Fault management system for automobile wheel manufacturing process

Abstract

The invention discloses a fault management system used in an automobile wheel manufacturing process, and relates to the technical field of manufacturing industry informatization. A record analysis unit is set to record processing data, cleaning data and processing data generated when a data cleaning unit cleans data; the model diagnosis unit analyzes diagnosis data generated when fault causes are diagnosed, and any dimension parameter of any type of fault causes is analyzed in the analysis process; the new backtracking duration of any dimension parameter in any kind of faults is determined based on the collection duration corresponding to the missing value and the abnormal value in the transmission process, and backtracking information is updated in sequence, so that the capacity of multi-dimension data collected and transmitted when the faults occur is effectively reduced, and the bandwidth consumption and the storage cost are reduced; and meanwhile, the effectiveness of data transmission is improved, the cloud data cleaning efficiency is improved, the data volume input into the fault diagnosis model is reduced, and the efficiency of the whole fault diagnosis process is optimized.

Description

A fault management system for automobile wheel manufacturing process Technical Field

[0001] This invention relates to the field of information technology in manufacturing, specifically a fault management system for the automobile wheel manufacturing process. Background Technology

[0002] With the rapid development of the automotive industry, the manufacturing quality of wheels, as a core component of the vehicle's driving system, directly affects the safety and reliability of the vehicle. The manufacturing of automotive wheels involves multiple processes such as forging, welding, machining, and painting, and is susceptible to various faults such as out-of-tolerance wheel hub dimensions, welding cracks, and surface scratches due to factors such as equipment condition fluctuations, process parameter deviations, and environmental interference. In order to achieve rapid fault location and handling, the industry generally adopts a data-driven fault management system. When a fault is detected, this type of system will automatically collect all time-series data within a preset backtracking time period, such as welding current, lathe speed, equipment vibration value, and ambient temperature and humidity, based on the pre-selected monitoring parameters according to the fault type, and upload them to the cloud. The cloud then uses a pre-trained fault diagnosis model to analyze the cause of the fault.

[0003] However, in order to ensure the accuracy of fault cause analysis, existing systems usually set a long backtracking time. This not only leads to a large amount of redundant information unrelated to the fault in the transmitted data, increasing bandwidth consumption and storage costs, but also reduces the computing efficiency of cloud diagnostic models. At the same time, the massive amount of data is affected by factors such as industrial network fluctuations and transmission link interruptions during transmission, significantly increasing the risk of data loss and anomalies. Furthermore, excessively long backtracking data will reduce the efficiency of cloud data cleaning.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] The purpose of this invention is to provide a fault management system for the manufacturing process of automobile wheels, in order to solve the problems mentioned in the background art.

[0006] This invention provides a fault management system for the automotive wheel manufacturing process, comprising:

[0007] The fault monitoring module is used to obtain multi-dimensional data of a certain type of fault when any of the preset fault types is detected during the manufacturing process of the target automobile wheel, based on the backtracking information of the fault type stored in the pre-stored backtracking information. The backtracking information includes several dimension parameters and their backtracking duration.

[0008] A cloud-based analytics platform is used to analyze multi-dimensional data of the aforementioned fault types after receiving such data. The cloud-based analytics platform includes a data cleaning unit, a model diagnosis unit, and a record analysis unit.

[0009] The data cleaning unit is used to detect the multi-dimensional data using an anomaly detection algorithm after receiving the multi-dimensional data. If invalid data is detected, the invalid data is deleted from the multi-dimensional data and the multi-dimensional data is completed using a preset completion strategy to obtain the cleaned data of the type of fault. The invalid data is used as the processing data of the type of fault.

[0010] The model diagnostic unit is used to input the cleaning data into the fault diagnosis model after receiving the cleaning data and processing data of the fault type, obtain the output diagnostic data of the fault type, the diagnostic data includes the fault cause and diagnosis time of the fault type, and summarize the diagnostic data, processing data and cleaning data of the fault type to generate the record data of the fault type.

[0011] The recording and analysis unit is used to store the recorded data of the type of fault after receiving it. For any type of fault, the recording and analysis unit analyzes the recorded data of the type of fault stored in it after reaching a fixed amount, and generates updated data of the type of fault.

[0012] The fault monitoring module is also used to update the backtracking information of the fault type stored in it after receiving the updated data.

[0013] Furthermore, the fault monitoring module is also used to collect and store the monitoring values ​​of all dimensional parameters preset by the administrator in real time.

[0014] Furthermore, the multi-dimensional data of the fault type includes time-series data of all dimension parameters included in the backtracking information of the fault type, and the time-series data of each dimension parameter includes all monitoring values ​​of the dimension parameter and their collection times within its corresponding backtracking duration.

[0015] Furthermore, the invalid data refers to missing and outlier values ​​generated by the fault monitoring module during the real-time collection of monitoring values ​​of all dimensions of parameters and the transmission to the cloud analysis platform.

[0016] Furthermore, for missing values ​​in invalid data, the completion strategy handles missing values ​​through deletion, filling, interpolation, and prediction model methods. For outliers in invalid data, the completion strategy handles outliers through deletion, replacement, and transformation methods.

[0017] Furthermore, for any type of fault, when the recorded data of the fault type stored in the recording and analysis unit reaches a preset fixed amount, the following steps are taken for analysis:

[0018] S11: Obtain all the recorded data of the fault type stored in the storage, extract the fault cause from each of them and remove duplicates, and mark all the remaining fault causes after deduplication as B1, B2, ..., Bb, where b≥1;

[0019] S12: Obtain all record data containing the fault cause B1 from all the recorded data, and label them as C1, C2, ..., Cc, where c≥1;

[0020] S13: Perform deviation analysis on the diagnostic time of recorded data C1, C2, ..., Cc to obtain the baseline time F1 of fault cause B1;

[0021] S14: Iterate through the recorded data C1, C2, ..., Cc, obtain the dimension parameters corresponding to all monitoring values ​​contained in the recorded data C1, C2, ..., Cc, and deduplicate all the obtained dimension parameters. Label all the remaining dimension parameters after deduplication as G1, G2, ..., Gg, where g≥1;

[0022] S15: Create an empty redundant set of dimension parameter G1 relative to the fault cause B1. Extract the acquisition durations with a frequency greater than or equal to P1 and belonging to dimension parameter G1 from the processing data contained in the record data C1, C2, ..., Cc, and remove duplicates. Mark all the remaining acquisition durations in ascending order of value as H1, H2, ..., Hh, where h≥1, and P1 is the preset standard analysis frequency.

[0023] S16: Select whether to add the acquisition duration H1 to the redundant set according to the preset selection rules;

[0024] S17: Following S16, select whether to add the collection durations H2, H3, ..., Hh to the redundancy set of dimension parameter G1, and obtain the final redundancy set of dimension parameter G1 relative to the fault cause B1.

[0025] S18: Extract the acquisition duration with the largest value from the redundant set as the tracing duration of the dimension parameter G1 relative to the fault cause B1.

[0026] S19: Calculate and obtain the backtracking time of dimension parameters G2, G3, ..., Gg relative to the fault cause B1 in sequence according to S11 to S18, and generate the analysis set of fault cause B1 based on the backtracking time of dimension parameters G1, G2, ..., Gg relative to the fault cause B1;

[0027] S110: Generate analysis sets of fault causes B2, B3, ..., Bb in sequence according to S11 to S19; summarize and compare the analysis sets of fault causes B1, B2, ..., Bb to generate updated data for the aforementioned fault type; generate updated data for the aforementioned fault type based on the new backtracking duration of all dimension parameters.

[0028] Furthermore, in step S16, the selection rules for whether to add the acquisition duration H1 to the redundancy set are as follows:

[0029] S161: Obtain all recorded data containing the collection duration H1 within the processed data of recorded data C1, C2, ..., Cc, and relabel them as I1, I2, ..., Ii, i≥1 respectively;

[0030] S162: Determine whether the collection duration H1 is a redundant duration based on the recorded data I1 according to the preset judgment rules. The judgment steps are as follows:

[0031] SS11: Extract the cleaned data contained in the recorded data I1, and remove all the monitoring values ​​and their collection durations that were collected before the monitoring values ​​that were collected before the collection duration H1 from all the monitoring values ​​belonging to the dimension parameter G1 contained in the cleaned data, so as to obtain the preliminary data of the recorded data I1 based on the collection duration H1.

[0032] SS12: Input the preparatory data into the fault diagnosis model, obtain the diagnostic data output by the fault diagnosis model, perform a first determination on the diagnostic data and the fault cause contained in the recorded data I1, determine whether the acquisition time H1 has an impact on the fault cause based on the first determination result, if the first determination result determines that the acquisition time H1 has no impact on the fault cause, perform a second determination on the diagnostic time in the diagnostic data and the baseline time F1, and determine whether the acquisition time H1 has an adverse impact on the fault time based on the second determination result;

[0033] SS13: If the first judgment result is that the acquisition duration H1 has no effect on the cause of the fault, and the second judgment result is that the acquisition duration H1 has no adverse effect on the fault duration, then the acquisition duration H1 is determined to be a redundant duration based on the recorded data I1; otherwise, the acquisition duration H1 is determined to be a non-redundant duration based on the recorded data I1.

[0034] S163: Following the steps from S161 to S162, determine whether the acquisition duration H1 is a redundant duration based on the recorded data I2, I3, ..., Ii. If the acquisition duration H1 is a redundant duration based on the recorded data I1, I2, ..., Ii, then select to add the acquisition duration H1 to the redundant set of dimension parameter G1; otherwise, select not to add the acquisition duration H1 to the redundant set of dimension parameter G1 relative to the fault cause B1.

[0035] Furthermore, in SS12, the first determination is as follows: extract the fault causes in the diagnostic data and the fault causes in the recorded data I1 respectively, compare the consistency of the two extracted fault causes, if the two fault causes are consistent, it is determined that the acquisition duration H1 has no effect on the fault cause, otherwise it is determined that the acquisition duration H1 has an effect on the fault cause, and at this time it is further determined that the acquisition duration H1 is not a redundant duration based on the recorded data I1.

[0036] Furthermore, in SS12, the second determination is as follows: extract the diagnostic time in the diagnostic data, calculate the absolute value of the difference between it and the baseline time F1, compare the absolute value with P2, if the absolute value is less than or equal to P2, then it is determined that the acquisition time H1 has no adverse effect on the fault time, otherwise it is determined that the acquisition time H1 has an adverse effect on the fault time, and P2 is a preset adverse time determination threshold.

[0037] Compared with existing technologies, it has the following advantages:

[0038] This invention involves setting up a fault monitoring module during the manufacturing process of a target automobile wheel. When any of the preset types of faults is detected, the module acquires multi-dimensional data of the fault. A data cleaning unit cleans the multi-dimensional data, removes invalid data, and completes the data. A model diagnosis unit diagnoses the cause of the fault.

[0039] This invention sets up a recording and analysis unit to analyze the processed data and cleaned data generated by the data cleaning unit during data cleaning, as well as the diagnostic data generated by the model diagnosis unit during fault cause diagnosis. During the analysis, for any dimension parameter of any type of fault cause, based on the acquisition time corresponding to the missing and outlier values ​​during transmission, a new backtracking time is determined for any dimension parameter of any type of fault, and the backtracking information is updated sequentially. In this way, while ensuring the accuracy of fault cause analysis, the amount of multi-dimensional data collected and transmitted when a fault occurs is effectively reduced, bandwidth consumption and storage costs are reduced, and the effectiveness of transmitted data is improved through data cleaning and completion strategies, the efficiency of cloud data cleaning is improved, and the amount of data input to the fault diagnosis model is reduced, thus optimizing the efficiency of the overall fault diagnosis process. Attached Figure Description

[0040] Figure 1 is a system block diagram of the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] Please refer to Figure 1. This application provides a fault management system for the automotive wheel manufacturing process, including a fault monitoring module and a cloud analysis platform.

[0043] The fault monitoring module is used to, during the manufacturing process of the target automobile wheel, when any of the preset types of faults is detected, acquire multi-dimensional data of the type of fault based on the backtracking information of the type of fault stored in the pre-stored database, and transmit the multi-dimensional data to the cloud analysis platform. The backtracking information includes the backtracking duration of several dimension parameters, the multi-dimensional data includes the time series data of several dimension parameters, and the time series data of each dimension parameter includes several monitoring values ​​of the dimension parameter and their acquisition time.

[0044] It should be noted that the fault monitoring module collects and stores the monitoring values ​​of all dimensions of parameters preset by the administrator in real time;

[0045] Based on the pre-stored backtracking information of the aforementioned fault type, the multi-dimensional data of the aforementioned fault type is obtained as follows:

[0046] For the backtracking duration of any dimension parameter included in the backtracking information of the aforementioned type of fault, the end point of the dimension parameter at the time of occurrence of the aforementioned type of fault is calculated by taking the time of occurrence of the aforementioned type of fault as the starting point and combining it with the backtracking duration. The formula for calculating the backtracking end point is the starting point minus the backtracking duration.

[0047] All monitoring values ​​of the dimension parameter collected at all times between the start point and the end point are obtained, including the monitoring values ​​of the dimension parameter collected at the times corresponding to the start point and the end point, to obtain the time series data of the dimension parameter;

[0048] In this application, the pre-defined types of faults are pre-set by managers based on historical experience and process knowledge. The necessity of this setting is that the causes of faults in the target automobile wheel manufacturing process are complex and diverse. By pre-defining fault types, it is intended to provide a clear diagnostic direction for the cloud analysis platform, enabling it to analyze and locate the root cause of each fault based on the collected multi-dimensional data.

[0049] In this application, the initial backtracking duration of each dimension parameter is preset by the administrator, and is defined as: the time range for collecting the monitoring values ​​of the dimension parameter backtracking backward from the time of the fault occurrence.

[0050] In this application, all dimension parameters included in the retrospective information of any type of fault are pre-selected by the management personnel, encompassing dimension parameters for analyzing all fault causes of the aforementioned fault, and the parameter types of all dimension parameters included in the retrospective information of any type of fault are not completely the same. In this application, the parameter types of dimension parameters include, but are not limited to, process parameter types, equipment status parameter types, environmental parameter types, etc.

[0051] It should be noted that there are several causes of any type of failure.

[0052] Among them, the dimensional parameters belonging to the process parameter type include, but are not limited to, welding current, cutting speed, etc.; the dimensional parameters belonging to the equipment status parameter type include, but are not limited to, motor vibration value, bearing temperature, etc.; and the dimensional parameters belonging to the environmental parameter type include, but are not limited to, workshop temperature and humidity, air source pressure, etc.

[0053] The cloud-based analytics platform is used to analyze various faults that occur during the manufacturing process of target automobile wheels in the cloud. The cloud-based analytics platform includes a data cleaning unit, a model diagnostic unit, and a record analysis unit.

[0054] After receiving the multi-dimensional data of the aforementioned fault type, the cloud analysis platform transmits it to the data cleaning unit. The data cleaning unit pre-stores a completion strategy to process missing and outlier values ​​in the multi-dimensional data.

[0055] After receiving the transmitted multi-dimensional data of the type of fault, the data cleaning unit first uses an anomaly detection algorithm to detect the multi-dimensional data. If invalid data is detected during the detection process, the invalid data is deleted from the multi-dimensional data and a preset completion strategy is used to complete the multi-dimensional data. During the completion process, the processing data of the type of fault is generated based on the removed invalid data. The multi-dimensional data obtained after completion is used as the cleaning data of the type of fault. The invalid data contains several monitoring values ​​and their collection durations. The collection duration of any monitoring value is calculated based on the collection time of the monitoring value and the interval between the dimension parameter to which the monitoring value belongs in the multi-dimensional data and the most recent collection time. In this application, the unit of collection duration is seconds.

[0056] For example, if a monitoring value is collected at 10:15:03 AM, and the nearest collection time for the dimension parameter to which the monitoring value belongs in the multi-dimensional data is 10:18:03 AM, then the collection duration of the monitoring value is 180 seconds.

[0057] It should be noted that the dimension parameters corresponding to all the monitoring values ​​contained in the invalid data are not exactly the same. For example, based on dimension parameter A1, the invalid data contains several monitoring values ​​of dimension parameter A1, and based on dimension parameter A2, the invalid data also contains several monitoring values ​​of dimension parameter A2.

[0058] Invalid data refers to missing and outlier values ​​generated during the real-time collection of monitoring values ​​of all dimensions of parameters by the fault monitoring module and the transmission to the cloud analysis platform.

[0059] It should be noted here that the anomaly detection algorithm can be one of the Z-Score method or the box plot method;

[0060] For missing values ​​in invalid data, the completion strategy handles missing values ​​through methods such as deletion, filling, interpolation, and prediction models. Here, deletion refers to directly deleting missing values ​​and their collection times, filling refers to filling with the mean, median, or mode, and prediction models refer to using regression models, KNN models, etc. to predict missing values.

[0061] For outliers in invalid data, the completion strategy handles outliers through methods such as deletion, replacement, and transformation. Here, deletion refers to directly deleting outliers and their acquisition time; replacement refers to replacing outliers with upper and lower limits to reduce their impact; and transformation refers to logarithmic or binning transformation to reduce the impact of outliers.

[0062] The data cleaning unit transmits the processed data and cleaned data of the fault types to the model diagnosis unit. The model diagnosis unit has a pre-trained fault diagnosis model stored in it, which is used to analyze the causes of several preset types of faults that occur during the manufacturing process of the target wheel.

[0063] After receiving the transmitted processing data and cleaning data of the type of fault, the model diagnosis unit inputs the cleaning data into the fault diagnosis model and obtains the diagnostic data of the type of fault output by the fault diagnosis model. The diagnostic data includes the fault cause of the type of fault and the diagnosis time. The diagnosis time starts from the moment the cleaning data is input into the fault diagnosis model and ends at the moment the diagnostic data is obtained.

[0064] The model diagnostic unit sends the diagnostic data of the type of fault to the electronic device of the administrator, who then processes the type of fault. The electronic device can be any one of a smartphone, laptop, or desktop computer.

[0065] The model diagnostic unit summarizes the diagnostic data of the type of fault, as well as the received processing data and cleaning data of the type of fault, to generate the record data of the type of fault. At this time, the record data not only includes the diagnostic data, but also the processing data and the cleaning data.

[0066] The model diagnostic unit transmits the recorded data of the fault type to the record analysis unit for storage;

[0067] For any type of fault, the recording and analysis unit analyzes the fault when the recorded data of the fault type stored in its internal storage reaches a preset fixed amount. The fixed amount is set by the management personnel based on the fault type to ensure that there is a sufficient amount of recorded data to fully cover the different manifestations, triggering scenarios and impact dimensions of the fault type, and to avoid the analysis conclusions being one-sided or biased due to insufficient sample size.

[0068] The analysis steps for analyzing recorded data of any type of fault after reaching a preset fixed amount are as follows:

[0069] S11: Obtain all the stored record data, extract the fault causes from them respectively, remove duplicates from all the extracted fault causes, and mark all the remaining fault causes after deduplication as B1, B2, ..., Bb, where b≥1. It should be noted that deduplication only means keeping only one of the same fault causes.

[0070] S12: Obtain all record data containing the fault cause B1 from all the recorded data, and label them as C1, C2, ..., Cc, where c≥1;

[0071] S13: Perform deviation analysis on the diagnostic time of recorded data C1, C2, ..., Cc to obtain the baseline time F1 for the cause of failure B1. The deviation analysis is as follows:

[0072] Retrieve the diagnostic time contained in the recorded data C1, C2, ..., Cc, and label them as D1, D2, ..., Dc respectively;

[0073] Using formula , calculate the deviation E1 of the diagnostic times D1, D2, ..., Dc when 1 ≤ d ≤ c, and compare the magnitudes of E1 and E. In the formula, D is the average value of Dd at this time, Dd represents each of the diagnostic times D1, D2, ..., Dc, and E is the standard deviation threshold of the preset time based on the fault cause B1;

[0074] If E1 ≥ E, then delete the corresponding Dd in descending order of |Dd - D| and calculate the deviation E1 of the remaining Dd. Compare the magnitude of the deviation E1 at this time with E again until E1 < E. Obtain the average value of all the remaining diagnostic times participating in the calculation of the deviation E1 at this time, and calibrate the average value as the reference time F1 of the fault cause B1;

[0075] S14: Traverse the recorded data C1, C2, ..., Cc, obtain the dimensional parameters corresponding to all the monitored values included in the recorded data C1, C2, ..., Cc, and remove duplicates from all the obtained dimensional parameters. Mark all the remaining dimensional parameters after duplicate removal as G1, G2, ..., Gg, where g ≥ 1;

[0076] S15: Create an empty redundancy set of the dimensional parameter G1 relative to the fault cause B1. Extract and remove duplicates from the acquisition durations in the processed data included in the recorded data C1, C2, ..., Cc whose frequency of occurrence of the acquisition duration is greater than or equal to P1 and belongs to the dimensional parameter G1. Mark all the remaining acquisition durations after duplicate removal in ascending order of value as H1, H2, ..., Hh, where h ≥ 1, and P1 is the preset standard analysis frequency;

[0077] Among them, the frequency of any acquisition duration refers to the total number of the processed data included in the recorded data C1, C2, ..., Cc that includes the acquisition duration;

[0078] S16: Select whether to add the acquisition duration H1 to the redundancy set according to the preset selection rule. The selection rule is as follows:

[0079] S161: Obtain all the recorded data in the processed data included in the recorded data C1, C2, ..., Cc that includes the acquisition duration H1, and re-mark them as I1, I2, ..., Ii, where i ≥ 1;

[0080] S162: Determine whether the acquisition duration H1 is a redundant duration based on the recorded data I1 according to the preset determination rule. The determination steps are as follows:

[0081] SS11: Extract the cleaned data contained in the recorded data I1, and remove all the monitoring values ​​and their collection durations that were collected before the monitoring values ​​that were collected before the collection duration H1 from all the monitoring values ​​belonging to the dimension parameter G1 contained in the cleaned data, so as to obtain the preliminary data of the recorded data I1 based on the collection duration H1.

[0082] For example, if the collection time H1 is 10:00 AM, then whether it is 9:00 AM, 9:30 AM, or an earlier collection time, as long as it is before 10:00 AM, it falls within this range, and the monitoring values ​​collected before 10:00 AM need to be removed.

[0083] SS12: Input the preparatory data into the fault diagnosis model, obtain the diagnostic data output by the fault diagnosis model, perform a first determination on the diagnostic data and the fault cause contained in the recorded data I1, determine whether the acquisition time H1 has an impact on the fault cause based on the first determination result, if the first determination result determines that the acquisition time H1 has no impact on the fault cause, perform a second determination on the diagnostic time in the diagnostic data and the baseline time F1, and determine whether the acquisition time H1 has an adverse impact on the fault time based on the second determination result;

[0084] The first determination is as follows: extract the fault causes in the diagnostic data and the fault causes in the recorded data I1 respectively, and compare the consistency of the two extracted fault causes. If the two fault causes are consistent, it is determined that the acquisition duration H1 has no effect on the fault cause; otherwise, it is determined that the acquisition duration H1 has an effect on the fault cause. At this time, it is further determined that the acquisition duration H1 is not a redundant duration based on the recorded data I1. The second determination is as follows: extract the diagnostic time in the diagnostic data, calculate the absolute value of the difference between it and the baseline time F1, and compare the absolute value with P2. If the absolute value is less than or equal to P2, it is determined that the acquisition duration H1 has no adverse effect on the fault time; otherwise, it is determined that the acquisition duration H1 has an adverse effect on the fault time. P2 is a preset adverse time determination threshold.

[0085] SS13: Determine whether the acquisition duration H1 based on the recorded data I1 is a redundant duration based on the first and second determination results, as follows: If the first determination result is that the acquisition duration H1 has no impact on the cause of the fault, and the second determination result is that the acquisition duration H1 has no adverse impact on the fault consumption time, then the acquisition duration H1 based on the recorded data I1 is determined to be a redundant duration; otherwise, the acquisition duration H1 based on the recorded data I1 is determined to be a non-redundant duration.

[0086] S163: Following the steps from S161 to S162, determine whether the acquisition duration H1 is a redundant duration based on the recorded data I2, I3, ..., Ii. If the acquisition duration H1 is a redundant duration based on the recorded data I1, I2, ..., Ii, then select to add the acquisition duration H1 to the redundant set of dimension parameter G1; otherwise, select not to add the acquisition duration H1 to the redundant set of dimension parameter G1 relative to the fault cause B1.

[0087] It should be noted that in SS12, if it is determined that the acquisition duration H1 is not a redundant duration based on the recorded data I1, then the determination of whether the acquisition duration H1 is a redundant duration based on the recorded data I2, I3, ..., Ii will stop, and the acquisition duration H1 will not be added to the redundant set at this time.

[0088] S17: Following S16, select whether to add the collection durations H2, H3, ..., Hh to the redundancy set of dimension parameter G1, and obtain the final redundancy set of dimension parameter G1 relative to the fault cause B1.

[0089] S18: Extract the acquisition duration with the largest value from the redundant set as the tracing duration of the dimension parameter G1 relative to the fault cause B1.

[0090] S19: Calculate and obtain the backtracking duration of dimension parameters G2, G3, ..., Gg relative to the fault cause B1 in sequence according to S11 to S18. Generate an analysis set of fault cause B1 based on the backtracking duration of dimension parameters G1, G2, ..., Gg relative to the fault cause B1. The analysis set contains dimension parameters G1, G2, ..., Gg and their backtracking duration relative to the fault cause B1.

[0091] S110: Generate analysis sets for fault causes B2, B3, ..., Bb sequentially according to S11 to S19. Summarize and compare the analysis sets for fault causes B1, B2, ..., Bb to generate updated data for the aforementioned fault type. The summary, comparison, and generation content are as follows: For each dimension parameter in the analysis sets for fault causes B1, B2, ..., Bb, extract all backtracking durations of the dimension parameter from the analysis sets for fault causes B1, B2, ..., Bb. Select the backtracking duration with the largest value as the new backtracking duration of the dimension parameter. Generate updated data for the aforementioned fault type based on the new backtracking durations of all dimension parameters.

[0092] The recording and analysis unit transmits the updated data of the fault type to the fault monitoring module. Upon receiving the transmitted updated data of the fault type, the fault monitoring module updates the backtracking duration of the corresponding dimension parameter in the backtracking information of the fault type stored within it.

[0093] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0094] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A fault management system for automobile wheel manufacturing process, characterized in that, include: The fault monitoring module is used to obtain multi-dimensional data of a certain type of fault when any of the preset fault types is detected during the manufacturing process of the target automobile wheel, based on the backtracking information of the fault type stored in the pre-stored backtracking information. The backtracking information includes several dimension parameters and their backtracking duration. A cloud-based analytics platform is used to analyze multi-dimensional data of the aforementioned fault types after receiving such data. The cloud-based analytics platform includes a data cleaning unit, a model diagnosis unit, and a record analysis unit. The data cleaning unit is used to detect the multi-dimensional data using an anomaly detection algorithm after receiving the multi-dimensional data. If invalid data is detected, the invalid data is deleted from the multi-dimensional data and the multi-dimensional data is completed using a preset completion strategy to obtain the cleaned data of the type of fault. The invalid data is used as the processing data of the type of fault. The model diagnostic unit is used to input the cleaning data into the fault diagnosis model after receiving the cleaning data and processing data of the fault type, obtain the output diagnostic data of the fault type, the diagnostic data includes the fault cause and diagnosis time of the fault type, and summarize the diagnostic data, processing data and cleaning data of the fault type to generate the record data of the fault type. The recording and analysis unit is used to store the recorded data of the aforementioned fault type after receiving it. For any fault type, the recording and analysis unit analyzes the recorded data of the aforementioned fault type stored in its internal storage after reaching a fixed amount, and generates updated data for the aforementioned fault type. The fault monitoring module is also used to update the backtracking information of the aforementioned fault type stored in its internal storage after receiving the updated data.

2. A fault management system for automobile wheel manufacturing process according to claim 1, characterized in that, The fault monitoring module is also used to collect and store the monitoring values ​​of all dimensional parameters preset by the administrator in real time.

3. A fault management system for automobile wheel manufacturing process according to claim 2, characterized in that, The multi-dimensional data of the fault type includes time-series data of all dimensional parameters included in the backtracking information of the fault type. The time-series data of each dimensional parameter includes all monitoring values ​​of the dimensional parameter within its corresponding backtracking duration and their acquisition times.

4. A fault management system for automobile wheel manufacturing process according to claim 1, characterized in that, The invalid data refers to missing and outlier values ​​generated by the fault monitoring module during the real-time collection of monitoring values ​​of all dimensions of parameters and the transmission to the cloud analysis platform.

5. A fault management system for automobile wheel manufacturing process according to claim 4, characterized in that, For missing values ​​in invalid data, the completion strategy handles missing values ​​through deletion, filling, interpolation, and prediction model methods. For outliers in invalid data, the completion strategy handles outliers through deletion, replacement, and transformation methods.

6. A fault management system for automobile wheel manufacturing process according to claim 1, characterized in that, For any type of fault, the analysis unit performs the following steps when the stored fault-related data reaches a preset fixed amount: S11: Obtain all stored fault-related data, extract the fault causes from each, and remove duplicates. Label all remaining fault causes after deduplication as B1, B2, ..., Bb, where b≥1; S12: Obtain all data containing fault cause B1 from the stored data, labeling them as C1, C2, ..., Cc, where c≥1; S13: Analyze the diagnostic time of the data C1, C2, ..., Cc. Deviation analysis yields the baseline time F1 for fault cause B1; S14: Traverse the recorded data C1, C2, ..., Cc, obtain the dimension parameters corresponding to all monitoring values ​​contained in the recorded data C1, C2, ..., Cc, and deduplicate all the obtained dimension parameters, labeling the remaining dimension parameters as G1, G2, ..., Gg, where g≥1; S15: Create an empty redundant set of dimension parameter G1 relative to fault cause B1, and extract the data from the processed data contained in the recorded data C1, C2, ..., Cc where the frequency of the acquisition time is greater than or equal to P1 and belongs to dimension parameter G1. The acquisition duration is collected and deduplicated. All remaining acquisition durations after deduplication are sequentially labeled as H1, H2, ..., Hh in ascending order of value, where h ≥ 1. P1 is a preset standard analysis frequency. S16: Select whether to add acquisition duration H1 to the redundancy set according to preset selection rules. S17: Select whether to add acquisition durations H2, H3, ..., Hh to the redundancy set of dimension parameter G1 according to S16, obtaining the final redundancy set of dimension parameter G1 relative to fault cause B1. S18: Extract the acquisition duration with the largest value from the redundancy set as the dimension parameter G1 relative to fault cause B1. S19: Calculate and obtain the backtracking duration of dimension parameters G2, G3, ..., Gg relative to the fault cause B1 in sequence according to S11 to S18. Generate the analysis set of fault cause B1 based on the backtracking duration of dimension parameters G1, G2, ..., Gg relative to the fault cause B1. S110: Generate the analysis sets of fault causes B2, B3, ..., Bb in sequence according to S11 to S19. Summarize and compare the analysis sets of fault causes B1, B2, ..., Bb to generate updated data for the aforementioned type of fault. Generate updated data for the aforementioned type of fault based on the new backtracking duration of all dimension parameters.

7. A fault management system for automobile wheel manufacturing process according to claim 6, characterized in that, S16, the selection rule for whether to add the acquisition duration H1 to the redundancy set is as follows: S161: Obtain all the record data containing the acquisition duration H1 in the processing data of the record data C1, C2, ..., Cc, and re-label them as I1, I2, ..., Ii, i≥1 respectively; S162: Determine whether the acquisition duration H1 is a redundant duration based on the record data I1 according to the preset judgment rule. The judgment steps are as follows: SS11: Extract the cleaning data contained in the record data I1, and remove all the monitoring values ​​and their acquisition durations that were collected before the monitoring values ​​of the acquisition duration H1 from all the monitoring values ​​belonging to the dimension parameter G1 contained in the cleaning data, so as to obtain the preliminary data of the record data I1 based on the acquisition duration H1; SS12: Input the preliminary data into the fault diagnosis model, obtain the diagnostic data output by the fault diagnosis model, and make a first judgment on the fault causes contained in the record data I1 based on the diagnostic data and the first judgment result to determine the acquisition duration H1.

1. Does it affect the cause of the fault? If the first determination result indicates that the acquisition duration H1 has no impact on the cause of the fault, the diagnostic time in the diagnostic data is compared with the baseline time F1 for a second determination. Based on the second determination result, it is determined whether the acquisition duration H1 has an adverse effect on the fault time. SS13: If the first determination result indicates that the acquisition duration H1 has no impact on the cause of the fault, and the second determination result indicates that the acquisition duration H1 has no adverse effect on the fault time, then the acquisition duration H1 is determined to be a redundant duration based on the recorded data I1. Otherwise... Then determine whether the acquisition duration H1 is a redundant duration based on the recorded data I1; S163: Determine whether the acquisition duration H1 is a redundant duration based on the recorded data I2, I3, ..., Ii in sequence according to the steps of S161 to S162. If the acquisition duration H1 is a redundant duration based on the recorded data I1, I2, ..., Ii, then select to add the acquisition duration H1 to the redundant set of dimension parameter G1; otherwise, select not to add the acquisition duration H1 to the redundant set of dimension parameter G1 relative to the fault cause B1.

8. A fault management system for automobile wheel manufacturing process according to claim 7, characterized in that, In SS12, the first determination is as follows: extract the fault causes in the diagnostic data and the fault causes in the recorded data I1 respectively, compare the two extracted fault causes for consistency. If the two fault causes are consistent, it is determined that the acquisition duration H1 has no effect on the fault cause; otherwise, it is determined that the acquisition duration H1 has an effect on the fault cause. At this time, it is further determined that the acquisition duration H1 is not a redundant duration based on the recorded data I1.

9. A fault management system for automobile wheel manufacturing process according to claim 7, characterized in that, In SS12, the second determination is as follows: extract the diagnostic time in the diagnostic data, calculate the absolute value of the difference between it and the baseline time F1, compare the absolute value with P2, if the absolute value is less than or equal to P2, it is determined that the acquisition time H1 has no adverse effect on the fault time, otherwise it is determined that the acquisition time H1 has an adverse effect on the fault time. P2 is a preset adverse time determination threshold.

Images