material feeding interval significantly affects normal production, the material feeding interval is used as a preset critical process characteristic. The material feeding interval is the interval between two consecutive material feedings.

[0037] Since multiple material feedings are inevitably involved in the production process corresponding to the baseline normal production batch, there are also multiple material feeding intervals. Therefore, we obtain multiple material feeding intervals (i.e., numerical values ​​of the material feeding intervals) from the baseline normal production batch. Then, we calculate the average value and range (i.e., the difference between the maximum and minimum values) of these material feeding intervals. Based on the average value and range of the material feeding intervals, we determine the baseline range for the material feeding intervals. ; in, The baseline range representing the feeding interval is the baseline range for preset key process characteristics; This represents the average duration of the feeding interval; This represents the range of feeding intervals; This represents the minimum of these feeding intervals; `max` represents the maximum value function. The purpose of setting the maximum value function is to: when the range is too small, switch to using the minimum of the feeding intervals. The baseline interval is defined as half the width of the baseline range, thus providing the normal fluctuation range of the feeding interval. Therefore, the baseline interval for the feeding interval is based on the average value. Centered on, with The numerical range is half-width to cover all historical performance under normal fluctuations in feeding interval duration.

[0038] It should be noted that if the above processing cannot obtain a baseline normal production batch, the subsequent determination process of the preset key process characteristics and their baseline range will not be carried out. Instead, based on the historical production process, one of the production process parameters that affect production quality will be manually selected from several production process parameters by experience as the preset key process characteristic. The baseline range of the preset key process characteristic will be the standard value of the preset key process characteristic as the center and the preset value (e.g., 10% of the standard value) as the half-width.

[0039] When production quality anomalies occur, they are often accompanied by preset critical process characteristics first deviating from the baseline range. By tracing the time point of the deviation and identifying the production anomaly events preceding the deviation point, the causal priority can be determined based on the order in which the events occurred before the deviation.

[0040] For any production anomaly event in the suspected factor combination, the time when the production anomaly event occurs (i.e., the start time of the event's occurrence period) is taken as the anomaly occurrence time corresponding to that production anomaly event. Therefore, each production anomaly event in the suspected factor combination corresponds to an anomaly occurrence time, and the anomaly occurrence times of different production anomalies may be the same or different.

[0041] During the production process within the associated time period of a production quality anomaly, multiple material feedings are performed according to production needs and in a sequential order, resulting in multiple feeding intervals. The determination time for each feeding interval is also determined (the determination time is the time of the latter feeding between two adjacent feedings, i.e., the cutoff time corresponding to the feeding interval). Therefore, after the anomaly occurs, it is determined whether each feeding interval falls outside its baseline range. If it falls outside the baseline range, it indicates a deviation in production quality. If a predetermined number of consecutive feeding intervals fall outside the baseline range, it indicates a confirmed production quality deviation. The predetermined number of consecutive intervals is set according to actual needs; this embodiment uses 3 as an example. Then, the first feeding interval is determined from these predetermined number of consecutive feeding intervals, and the determination time corresponding to this first feeding interval is taken as the first time. This first time is designated as the anomaly determination time, i.e., the deviation start time. It should be noted that if, after the time of the anomaly, there are no consecutive preset number of feeding intervals outside the baseline range in terms of time sequence, then the time of the production quality anomaly can be taken as the time of anomaly determination.

[0042] Step S5: Based on the time of anomaly occurrence and the time of anomaly determination, determine the causal priority of each dimension in the suspected factor combination.

[0043] Since each production anomaly event in the suspected factor combination corresponds one-to-one with each dimension, the causal priority of each dimension in the suspected factor combination is determined based on the occurrence time of each production anomaly event and the determined anomaly time. Accordingly, the causal priorities of each dimension include the causal priority of personnel changes, material switching, production equipment anomalies, and tooling replacements.

[0044] For the causal priority of personnel changes, the production anomaly event corresponding to the personnel change dimension in the suspected factor combination is identified, and the time of occurrence of the anomaly event is determined. The time interval between the anomaly occurrence time and the anomaly determination time is defined as the anomaly time interval. In this embodiment, the time interval represents the duration between two corresponding moments. The causal priority of personnel changes needs to be obtained through comprehensive analysis combining the anomaly time interval and steady-state observations. Specifically: This embodiment sets a preset interval threshold for comparison with abnormal time intervals to determine whether the abnormal time interval is too long. If the abnormal time interval is greater than the preset interval threshold, it indicates that the production line experienced a relatively long period of steady-state operation before the deviation occurred after the personnel change, indicating a weak causal relationship between the personnel change and the deviation. Furthermore, the larger the abnormal time interval, the weaker the causal relationship between the personnel change and the deviation, and the lower the causal priority of the personnel change. The causal priority of the personnel change is negatively correlated with the abnormal time interval. If the abnormal time interval is less than or equal to the preset interval threshold, it indicates that the production line experienced a relatively short period of steady-state operation after the personnel change, indicating a strong causal relationship between the personnel change and the deviation. The preset interval threshold can be obtained by calibrating the time intervals in historical data, such as 30 minutes.

[0045] When the abnormal time interval exceeds a preset threshold, it is determined whether the duration of each feeding interval within the period between the time of the abnormality occurrence and the time of abnormality determination falls within the baseline interval. This determines the number of feeding interval durations within the baseline interval. Simultaneously, the total number of feeding interval durations within the period between the time of the abnormality occurrence and the time of abnormality determination is determined. The ratio of the number of feeding interval durations within the baseline interval to the total number of feeding interval durations is calculated as the percentage of feeding interval durations within the baseline interval. A larger percentage indicates a more stable production status within the period between the time of the abnormality occurrence and the time of abnormality determination, a weaker causal relationship between personnel changes and deviations, and a lower causal priority for personnel changes. The causal priority of personnel changes is negatively correlated with this percentage.

[0046] Therefore, based on the abnormal time intervals and the proportion of abnormal occurrences, the causal priority of personnel changes is obtained. Based on the logical analysis above, the following is one method for calculating the causal priority of personnel changes: ; in, Indicates the causal priority of personnel changes; Indicates an abnormal time interval; This represents the preset upper limit of the abnormal time interval. This upper limit of the abnormal time interval is used as the dividing line between whether there is a causal relationship between production abnormal events and offsets in various dimensions. This upper limit of the abnormal time interval can be obtained by experience or by statistical analysis of historical abnormal time intervals. In this embodiment, 1 hour is used as an example. This indicates the percentage of feeding intervals that fall within the baseline range; This represents the adjustment factor related to personnel changes.

[0047] When the abnormal time interval is less than or equal to a preset interval threshold, it indicates that the production line experienced a deviation after a short period of steady-state operation following the personnel change. This can be interpreted as the production anomaly event corresponding to the personnel change being a sudden event, and the causal priority of the personnel change is obtained solely based on the abnormal time interval. The causal priority of the personnel change is negatively correlated with the abnormal time interval. The following is one method for calculating the causal priority of the personnel change: .

[0048] For the causal priority of material switching, the production anomaly event corresponding to the material switching is a sudden event. The abnormal time interval between the occurrence time and the determination time of the anomaly event corresponding to the material switching is determined, and the causal priority of the material switching is obtained based on the abnormal time interval. The shorter the abnormal time interval, the stronger the causal relationship between the material switching and the offset, and the higher the causal priority of the material switching. The causal priority of the material switching is negatively correlated with the abnormal time interval. One method for calculating the causal priority of material switching is given below: ; in, Indicates the causal priority of material switching; This represents the adjustment coefficient related to material switching.

[0049] For the causal priority of tooling change, the production anomaly event corresponding to the tooling change is a sudden event. The causal priority of the tooling change is determined based on the time interval between the occurrence and determination of the anomaly event. The shorter the time interval, the stronger the causal relationship between the tooling change and the offset, and the higher the causal priority of the tooling change. The causal priority of the tooling change is negatively correlated with the time interval. One method for calculating the causal priority of tooling change is given below: ; in, Indicates the causal priority of tooling changes; This indicates the adjustment coefficient related to tooling changes.

[0050] To determine the causal priority of production equipment anomalies, the status change records of the production equipment are first extracted from the ERP system. The start time of the fault or abnormal state (i.e., the start time of the production equipment abnormality time mentioned above) is taken as the occurrence time of the production abnormal event corresponding to the production equipment anomaly. The production abnormal event corresponding to the production equipment anomaly is a sudden event. The abnormal time interval between the occurrence time of the production abnormal event corresponding to the production equipment anomaly and the abnormal determination time is determined, and the causal priority of the production equipment anomaly is obtained based on the abnormal time interval. The shorter the abnormal time interval, the stronger the causal correlation between the production equipment anomaly and the offset, and the higher the causal priority of the production equipment anomaly. The causal priority of the production equipment anomaly is negatively correlated with the abnormal time interval. The following is one method for calculating the causal priority of the production equipment anomaly: ; in, Indicates the causal priority of changes in production equipment; This represents the adjustment coefficient related to changes in production equipment.

[0051] It should be noted that in the calculation method of the causal priority of each dimension mentioned above, if the abnormal time interval is greater than the upper limit of the abnormal time interval, it means that the time interval between the occurrence time of the abnormal production event of the corresponding dimension and the occurrence offset is too long. It is determined that there is no causal relationship between the production abnormal event and the offset of the corresponding dimension, and the causal priority of the corresponding dimension is directly set to 0.

[0052] The adjustment coefficients related to personnel changes mentioned above Adjustment coefficients related to material switching Adjustment coefficients related to tooling changes Adjustment coefficients related to production equipment anomalies The adjustment coefficient is determined based on the strength of the causal relationship causing the shift in each dimension; the stronger the causal relationship, the larger the value. In one specific embodiment, the adjustment coefficient is set to a range of 0.8 to 1.2. The strength of the causal relationship causing the shift in each dimension can be set by the operator as needed. In one specific embodiment, since abrupt events are more likely to cause shifts, the adjustment coefficient related to abrupt events is slightly larger. Therefore, the adjustment coefficient related to material switching... Adjustment coefficients related to tooling changes Adjustment coefficients related to production equipment anomalies Greater than the adjustment coefficient related to personnel changes Adjustment coefficient related to material switching Adjustment coefficients related to tooling changes Adjustment coefficients related to production equipment anomalies They can be equal, as an example: adjustment coefficients related to material switching. Adjustment coefficients related to tooling changes Adjustment coefficients related to production equipment anomalies Set to 1.1, the adjustment factor related to personnel changes. The value is set to 0.9. Alternatively, without considering the strength of the causal relationship causing the shift in each dimension, all four adjustment coefficients can be left unset, i.e., all four adjustment coefficients can be set to a value of 1. This yields the causal priority of each dimension in the suspected factor combination.

[0053] It should be noted that for a certain dimension, if in an extreme case the time of anomaly determination is before the time of anomaly occurrence corresponding to that dimension, this is considered an abnormal situation. In this case, the causal priority of that dimension will no longer be calculated, but the causal priority of that dimension will be directly set to 0.

[0054] Step S6: Extract production anomaly events from the production anomaly dimension. The production anomaly dimension is the dimension corresponding to the highest causal priority.

[0055] For the causal priority of each dimension in the suspected factor combination, the higher the causal priority, the more likely the corresponding production anomaly event is to be the root cause of the production quality anomaly. Therefore, the highest causal priority is determined from the four dimensions of the suspected factor combination, and the dimension corresponding to the highest causal priority is identified as the production anomaly dimension, thereby extracting the production anomaly events of the production anomaly dimension in the suspected factor combination. Using the above method, the production anomaly events of the production anomaly dimension in each suspected factor combination within the associated time period of the production quality anomaly are obtained and transmitted to the ERP system to realize the mining of enterprise production big data. Subsequently, based on the production anomaly events of the production anomaly dimension in each suspected factor combination within the associated time period of the production quality anomaly, the production quality anomaly situation can be further analyzed, and corresponding production process parameters can be adjusted to optimize subsequent production quality.

[0056] It should be noted that if there are at least two highest causal priorities, then all dimensions corresponding to the highest causal priorities will be considered as production anomaly dimensions. Conversely, if the causal priority of all dimensions is 0, then the production anomaly dimension will not be determined; instead, a causal priority anomaly alarm signal will be directly output to facilitate staff investigation of the cause.

[0057] In addition, this embodiment also provides an enterprise production big data mining system based on an ERP system, including: a memory and a processor; the memory is connected to the processor, and the memory is used to store program instructions; the processor is used to implement the steps in the above-described embodiment of the enterprise production big data mining method based on an ERP system when the program instructions are executed.

Claims

1. A method for mining enterprise production big data based on an ERP system, characterized in that, include: Identify production anomaly events across multiple dimensions within the associated timeframe at the moment of production quality anomaly. Determine the temporal overlap between any production anomaly events in each pair of dimensions; Determine a combination of suspicious factors, wherein the combination of suspicious factors consists of one of the production anomalies in each dimension, and the time overlap of every two production anomalies in the combination of suspicious factors meets a preset condition. The timing of anomalies and the timing of anomaly determination are determined by the abnormal situations in the time sequence of pre-defined key process characteristics; Based on the time of occurrence of the anomaly and the time of determination of the anomaly, the causal priority of each dimension in the combination of suspect factors is determined. Extract production anomaly events from the production anomaly dimension, where the production anomaly dimension corresponds to the dimension with the highest causal priority.

2. The method for mining enterprise production big data based on an ERP system according to claim 1, characterized in that, The process of obtaining the time overlap includes: The intersection and union durations of the time intervals of two candidate production anomaly events are determined respectively; the two candidate production anomaly events are any production anomaly events in any two dimensions. The time overlap of the two candidate production anomalies is obtained by calculating the ratio of the intersection duration to the union duration.

3. The method for mining enterprise production big data based on an ERP system according to claim 1, characterized in that, The time overlap meets the preset conditions, including: the time overlap is greater than or equal to the preset time overlap threshold.

4. The method for mining enterprise production big data based on an ERP system according to claim 1, characterized in that, The process of acquiring the preset key process features includes: Determine the baseline normal production batch from all production batches prior to the time of the aforementioned production quality anomaly. Preset key process characteristics are determined based on the benchmark normal production batch.

5. The method for mining enterprise production big data based on an ERP system according to claim 1, characterized in that, The time of the anomaly is the moment when the production anomaly event occurs. The process of obtaining the anomaly determination time includes: Determine the baseline range of the preset key process features; After the time when the anomaly occurs, the time when the anomaly is determined is determined by comparing each preset key process feature in chronological order to see if it is outside the baseline interval.

6. The method for mining enterprise production big data based on an ERP system according to claim 5, characterized in that, The step of determining the anomaly determination time by comparing each preset key process feature in chronological order to see if it is outside the baseline interval includes: When a predetermined number of consecutive key process features are outside the baseline range, the first moment in the time sequence of the predetermined number of key process features is determined as the anomaly determination moment.

7. The method for mining enterprise production big data based on an ERP system according to claim 5, characterized in that, The multiple dimensions are: personnel changes, material switching, production equipment changes, and tooling replacement; The causal priorities of each dimension include the causal priorities of personnel changes, material switching, production equipment changes, and tooling replacements.

8. The method for mining enterprise production big data based on an ERP system according to claim 7, characterized in that, The process of obtaining the causal priority of the personnel change includes: Determine the time interval between the time the anomaly occurred and the time the anomaly was determined; When the abnormal time interval is greater than a preset interval threshold, the percentage of preset key process features that are within the baseline interval is determined during the time period between the time of the abnormality occurrence and the time of the abnormality determination; the causal priority of the personnel change is obtained based on the abnormal time interval and the percentage of the number; the causal priority of the personnel change is negatively correlated with both the abnormal time interval and the percentage of the number. When the abnormal time interval is less than or equal to the preset interval threshold, the causal priority of the personnel change is obtained based on the abnormal time interval; the causal priority of the personnel change is negatively correlated with the abnormal time interval.

9. The method for mining enterprise production big data based on an ERP system according to claim 7, characterized in that, The process of obtaining the causal priority of material switching, production equipment anomalies, and tooling replacements includes: The causal priority of the material switching, production equipment anomaly, and tooling replacement is obtained based on the abnormal time interval between the time of the anomaly occurrence and the time of the anomaly determination; the causal priority of the material switching, production equipment anomaly, and tooling replacement is negatively correlated with the abnormal time interval.

10. The method for mining enterprise production big data based on an ERP system according to claim 1, characterized in that, The associated time period is a preset duration period with the time of the production quality abnormality as the cutoff time.

Images