Process inspection method and device for tightening workpieces, electronic equipment and storage medium

Abstract

This application provides a process inspection method, apparatus, electronic device, and computer-readable storage medium for tightening workpieces. The method includes: acquiring first process context information and first tightening process data of the workpiece being tightened; constructing features based on the first process context information and the first tightening process data to obtain a first tightening feature set and a first temporal feature set of the workpiece; acquiring a first verification result from a rule engine, a second verification result from a similarity model, a third verification result from a feature classification model, a fourth verification result from a temporal depth model, and a fifth verification result from an unsupervised anomaly detection model based on the first tightening feature set and the first temporal feature set; and determining the process inspection result of the tightened workpiece based on the first, second, third, fourth, and fifth verification results. This application can improve the accuracy of tightening result determination and reduce the risk of abnormal tightening of workpieces.

Description

Technical Field

[0001] This application relates to the field of intelligent assembly technology, specifically to a process inspection method, apparatus, electronic device, and computer-readable storage medium for tightening workpieces. Background Technology

[0002] Intelligent tightening tools are industrial assembly equipment that integrates sensors, controllers, communication modules, and tightening components. They can detect parameters such as torque, angle, and speed through sensors to ensure that fasteners meet preset process requirements (torque / angle).

[0003] In related technologies, the final torque / angle is used to determine whether the tightened workpiece meets the process requirements. However, even if the final torque / angle is within the predetermined range, there are still cases where the tightened workpiece does not meet the process requirements. The method of determining whether the tightened workpiece meets the process requirements by identifying the final torque / angle has a relatively high misjudgment rate, resulting in a high risk of failure in the final tightened workpiece. Summary of the Invention

[0004] This application provides a process inspection method, electronic device, and computer-readable storage medium for tightening workpieces, which can improve the accuracy of tightening result judgment and reduce the risk of abnormal tightening of workpieces.

[0005] In a first aspect, this application provides a process inspection method for tightened workpieces, the method comprising: The first process context information and the first tightening process data of the workpiece to be tightened are obtained. The first process context information includes identification information, configuration information and batch information. The first tightening process data includes the torque curve, angle curve, speed curve and stage mark of the entire tightening process of the workpiece to be tightened. Based on the first process context information and the first tightening process data, feature construction is performed to obtain the first tightening feature set and the first temporal feature of the workpiece being tightened in this operation. Based on the first tightening feature set and the first temporal features, the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the temporal deep model, and the fifth verification result of the unsupervised anomaly detection model are obtained. Based on the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result, a fusion decision is made to obtain the process inspection result of the workpiece being tightened. The fusion decision process includes making a compliance decision on the workpiece being tightened based on the first verification result, and determining whether the workpiece being tightened has at least one of the following information: risk, risk level, abnormal stage, abnormal attribution, and process optimization suggestion information.

[0006] Secondly, this application also provides a process inspection device for tightening workpieces, comprising: The acquisition unit is used to acquire the first process context information and the first tightening process data of the workpiece being tightened. The first process context information includes identification information, configuration information and batch information. The first tightening process data includes the torque curve, angle curve, rotation speed curve and stage markers of the entire tightening process of the workpiece being tightened. The construction unit is used to construct features based on the first process context information and the first tightening process data to obtain the first tightening feature set and the first temporal feature of the workpiece being tightened this time. The identification unit is used to obtain the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the temporal depth model, and the fifth verification result of the unsupervised anomaly detection model based on the first tightening feature set and the first temporal features. The decision-making unit is used to make a fusion decision based on the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result to obtain the process inspection result of the workpiece being tightened. The fusion decision-making process includes making a compliance decision on the workpiece being tightened based on the first verification result, and determining whether the workpiece being tightened has at least one of the following information: risk, risk level, abnormal stage, abnormal attribution, and process optimization suggestion information.

[0007] Thirdly, this application also provides an electronic device, which includes a processor and a memory, wherein the memory stores a computer program, and the processor executes any of the tightening workpiece process inspection methods provided in this application when it calls the computer program in the memory.

[0008] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, the computer program being loaded by a processor to execute the process inspection method for tightening workpieces.

[0009] In this application, in a first aspect, a first tightening feature set and a first temporal feature are constructed using first process context information and first tightening process data. These are used to obtain the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the temporal depth model, and the fifth verification result of the unsupervised anomaly detection model, thereby determining the process inspection result of the tightened workpiece. Since the first tightening feature set and the first temporal feature contain process context information and tightening process data, they can identify hidden anomalies that are difficult to detect by the traditional final value method. Thus, in the scenario where the final value is qualified but the process is abnormal, abnormal tightened workpieces can be identified, thereby reducing the misjudgment rate of tightening results (i.e., process inspection results) to a certain extent. Secondly, by using the first verification result based on the rule engine, the second verification result based on the similarity model, the third verification result based on the feature classification model, the fourth verification result based on the time-series deep model, and the fifth verification result based on the unsupervised anomaly detection model, the process inspection result of the tightened workpiece is determined. This can integrate multiple models or engines to make decisions on the process inspection result of the tightened workpiece, reduce the problem of misjudgment that is prone to occur when making decisions based on a single model, thereby improving the accuracy of the tightening result judgment and reducing the risk of abnormalities in the tightened workpiece. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic block diagram of the structure of an electronic device provided in an embodiment of this application; Figure 2 This is a schematic flowchart of a process inspection method for tightening workpieces provided in an embodiment of this application; Figure 3 This is an illustrative diagram illustrating the verification results of the multi-model or engine fusion decision-making process provided in the embodiments of this application; Figure 4 This is a schematic diagram of an embodiment of determining the first verification result provided in this application; Figure 5 This is a schematic diagram of an embodiment of determining the second verification result provided in this application. Figure 6 This is a schematic diagram of an embodiment of the process inspection device for tightening workpieces provided in this application. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0014] In the description of the embodiments of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0015] To enable any person skilled in the art to implement and use this application, the following description is provided. In this description, details are set forth for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be implemented without using these specific details. In other instances, well-known processes will not be described in detail to avoid obscuring the description of the embodiments of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in the embodiments of this application.

[0016] This application provides a process inspection method for tightened workpieces, an electronic device, and a computer-readable storage medium. The electronic device may be a server, mobile phone, computer, etc.

[0017] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0018] Figure 1 This is a schematic block diagram of the structure of an electronic device provided in an embodiment of this application.

[0019] like Figure 1As shown, the electronic device 100 includes a processor 101 and a memory 102, which are connected via a bus 103, such as a PCIe (Peripheral Component Interconnect Express) bus. The electronic device 100 can establish a communication connection with the tightening device via wired or wireless communication to receive tightening operation data (such as tightening process data) from the tightening device.

[0020] Specifically, processor 101 provides computing and control capabilities to support the operation of the entire electronic device 100. Processor 101 can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0021] Specifically, the memory 102 can be a Flash chip, a read-only memory (ROM) disk, an optical disk, a USB flash drive, or a portable hard drive, etc.

[0022] Those skilled in the art will understand that Figure 1 The structures shown are merely block diagrams of some structures related to the embodiments of this application and do not constitute a limitation on the electronic devices to which the embodiments of this application are applied. Specific electronic devices may include more or fewer components than those shown in the figures, or combine certain components, or have different component arrangements.

[0023] The processor 101 is configured to run a computer program stored in the memory 102, and when executing the computer program, implement any of the process inspection methods for tightening workpieces provided in the embodiments of this application. For example, the processor 101 is configured to run a computer program stored in the memory 102, and when executing the computer program, it can implement the following steps: The process involves acquiring the first process context information and the first tightening process data of the workpiece to be tightened. The first process context information includes identification information, configuration information, and batch information. The first tightening process data includes the torque curve, angle curve, rotational speed curve, and stage markers for the entire tightening process of the workpiece. Based on the first process context information and the first tightening process data, feature construction is performed to obtain the first tightening feature set and the first temporal feature of the workpiece. Based on the first tightening feature set and the first temporal feature, the first verification result of the rule engine, the second verification result of the similarity model, and the feature classification model are obtained. The third verification result, the fourth verification result of the time-series deep model, and the fifth verification result of the unsupervised anomaly detection model are used to make a fusion decision based on the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result to obtain the process inspection result of the tightened workpiece. The fusion decision process includes making a compliance decision for the tightened workpiece based on the first verification result, and determining whether the tightened workpiece has at least one of the following information: risk, risk level, anomaly stage, anomaly attribution, and process optimization suggestion information based on the second verification result, the third verification result, the fourth verification result, and the fifth verification result.

[0024] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the electronic device described above can be referred to the corresponding process in the following embodiment of the process inspection method for tightening workpieces, and will not be repeated here.

[0025] The following will be based on Figure 1 Taking the electronic device shown as the execution subject of the process inspection method for tightening workpieces as an example, the process inspection method for tightening workpieces provided in this application embodiment will be described in detail. For the sake of simplification and ease of description, the execution subject will be omitted in the subsequent method embodiments.

[0026] Please see Figure 2 , Figure 2 This is a flowchart illustrating a process inspection method for tightened workpieces according to an embodiment of this application. The process inspection method for tightened workpieces includes steps 201-204, wherein: 201. Obtain the first process context information and the first tightening process data of the workpiece being tightened.

[0027] The workpiece being tightened refers to the workpiece that has completed the tightening operation and is awaiting process result inspection. It is the object of process result inspection, such as the engine cylinder head after the cylinder head bolts of an automobile engine are tightened, the battery pack after the bolts of a new energy battery pack are tightened, and the flange assembly after the flange bolts of mechanical equipment are tightened.

[0028] The first process context information refers to non-process auxiliary information related to the tightening operation of the workpiece in this tightening operation. It is a collection of various information describing the tightening operation scenario, configuration, and batch, used to adapt to different tightening scenarios and reduce judgment distortion caused by cross-scenario mixed judgment. The first process context information includes identification information, configuration information, and batch information. The first tightening process data includes the current torque curve, current angle curve, current speed curve, and stage markers.

[0029] The identification information may include, but is not limited to, tightening gun numbers (e.g., gun numbers: G01, G02), workstation numbers (e.g., workstation numbers: W05, W06), and production line numbers (e.g., production line number: L02). Configuration information may include, but is not limited to, tightening procedure numbers (e.g., tightening procedure number: P1008) and process version numbers (e.g., process version number: V2.1). Batch information may include, but is not limited to, product model (e.g., automotive engine model: EA888), product batch number (e.g., B2026040601), and production shift (e.g., morning shift 8:00-16:00, afternoon shift 16:00-24:00).

[0030] The first tightening process data refers to the set of dynamic data related to the tightening action collected in real time throughout the entire tightening operation. It is data that reflects the state of the tightening process and may include, but is not limited to, the torque curve, angle curve, speed curve, and stage markers for the current tightening process of the current workpiece.

[0031] The torque curve in this context is used to indicate the change of torque over time during the tightening operation (i.e., the entire tightening operation of the workpiece). Specifically, the torque curve can be a curve formed by plotting the tightening operation time on the horizontal axis and the torque value detected in real time during the tightening process on the vertical axis.

[0032] The current angle curve is used to indicate the change of the angle during the tightening operation over time. Specifically, the current angle curve can be a curve formed by plotting the tightening operation time on the horizontal axis and the bolt rotation angle value detected in real time during the tightening process on the vertical axis.

[0033] The speed curve in this example is used to indicate the change in speed over time during the tightening operation. Specifically, this speed curve can be plotted with the tightening operation time on the x-axis and the real-time monitored output speed of the tightening gun on the y-axis.

[0034] The stage markers are used to indicate the time periods corresponding to different process stages (such as the tooth finding stage, the fitting stage, and the final tightening stage) throughout the entire tightening operation. They can be generated in real time during the tightening process or marked uniformly after the operation is completed.

[0035] For example, firstly, a single tightening result package is received: after the tightening operation of the workpiece is completed, a single tightening result package sent by the tightening gun is received. The single tightening result package includes information related to the tightening process, such as sampling time, torque curve, angle curve, and speed curve, as well as information related to the operation scenario, such as program number, station number, gun number, and product batch, to ensure that the data covers comprehensive information related to the tightening process and the operation scenario.

[0036] Then, protocol parsing and data verification: The received tightening result packet is parsed to complete field mapping and process version matching. At the same time, legality checks, field corrections, missing packet / missing value checks, timestamp corrections, and process mappings are performed to obtain the parsed and verified result packet. Among them, the legality check is used to confirm that the data format and field content meet the preset standards; the missing packet / missing value check is used to investigate and supplement missing data; the timestamp correction is used to unify the data time precision; and the process mapping is used to associate the standard parameters of the corresponding process version to ensure that the parsed data is accurate, complete, and usable.

[0037] Next, data classification, extraction, and integration: From the parsed and verified result package, information related to the work scenario (such as program number, workstation number, gun number, product batch, etc.) is extracted as the first process context information, and information related to the tightening process (such as torque curve, angle curve, speed curve, etc.) is extracted as the first tightening process data. Among them, the torque curve, angle curve, and speed curve in the result package correspond to the current torque curve, current angle curve, and current speed curve. At the same time, the stage markers of the tightening operation are extracted, and the stage markers are aligned with the time axis of each curve, thereby marking the process stage of each curve. For example, the current torque curve will be marked with the tooth finding stage (such as the t1-t2 time period), the fitting stage (such as the t2-t3 time period), and the final tightening stage (such as the t3-t4 time period).

[0038] Finally, data standardization processing: The first process context information and the first tightening process data extracted after classification are integrated and further standardized according to the preset data format to ensure that the data has uniform field naming, units, and timestamp accuracy, forming a complete data package of the workpiece being tightened (the complete data package includes the first process context information and the first tightening process data of the workpiece being tightened), and then proceeding to step 202 for feature construction.

[0039] 202. Based on the first process context information and the first tightening process data, feature construction is performed to obtain the first tightening feature set and the first temporal feature of the workpiece being tightened.

[0040] For example, the first tightening feature set includes a first basic tightening feature, a first time-related feature, a first tightening process feature, and a first contextual process feature; the first basic tightening feature, the first time-related feature, and the first tightening process feature of the workpiece being tightened can be determined based on the first tightening process data; the first contextual process feature of the workpiece being tightened can be determined based on the first process context information; and the first time-series feature can be obtained by performing equal-length serialization processing based on the first time-related feature.

[0041] The first tightening feature set may include, but is not limited to, first basic tightening features, first time-related features, first tightening process features, and first contextual process features. The following sections describe how to construct the first basic tightening features, first time-related features, first tightening process features, and first contextual process features: I. First basic tightening characteristics: The first basic tightening characteristics include the final value of torque and angle, the duration of each process stage, the total process duration, and the stage switching characteristics. The stage switching characteristics include at least one of the following: stage switching physical quantity, stage switching duration, and stage switching sequence.

[0042] Among them, the final torque value: by comparing the final torque value of the workpiece being tightened with the final torque value window, it can be determined whether the final torque value of the workpiece being tightened is within the allowable fluctuation range of the process, thereby determining whether there is any abnormality in the workpiece being tightened. For example, the average or maximum torque value within a preset window at the end of the final tightening stage of the workpiece tightening operation can be used as the final torque value of the workpiece being tightened.

[0043] Among them, the final angle value: by comparing the final angle value of the workpiece being tightened with the final angle value window, it can be determined whether the final angle value of the workpiece being tightened is within the allowable fluctuation range of the process, thereby determining whether there is any abnormality in the workpiece being tightened. For example, the average or maximum angle value within the preset window at the end of the final tightening stage of the tightening operation can be used as the final angle value of the workpiece being tightened.

[0044] The duration of each process stage refers to the time spent on each process stage during the tightening of the workpiece. For example, the duration of each process stage can be obtained by statistical analysis based on the stage markers, such as the duration of the tooth-finding stage, the fitting stage, and the final tightening stage.

[0045] The total process time refers to the total time consumed from the start of the tightening operation to its complete end (i.e., fulfilling the preset tightening process requirements and reaching the final tightening standard). For example, it can be obtained by: based on the stage markers in the first tightening process data, calculating the duration of each process stage (such as the tooth-finding stage, the mating stage, and the final tightening stage), and summing the durations of all process stages to obtain the total process time; or by directly using the sampling timestamps in the first tightening process data, taking the difference between the initial timestamp at the start of the tightening operation and the termination timestamp at the end of the tightening operation to obtain the total process time.

[0046] Among them, the stage switching characteristics refer to the set of relevant characteristic parameters when the workpiece is switched from one process stage to another during the tightening operation. They are used to reflect the rationality and standardization of the switching between process stages and serve as the basis for identifying stage switching anomalies (such as switching timing deviations or switching signal anomalies) during the tightening process. Specifically, they may include, but are not limited to, stage switching physical quantities, stage switching duration, and stage switching sequence.

[0047] The physical quantity for stage switching refers to the physical parameters that trigger the tightening operation to switch from the current process stage to the next process stage, and it must be strictly matched with the preset tightening process rules. For example, the physical quantity for stage switching includes at least one of torque value, angle value, and speed value, which can be determined according to the tightening process requirements. For example, the trigger physical quantity for switching from the tooth finding stage (M1) to the mating stage (M2) is the torque value. When the torque value detected in real time during the tightening process reaches the preset mating trigger torque (e.g., 5 N·m), the switch from the tooth finding stage to the mating stage is triggered. The trigger physical quantity for switching from the mating stage (M2) to the final tightening stage (M3) can also be the torque value. When the real-time torque value reaches the preset final tightening trigger torque (e.g., 20 N·m), the switch to the final tightening stage is triggered.

[0048] Stage switching time refers to the time consumed from the end of the current tightening operation to the complete transition to the next stage. Specifically, it's the time difference between the moment the stage switching is triggered (e.g., when the physical quantity of the stage switching reaches a preset threshold) and the moment the characteristic parameters (torque, angle, speed) of the next stage stabilize within the preset range of that stage. Stage switching time can be used to measure the smoothness and speed of stage switching. Excessive switching time may indicate abnormalities such as delayed response of the tightening equipment or unreasonable process parameter settings. For example, if the trigger time for switching from the tooth-finding stage to the mating stage is 3.0s, and the parameter stabilization time in the mating stage is 3.1s, then the switching time for this stage is 0.1s. If the pre-configured target switching time threshold is ≤0.2s, then this switching time meets the process requirements.

[0049] The stage switching sequence refers to the order in which the various process stages are switched during this tightening operation. Following preset tightening process rules is fundamental to ensuring tightening quality. For example, the switching sequence could be the thread-finding stage – fitting stage – final tightening stage. By constructing this stage switching sequence feature, tightening anomalies caused by stage misalignment can be reduced (e.g., directly entering the final tightening stage without completing the thread-finding stage may lead to stripped bolt threads or insecure tightening). For example, the specific method for obtaining this sequence can be: based on the stage markers in the first tightening process data, extract the order in which each process stage appears to obtain the stage switching sequence. Subsequently, compare the stage switching sequence with the preset target stage switching rules to determine whether the switching sequence complies with regulations.

[0050] II. First-time related features: First-time related features include at least one of the current torque curve, current angle curve, current speed curve, and current torque-angle coupling curve.

[0051] For example, the torque curve, angle curve, and speed curve can be directly extracted from the data of the first tightening process.

[0052] The torque-angle coupling curve is used to indicate the change of torque with angle during this tightening operation. For example, it can be a curve with the bolt rotation angle value detected in real time during the tightening process as the abscissa and the torque value detected in real time as the ordinate. It can be generated from the first tightening process data based on the extracted torque and angle data.

[0053] III. Characteristics of the first tightening process: The characteristics of the first tightening process include at least one of the following: stiffness characteristics, ripple frequency domain characteristics, and torque-angle coupling characteristics.

[0054] (i) Stiffness characteristics: These are features extracted based on torque-angle coupling curves. They are used to capture the overall stiffness of the tightened workpiece, bolts, and connectors, thereby identifying hidden anomalies that are difficult to detect using traditional final value methods.

[0055] Among them, stiffness characteristics can be extracted based on the torque-angle coupling curve of the workpiece being tightened. Stiffness characteristics include, but are not limited to, at least one of the following: overall stiffness value (i.e., the fitting slope of the linear segment of the torque-angle coupling curve), stiffness consistency (i.e., the deviation rate between the local stiffness value and the overall stiffness value within the linear segment), and stiffness mutation point (i.e., the angle or torque position where the stiffness value changes significantly in the torque-angle coupling curve).

[0056] The linear segment refers to the curve segment in the torque-angle coupling curve that corresponds to the early stage of the contact segment to the final tightening segment during the entire tightening process.

[0057] The stiffness value of a local point within the linear segment is equal to the fitting slope dT / dA at that local point (i.e., dT / dA is the rate of change of torque with angle, where dT and dA represent the small changes in torque and angle, respectively, T is torque, A is angle, and d is the differential sign). The larger the slope, the higher the stiffness value.

[0058] (ii) Ripple frequency domain characteristics: These are characteristics obtained by combining ripple index and frequency domain index extracted from the torque curve. They are used to explore the potential micro-fluctuation patterns of torque signal throughout the tightening process, capture details such as the operating status of the tightening gun, thread machining accuracy, and lubrication status, thereby capturing process fluctuations that cannot be captured by the final value of angle and the final value of torque, and thus identifying hidden anomalies that are difficult to detect by the traditional final value method.

[0059] Among them, the ripple index includes, but is not limited to, at least one of the following: ripple amplitude (i.e., the difference between the peak and trough values ​​of the torque curve fluctuation), ripple frequency (i.e., the number of fluctuations of the torque curve per unit time), and ripple root mean square (i.e., the root mean square value of the ripple signal, used to quantify the overall intensity of the fluctuation);

[0060] Among them, the frequency domain index is the spectral feature extracted after performing a frequency domain transformation (such as FFT) on the ripple signal.

[0061] Among them, the ripple signal is the result of subtracting the trend component from the torque curve (the trend component is obtained by smoothing the torque curve, which is used to reflect the overall upward trend of the torque curve).

[0062] (iii) Torque-angle coupling characteristics: These are features extracted from the torque-angle coupling curve. They are used to capture the mutual influence and synchronous change of torque and angle during the entire tightening process (finding teeth, fitting, and final tightening), thereby identifying hidden anomalies that are difficult to detect by the traditional final value method.

[0063] IV. First Context Process Features: These are numerical features obtained by encoding the first process context information based on the tightening operation, such as unique heat encoding, tag encoding, and embedded encoding.

[0064] The first process context information includes, but is not limited to, at least one of the following: identification information (such as gun number, workstation number, production line number), configuration information (such as tightening procedure number, process version number), and batch information (such as product model, product batch number, production shift).

[0065] The first temporal-series feature is obtained by performing equal-length serialization on the first time-related features. This unifies curves with different durations and numbers of sampling points into a fixed-length sequence of data, thereby adapting to the input requirements of the temporal-series deep model and ensuring the stability and consistency of the model's judgment. For example, the current torque curve, current angle curve, current speed curve, and current torque-angle coupling curve can be normalized, interpolated, truncated, or resampled to obtain the first temporal-series feature with a uniform length.

[0066] 203. Based on the first tightening feature set and the first temporal features, obtain the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the temporal depth model, and the fifth verification result of the unsupervised anomaly detection model.

[0067] Please refer to Figure 3 , Figure 3 This is an illustrative diagram illustrating the verification results of the multi-model or engine fusion decision-making process provided in the embodiments of this application.

[0068] For example, please refer to Figure 4 , Figure 4 This is a schematic diagram of an embodiment of the present application that provides for determining the first verification result. The first verification result is determined through the following steps A1 to A5: A1. Using the rule engine, determine the final value verification result based on the pre-configured target final value window and the final torque and angle values ​​of the workpiece being tightened.

[0069] The target final value window is a final value window that matches the process scenario parameters of the workpiece being tightened. The target final value window includes at least one of the torque final value window and the angle final value window.

[0070] For example, the rule engine can call a pre-configured target final value window to extract the final torque and final angle values ​​of the workpiece being tightened. It then compares the final torque value with the final torque value window. If the final torque value is within the window's range, the final torque value is deemed acceptable (e.g., a final torque value of 5 N·m corresponds to a final torque value window of 3-7 N·m, indicating acceptable torque). If the final torque value is outside the window's range, the final torque value is deemed unacceptable. The final angle value of the workpiece being tightened is compared with the final angle value window. If the final angle value is within the window, the final angle value is considered acceptable. If it is outside the window, the final angle value is considered unacceptable. Based on the comparison results of the final torque and final angle values, the final value is determined to meet the process requirements. Finally, the final value verification result is output. For example, if both the final torque and final angle values ​​are acceptable, the final value is considered to meet the process requirements, and the final value verification result is considered acceptable.

[0071] A2. Using the rule engine, determine the stage verification result based on the pre-configured target duration range for each stage and the duration value of each process stage.

[0072] The target duration interval for each stage is the target duration interval corresponding to each process stage, and the target duration interval for each stage is the duration interval that matches the process scenario parameters.

[0073] The process scenario parameters may include, but are not limited to, one or more of the following: tightening gun number, workstation number, production line number, tightening procedure number, process version, product model, and production shift.

[0074] For example, the rule engine calls the pre-configured target time ranges corresponding to each process stage, extracts the time value of each process stage one by one, and compares the time value of each process stage with the target time range of the corresponding stage. If the time value of a certain process stage of tightening the workpiece is within the target time range of the corresponding stage, the time value of that process stage is determined to be qualified (for example, the time value of the tooth finding stage is 2s, the corresponding tooth finding stage time range is 1-3s, so the tooth finding stage time value is determined to be qualified); if the time value of a certain process stage of tightening the workpiece is outside the target time range of the corresponding stage, the time value of that process stage is determined to be unqualified; based on the comparison results of the time values ​​of each process stage, it is determined whether the time consumption of each stage meets the process requirements, and finally the stage verification result is output. For example, only when the time values ​​of all process stages of tightening the workpiece are qualified are the time consumption of each stage considered to meet the process requirements, and the stage verification result is qualified.

[0075] A3. Using the rule engine, determine the timeout verification result based on the pre-configured target total duration range and the total process duration value.

[0076] The target total duration interval is the total duration interval that matches the process scenario parameters.

[0077] For example, the rule engine can call a pre-configured target total time range, extract the total process time value of the current tightening of the workpiece, and compare the total process time value of the current tightening of the workpiece with the target total time range. If the total process time value of the current tightening of the workpiece is within the target total time range, the total process time value is determined to be qualified (for example, if the total process time value is 8 seconds, corresponding to the target total time range of 5-10 seconds, the total process time value is determined to be qualified); if the total process time value of the current tightening of the workpiece is outside the target total time range, the total process time value is determined to be unqualified. Based on the comparison result of the total process time value, it is determined whether the overall tightening time is compliant, and finally the timeout verification result is output. For example, if the total process time value of the current tightening of the workpiece is qualified, the overall tightening time is considered compliant, and the timeout verification result is qualified.

[0078] A4. Using the rule engine, determine the stage switching verification result based on the pre-configured target stage switching rules and the stage switching features.

[0079] The target stage switching rule is a stage switching rule that matches the process scenario parameters.

[0080] For example, the rule engine can call a pre-configured target stage switching rule (the target stage switching rule is used to limit the specific stage switching physical quantity, stage switching duration, and stage switching sequence), extract the stage switching features of the workpiece being tightened, and compare each stage switching feature of the workpiece being tightened (such as stage switching physical quantity, stage switching duration, and stage switching sequence) with the corresponding requirements of the target stage switching rule. If a certain stage switching feature of the workpiece being tightened matches the target stage switching rule, then the stage switching parameter is deemed qualified (for example, if the stage switching duration is 0.1s, and the corresponding target switching duration threshold is ≤0.2s, then the stage switching duration is deemed qualified). If a certain stage switching feature of the workpiece being tightened does not match the target stage switching rule, then the stage switching feature is deemed unqualified. Based on the comparison results of each stage switching feature, the engine determines whether the stage switching is standardized and finally outputs the stage switching verification result. For example, if all stage switching features of the workpiece being tightened are qualified, then the stage switching is considered standardized, and the stage switching verification result is qualified.

[0081] A5. Based on the final value verification result, the stage verification result, the timeout verification result, and the stage switching verification result, determine the first verification result.

[0082] The first verification result includes a first risk classification label. If at least one of the final value verification result, the stage verification result, the timeout verification result, and the stage switching verification result is unqualified, then the first risk classification label is risk.

[0083] The first risk classification label is a classification label determined by the rule engine to indicate whether there is a risk in the tightening of the workpiece.

[0084] For example, the rule engine summarizes the final value verification results, stage verification results, timeout verification results, and stage switching verification results, and performs logical judgment on each verification result. If any one of the verification results is unqualified, the first risk classification label is risk; if all the verification results, such as the final value verification result, stage verification result, timeout verification result, and stage switching verification result, are qualified, the first risk classification label is normal.

[0085] For example, please refer to Figure 5 , Figure 5 This is a schematic diagram of an embodiment of the present application providing the determination of a second verification result. The second verification result includes a first risk score, a second risk classification label, and / or the anomaly probability of each process stage. The second verification result is determined through the following steps B1 to B5: B1. Using the similarity model, the torque reference curve is matched with the current torque curve according to the pre-configured torque reference curve, and the first curve similarity and the first deviation stage are output.

[0086] The process involves several steps: First, a standardized torque curve is generated using torque curves from multiple standard samples in a pre-defined standard template library. This curve is generated through feature extraction, curve normalization, and mean fitting. The standard samples are tightened workpieces with normal process inspection results. This standardized torque curve is configured as the torque reference curve for the similarity model. Second, a standardized torque-angle coupling curve is generated using torque-angle coupling curves from multiple standard samples in the pre-defined standard template library. This curve is configured as the torque-angle coupling reference curve for the similarity model. Third, a standardized torque-angle coupling curve is generated using torque-angle coupling curves from multiple standard samples in the pre-defined standard template library. This standardized torque-angle coupling curve is configured as the torque-angle coupling reference curve for the similarity model. The scene parameters for each standard sample used to generate the standardized torque curve are the same as the scene parameters for the workpiece being tightened. These scene parameters include at least one of the following: tightening gun number, workstation number, tightening program number, process version, product model, and production shift. The scenario parameters for each standard sample used to generate the standardized angle curve are the same as the scenario parameters for the workpiece being tightened in this instance. Similarly, the scenario parameters for each standard sample used to generate the standardized torque-angle coupling curve are the same as the scenario parameters for the workpiece being tightened in this instance.

[0087] The first curve similarity refers to the similarity between the current torque curve and the torque reference curve.

[0088] The first deviation stage refers to the process stage with deviation determined based on the current torque curve and the torque reference curve. If there is a certain deviation (such as the deviation being greater than a preset deviation threshold) between the curve segment corresponding to a certain process stage of the current torque curve (such as curve segment 1 corresponding to the tooth-finding stage) and the curve segment corresponding to the same process stage of the torque reference curve (such as curve segment 2 corresponding to the tooth-finding stage), then that process stage is regarded as the first deviation stage.

[0089] The torque reference curve is a standardized torque curve that matches the process parameters of the workpiece being tightened. It can be constructed based on standard samples from a pre-set standard template library. For example, the similarity model can call the pre-configured torque reference curve, extract the torque curve of the workpiece being tightened, perform feature alignment and similarity calculation (such as cosine similarity and Euclidean distance calculation) on the two curves, and finally output the first curve similarity (for example, the similarity calculation result is 0.92, the closer to 1, the more similar the two curves are).

[0090] The specific network structure of the similarity model can be set according to the actual business scenario requirements. For example, a DTW / Soft-DTW network structure can be used. The similarity model can use algorithms such as dynamic time warping, piecewise Euclidean distance, correlation coefficient, or cosine similarity to match curves, thereby locating the deviation stage based on the stage marker.

[0091] B2. Using the similarity model, match the current angle curve with the pre-configured angle reference curve to output the second curve similarity and the second deviation stage.

[0092] The second curve similarity refers to the similarity between the current angle curve and the angle reference curve.

[0093] The second deviation stage refers to the process stage with deviation determined based on the current angle curve and the angle reference curve. If there is a certain deviation (such as the deviation being greater than the preset deviation threshold) between the curve segment corresponding to a certain process stage of the current angle curve (such as curve segment 3 corresponding to the bonding stage) and the curve segment corresponding to the same process stage of the angle reference curve (such as curve segment 4 corresponding to the bonding stage), then that process stage is regarded as the second deviation stage.

[0094] The angle reference curve is a standardized angle curve that matches the process parameters of the tightening workpiece in this scenario. It can be constructed based on standard samples from a preset standard template library. For example, the similarity model can call the pre-configured angle reference curve, extract the current angle curve of the tightening workpiece, align the two curves, calculate the similarity, and finally output the second curve similarity.

[0095] B3. Using the similarity model, the torque angle coupling reference curve is matched with the current torque angle coupling curve according to the pre-configured torque angle coupling reference curve, and the third curve similarity and third deviation stage are output.

[0096] The third curve similarity refers to the similarity between the current torque-angle coupling curve and the torque-angle coupling reference curve.

[0097] The second deviation stage refers to the process stage with deviation determined based on the current torque-angle coupling curve and the torque-angle coupling reference curve. If there is a certain deviation (e.g., the deviation is greater than a preset deviation threshold) between the curve segment corresponding to a certain process stage of the first torque-angle coupling curve (such as curve segment 5 corresponding to the final tightening stage) and the curve segment corresponding to the same process stage of the torque-angle coupling reference curve (such as curve segment 6 corresponding to the final tightening stage), then this process stage is designated as the third deviation stage.

[0098] The torque-angle coupling reference curve is a standardized torque-angle coupling curve that matches the process parameters of the tightening workpiece in this case. It can be constructed based on standard samples from a preset standard template library. For example, the similarity model can call the pre-configured torque-angle coupling reference curve, extract the torque-angle coupling curve of the workpiece in this case, align the features of the two curves, calculate the similarity, and finally output the similarity of the third curve.

[0099] In this way, the similarity model can be used to locate the deviation stages of the current torque curve, current angle curve, and current torque-angle coupling curve, thereby uncovering the process stages where anomalies truly exist and improving the accuracy of identifying abnormal stages in the current workpiece tightening.

[0100] B4. Based on the first curve similarity, the second curve similarity, and the third curve similarity, determine the first risk score and / or the second risk classification label.

[0101] The first risk score is the risk score of the workpiece being tightened in this instance, determined based on the second verification result. For example, the first risk score can be determined based on the first curve similarity, the second curve similarity, and the third curve similarity. The first curve similarity, the second curve similarity, and the third curve similarity are all negatively correlated with the first risk score; that is, the greater the first curve similarity, the smaller the first risk score; the greater the second curve similarity, the smaller the first risk score; and the greater the third curve similarity, the smaller the first risk score.

[0102] The second risk classification label is determined based on a similarity model to indicate whether there is a risk in the tightening of the workpiece. In some embodiments, classification can be based on a first risk score of the workpiece being tightened. If the first risk score is greater than a preset risk score threshold, the second risk classification label is "risk"; if the first risk score is less than or equal to the preset risk score threshold, the second risk classification label is "normal". In some embodiments, classification can be based on a first curve similarity, a second curve similarity, and a third curve similarity. If the first curve similarity, the second curve similarity, or the third curve similarity is less than a preset similarity threshold, the second risk classification label is "risk"; if the first curve similarity, the second curve similarity, and the third curve similarity are all greater than or equal to the preset similarity threshold, the second risk classification label is "normal".

[0103] B5. Based on the first deviation stage, the second deviation stage, and the third deviation stage, determine the first abnormal probability of each process stage of the current workpiece tightening.

[0104] For example, based on the first deviation stage, the second deviation stage, and the third deviation stage, the number of times each process stage is judged as a deviation stage can be counted; and the ratio of the number of times each process stage is judged as a deviation stage to the number of curve types used for deviation stage judgment can be used as the anomaly probability of that process stage. For example, assuming there are a total of 3 process stages, including the tooth finding stage, the bonding stage, and the final tightening stage, and the number of curve types used for deviation stage judgment is 3 (i.e., torque curve, angle curve, torque-angle coupling curve), the first deviation stage is the tooth finding stage, the second deviation stage is the tooth finding stage and the bonding stage, and the third deviation stage is the tooth finding stage, then the number of times the tooth finding stage, the bonding stage, and the final tightening stage are judged as deviation stages are 3, 1, and 0, respectively. Therefore, the anomaly probabilities of the tooth finding stage, the bonding stage, and the final tightening stage are 3 / 3, 1 / 3, and 0 / 3, respectively.

[0105] For example, the third verification result is determined through the following steps C1~C2: C1. Based on the first tightening process features and the first context process features, a combined encoding is performed to obtain the first multi-dimensional feature vector of the workpiece being tightened.

[0106] C2. Using the feature classification model, classify based on the first multi-dimensional feature vector and output the third verification result.

[0107] The third verification result includes a second risk score, a third risk classification label, and the probability and / or quality level of each candidate anomaly type. Candidate anomaly types include at least one of the following: under-tightening, over-tightening, stripped threads, floating threads, missed tightening, fit abnormalities, and tool abnormalities.

[0108] In some embodiments, the third verification result includes a second risk score of the workpiece being tightened, which can be obtained by predicting the second risk score of the workpiece being tightened based on the first multi-dimensional feature vector using a feature classification model.

[0109] In some embodiments, the third verification result also includes a third risk classification label, which is classified based on the second risk score of the workpiece being tightened. If the second risk score is greater than a preset risk score threshold, the third risk classification label is risky; if the second risk score is less than or equal to the preset risk score threshold, the third risk classification label is normal.

[0110] The third risk classification label is a classification label determined by a similarity model to indicate whether there is a risk in the workpiece being tightened.

[0111] In some embodiments, the third verification result may further include the anomaly probability of the candidate anomaly type of the workpiece being tightened. This anomaly probability is obtained by predicting the candidate anomaly type of the workpiece being tightened based on the first multi-dimensional feature vector using a feature classification model. Thus, anomaly type identification can be achieved through the feature classification model, thereby uncovering the attribution of the anomaly in the workpiece being tightened.

[0112] In some embodiments, the third verification result may also include a quality level, such as a quality level divided into four levels: A, B, C, and D. The quality level of the workpiece being tightened can be obtained by classifying it based on the first multi-dimensional feature vector using a feature classification model, thereby classifying the quality level of the workpiece being tightened.

[0113] The specific network structure of the feature classification model can be set according to the actual business scenario requirements. For example, XGBoost, LightGBM, CatBoost and other network structures can be used.

[0114] For example, the fourth verification result is determined by the following step D1: D1. Based on the first temporal features, anomaly detection is performed using the temporal depth model to obtain the fourth verification result.

[0115] In some embodiments, the fourth verification result includes the second anomaly probability of each process stage. Anomaly prediction is performed based on the first temporal features using a time-series depth model, and the second anomaly probability of each process stage is output. Thus, the time-series depth model enables anomaly identification in the time-series process, improving the comprehensiveness of anomaly identification for this workpiece tightening operation.

[0116] In some embodiments, the fourth verification result includes a third risk score, which is generated by using a time-series depth model to predict anomalies based on the first time-series features and outputting the third risk score for the workpiece being tightened.

[0117] In some embodiments, the fourth verification result includes a fourth risk classification label. For example, it can be classified based on the third risk score of the workpiece being tightened. If the third risk score is greater than a preset risk score threshold, the fourth risk classification label is risky; if the third risk score is less than or equal to the preset risk score threshold, the fourth risk classification label is normal.

[0118] Among them, the fourth risk classification label is a classification label for whether there is a risk in the tightening of the workpiece, which is determined based on the time-series depth model.

[0119] The specific network structure of the temporal deep model can be set according to the actual business scenario requirements. For example, network structures such as ID-CNN, TCN, and CNN-LSTM can be used.

[0120] For example, the fifth verification result is determined through the following steps E1~E2: E1. Based on the first tightening process features and the first context process features, a combined encoding is performed to obtain the first multi-dimensional feature vector of the workpiece being tightened.

[0121] E2. Using the unsupervised anomaly detection model, anomaly detection is performed based on the first multi-dimensional feature vector and the first temporal feature, and the fifth verification result is output.

[0122] In some embodiments, the fifth verification result includes a fourth risk score, which indicates the risk score of the workpiece being tightened. The unsupervised anomaly detection model performs anomaly detection based on the first multi-dimensional feature vector and the first temporal feature, outputting the fourth risk score of the workpiece being tightened.

[0123] In some embodiments, the fifth verification result also includes a fifth risk classification label. For example, it can be classified based on the fourth risk score of the workpiece being tightened. If the fourth risk score is greater than a preset risk score threshold, the fifth risk classification label is risky; if the fourth risk score is less than or equal to the preset risk score threshold, the fifth risk classification label is normal.

[0124] The fifth risk classification label is a classification label for whether there is a risk in the tightening of the workpiece, determined by an unsupervised anomaly detection model.

[0125] The specific network structure of an unsupervised anomaly detection model can be set according to the actual business scenario requirements. For example, network structures such as Isolation Forest and AutoEncoder can be used. In this way, the unsupervised anomaly detection model can discover novel anomalies not covered by the training labels, reducing the problem of limited anomaly coverage by the training labels of similarity models, feature classification models, and temporal deep models. This allows for the discovery of unpredictable unknown anomalies, thereby improving the anomaly recognition rate of the final process inspection results.

[0126] Therefore, the decision is made by integrating the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the time-series deep model, and the fifth verification result of the unsupervised anomaly detection model. This allows for compliance vetoes using the rule engine, the discovery of truly abnormal process stages using the similarity model, the discovery of the anomaly attribution for the tightened workpiece using the feature classification model, the identification of anomalies in the time-series process using the time-series deep model, and the discovery of unknown anomalies using the unsupervised anomaly detection model. This approach can improve the comprehensiveness and accuracy of anomaly identification for the tightened workpiece.

[0127] 204. Based on the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result, a fusion decision is made to obtain the process inspection result of the tightened workpiece.

[0128] The integrated decision-making process includes making a compliance decision on the workpiece being tightened based on the first verification result, and determining, based on the second, third, fourth, and fifth verification results, whether the workpiece being tightened has at least one of the following: risk level, risk grade, abnormal stage, abnormal attribution, or process optimization suggestion information. The implementation method for making a compliance decision on the workpiece being tightened can refer to the relevant explanation in step 2041A; the method for determining whether the workpiece being tightened has a risk can refer to the relevant explanation in step 2042A; the method for determining the risk grade of the workpiece being tightened can refer to the relevant explanation in step 2043A; and the method for determining the abnormal stage, abnormal attribution, and process optimization suggestion information of the workpiece being tightened can refer to the relevant explanation in step 2044B.

[0129] The process inspection results are used to indicate whether the workpiece being tightened meets the process requirements, such as whether the workpiece being tightened is a compliant product or whether there are any risks.

[0130] There are several ways to implement step 204, including, for example: (1) In some embodiments, the process inspection results are used to indicate whether the workpiece being tightened is a compliant product, and whether there is any risk and the risk level if the workpiece being tightened is a compliant product. In this case, step 204 may specifically include the following steps 2041A~2043A: 2041A. Based on the first verification result, conduct a compliance test on the workpiece being tightened.

[0131] In some embodiments, compliance checks are performed using the first verification result. For example, compliance checks are performed using the first risk classification label contained in the first verification result. If the first risk classification label is normal, the workpiece being tightened is determined to be a compliant product; if the first risk classification label is risky, the workpiece being tightened is determined to be a non-compliant product. In this way, compliance is vetoed using the rule engine.

[0132] In some embodiments, the first verification result and other verification results (such as at least one of the second, third, fourth, and fifth verification results) can be combined to perform compliance testing on the tightened workpiece. For example, the first verification result may include a first risk classification label, the second verification result may include a second risk classification label, the third verification result may include a third risk classification label, the fourth verification result may include a fourth risk classification label, and the fifth verification result may include a fifth risk classification label. If all of the first, second, third, fourth, and fifth risk classification labels are normal, the tightened workpiece is determined to be a compliant product. If any one of the first, second, third, fourth, and fifth risk classification labels is risky, the tightened workpiece is determined to be a non-compliant product.

[0133] 2042A. If the workpiece being tightened is a compliant product, then based on the second verification result, the third verification result, the fourth verification result, and the fifth verification result, determine whether there is any risk to the workpiece being tightened.

[0134] In some embodiments, the second verification result specifically includes a first risk score, the third verification result specifically includes a second risk score, the fourth verification result specifically includes a third risk score, and the fifth verification result specifically includes a fourth risk score. The first, second, third, and fourth risk scores can be weighted and summed to obtain the target risk score for the workpiece being tightened. If the target risk score for the workpiece being tightened is greater than a preset risk score threshold, then the workpiece being tightened is determined to have a risk; if the target risk score for the workpiece being tightened is less than or equal to the preset risk score threshold, then the workpiece being tightened is determined to have no risk.

[0135] In some embodiments, the second verification result includes a second risk classification label, the third verification result includes a third risk classification label, the fourth verification result includes a fourth risk classification label, and the fifth verification result includes a fifth risk classification label. The risk label percentage (i.e., the ratio of the number of risk classification labels indicating risk to the total number of risk classification labels) can be calculated based on the second, third, fourth, and fifth risk classification labels. If the risk label percentage is greater than a preset percentage threshold (e.g., 1 / 2), it is determined that there is a risk in the workpiece being tightened; if the risk label percentage is less than or equal to the preset percentage threshold, it is determined that there is no risk in the workpiece being tightened.

[0136] 2043A. If the workpiece being tightened is a compliant product and there is a risk, then the risk level of the workpiece being tightened is determined based on the first risk score, the second risk score, the third risk score, and the fourth risk score.

[0137] For example, the first risk score, the second risk score, the third risk score and the fourth risk score can be weighted and summed to obtain the target risk score of the workpiece to be tightened. Then, the target risk score is directly used as the risk level, or the risk level of the workpiece to be tightened is finally determined according to the level scoring range that the target risk score falls into (e.g., high, medium and low risks correspond to scoring ranges 1, 2 and 3 respectively).

[0138] (2) In some embodiments, the process inspection results, in addition to indicating whether the workpiece being tightened is a compliant product and whether there is a risk, can also indicate the abnormal stage, the cause of the abnormality, and / or process optimization suggestions when the workpiece is a non-compliant product or when there is a risk. In this case, step 204 may specifically include the following steps 2041B to 2044B: 2041B. Based on the first verification result, conduct a compliance test on the workpiece being tightened.

[0139] 2042B. If the workpiece being tightened is a compliant product, then based on the second verification result, the third verification result, the fourth verification result, and the fifth verification result, determine whether there is any risk associated with the workpiece being tightened.

[0140] 2043B. If the workpiece being tightened is a compliant product and there is a risk, then the risk level of the workpiece being tightened is determined based on the first risk score, the second risk score, the third risk score, and the fourth risk score.

[0141] The implementation of steps 2041B to 2043B is similar to that of steps 2041A to 2043A. For details, please refer to the relevant explanations above, which will not be repeated here.

[0142] 2044B. If the workpiece being tightened is a non-compliant product, or if the workpiece being tightened is a compliant product but there is a risk, then the second verification result, the third verification result, the fourth verification result and / or the fifth verification result are combined to obtain the abnormal stage, abnormal attribution and / or process optimization suggestion information of the workpiece being tightened.

[0143] Taking the process inspection results as an example of indicating abnormal stages in the current workpiece tightening, the second verification result specifically includes the first abnormal probability of each process stage, and the fourth verification result includes the second abnormal probability of each process stage. The second and fourth verification results can be combined to determine the abnormal stage of the current workpiece tightening. For example, based on the first and second abnormal probabilities of each process stage, a target abnormal probability for each process stage is determined. Based on the target abnormal probability of each process stage, it is determined whether each process stage is an abnormal stage: if the target abnormal probability of process stage i is greater than a preset abnormal probability threshold, then process stage i is determined to be an abnormal stage; if the target abnormal probability of process stage i is less than or equal to the preset abnormal probability threshold, then process stage i is determined to be a normal stage. This process can be repeated to filter out all abnormal stages in the current workpiece tightening.

[0144] For example, the process inspection results are also used to indicate the cause of the abnormality in the tightening of the workpiece. The third verification result also includes the probability of each candidate abnormality type. The top N candidate abnormality types with the highest probabilities can be output as the cause of the abnormality in the tightening of the workpiece.

[0145] Taking the process inspection results as an example of using them to indicate process optimization suggestions for the tightened workpiece, the second verification result specifically includes the first anomaly probability of each process stage, and the fourth verification result includes the second anomaly probability, curve similarity (including the first curve similarity, the second curve similarity, and the third curve similarity), and stage deviation (including the first deviation stage, the second deviation stage, and the third deviation stage) of each process stage. The second and fourth verification results can be combined to determine the abnormal stage of the tightened workpiece. The third verification result also includes the probability of each candidate anomaly type, and the fifth verification result includes the fifth risk classification label. The top N candidate anomaly types with the highest probabilities of the candidate anomaly types output by the third verification result are used as the anomaly attribution for the tightened workpiece, and the fifth verification result is used to assist in verifying the accuracy of the anomaly attribution (if the fifth risk classification label is risky, it proves that the top N candidate anomaly types belong to the anomaly attribution for the tightened workpiece; if the fifth risk classification label is normal, it proves that the top N candidate anomaly types do not belong to the anomaly attribution for the tightened workpiece). The above verification results can be combined to obtain the process optimization suggestions for the tightened workpiece. For example, firstly, based on the anomaly probabilities of the second and fourth verification results, the target anomaly probability for each process stage is determined, and anomaly stages are selected. Secondly, combining the third and fifth verification results, the top N candidate anomaly types by probability are selected as core anomaly attributions, and the optimization targets are determined by combining the curve similarity and stage deviation of the second verification results. Finally, based on the deviation value and deviation rate between the current tightening characteristics and the standard process template, the optimization direction and adjustment range are determined, and targeted process optimization suggestions are generated. The process optimization suggestions include at least one of tightening parameter adjustment suggestions, process template optimization suggestions, and process control optimization suggestions. The tightening parameter adjustment suggestions include at least one of torque final value window and angle final value window adjustment suggestions. The process template optimization suggestions include at least one of torque reference curve, angle reference curve, and torque-angle coupling reference curve adjustment suggestions. For example, process optimization suggestions include tightening parameter adjustment suggestions, such as adjustments to the final torque value window. If the anomaly is attributed to "low final torque value," and the final tightening stage is an abnormal stage, then a suggestion to "increase the lower limit of the target torque in the final tightening stage" is generated to automatically propose adjustments to the final torque value window. Similarly, process optimization suggestions include tightening parameter adjustment suggestions, such as adjustments to the final angle value window. If the anomaly is attributed to "low final angle value," and the final tightening stage is an abnormal stage, then a suggestion to "increase the lower limit of the target angle in the final tightening stage" is generated to automatically propose adjustments to the final angle value window.For example, process optimization suggestions include process template optimization suggestions. If the similarity between multiple types of curves and pre-configured reference curves (i.e., torque reference curve, angle reference curve, and torque-angle coupling reference curve) is lower than a preset threshold, and deviations exist in multiple stages, then a suggestion to "update the preset standard template library based on recent qualified samples and regenerate standardized curves" is generated. This allows for the automatic provision of adjustment suggestions for at least one of the torque reference curve, angle reference curve, and torque-angle coupling reference curve. Another example is process control optimization suggestions. If the anomaly is attributed to "stiffness mutation anomaly," and the anomaly occurs during the final tightening stage, then a suggestion to "increase the sampling frequency during the final tightening stage and adjust the torque filtering parameters" is generated.

[0146] Furthermore, the process inspection results can also include the quality grade, the third verification results can also include the quality grade, and the quality grade included in the third verification results can also be output as the quality grade of the workpiece being tightened.

[0147] As can be seen from the above, firstly, by utilizing the first process context information and the first tightening process data to construct a first tightening feature set and a first temporal feature set, these are used to obtain the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the temporal depth model, and the fifth verification result of the unsupervised anomaly detection model, thereby determining the process inspection result of the tightened workpiece. Since the first tightening feature set and the first temporal feature set contain process context information and tightening process data, they can identify hidden anomalies that are difficult to detect using traditional final value methods. In scenarios where the final value is qualified but the process is abnormal, abnormal tightened workpieces can be identified, such as when the rule engine determines that the final value verification result is qualified. If the second verification result of the similarity model (such as the deviation stage identified based on curve similarity, the first risk score, the second risk classification label, and the first anomaly probability of each process stage), the third verification result of the feature classification model (such as the second risk score, the third risk classification label, the probability of each candidate anomaly type, and the quality level), the fourth verification result of the time-series depth model (such as the second anomaly probability of each process stage, the third risk score, and the fourth risk classification label), or the fifth verification result of the unsupervised anomaly detection model (such as the fourth risk score and the fifth risk classification label) indicates an anomaly, it is still possible to identify workpieces that are qualified in the final value but abnormal in the process. Therefore, it can reduce the misjudgment rate of tightening results (i.e., process inspection results) to a certain extent. Secondly, by using the first verification result based on the rule engine, the second verification result based on the similarity model, the third verification result based on the feature classification model, the fourth verification result based on the time-series deep model, and the fifth verification result based on the unsupervised anomaly detection model, the process inspection result of the tightened workpiece is determined. This can integrate multiple models or engines to make decisions on the process inspection result of the tightened workpiece, reduce the problem of misjudgment that is prone to occur when making decisions based on a single model, thereby improving the accuracy of the tightening result judgment and reducing the risk of abnormalities in the tightened workpiece.

[0148] Furthermore, to better implement the process inspection method for tightening workpieces in the embodiments of this application, based on the process inspection method for tightening workpieces, the embodiments of this application also provide a process inspection device for tightening workpieces, such as... Figure 6 The diagram shown is a schematic representation of an embodiment of the process inspection device 600 for tightening workpieces provided in this application. The process inspection device 600 for tightening workpieces includes: The acquisition unit 601 is used to acquire the first process context information and the first tightening process data of the workpiece being tightened. The first process context information includes identification information, configuration information and batch information. The first tightening process data includes the torque curve, angle curve, rotation speed curve and stage markers of the entire tightening process of the workpiece being tightened. Construction unit 602 is used to construct features based on the first process context information and the first tightening process data to obtain the first tightening feature set and the first temporal feature of the workpiece being tightened this time; The identification unit 603 is used to obtain, based on the first tightening feature set and the first temporal features, the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the temporal depth model and the fifth verification result of the unsupervised anomaly detection model; Decision unit 604 is used to make a fusion decision based on the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result to obtain the process inspection result of the workpiece being tightened. The fusion decision process includes making a compliance decision on the workpiece being tightened based on the first verification result, and determining whether the workpiece being tightened has at least one of the following information: risk, risk level, abnormal stage, abnormal attribution, and process optimization suggestion information.

[0149] In some embodiments, the first tightening feature set includes a first basic tightening feature, a first time-related feature, a first tightening process feature, and a first contextual process feature; the construction unit 602 is used for: Based on the first tightening process data, the first basic tightening characteristics of the workpiece being tightened are determined. The first basic tightening characteristics include the final torque value and the final angle value, the duration of each process stage, the total process duration, and the stage switching characteristics. The stage switching characteristics include at least one of the stage switching physical quantity, the stage switching duration, and the stage switching sequence. Based on the first tightening process data, the first time-related features of the workpiece being tightened are determined, wherein the first time-related features include at least one of the current torque curve, current angle curve, current rotational speed curve, and current torque-angle coupling curve; Based on the first tightening process data, the first tightening process characteristics of the workpiece being tightened are determined, wherein the first tightening process characteristics include at least one of stiffness characteristics, ripple frequency domain characteristics, and torque-angle coupling characteristics. Based on the first process context information, the first context process feature of the workpiece being tightened is determined; Based on the first time-related features, equal-length serialization processing is performed to obtain the first time-series features.

[0150] In some embodiments, the identification unit 603 is used for: The rule engine determines the final value verification result based on the pre-configured target final value window and the final torque and angle values ​​of the workpiece being tightened. The target final value window is a final value window that matches the process scenario parameters of the workpiece being tightened. The target final value window includes at least one of the torque final value window and the angle final value window. The rule engine determines the stage verification result based on the pre-configured target duration range for each stage and the duration value of each process stage, wherein the target duration range for each stage is a duration range that matches the process scenario parameters. The timeout verification result is determined by the rule engine based on the pre-configured target total duration range and the total process duration value, wherein the target total duration range is the total duration range that matches the process scenario parameters; The rule engine determines the stage switching verification result based on the pre-configured target stage switching rule and the stage switching feature, wherein the target stage switching rule is a stage switching rule that matches the process scenario parameters. Based on the final value verification result, the stage verification result, the timeout verification result, and the stage switching verification result, the first verification result is determined, wherein the first verification result includes a first risk classification label. If at least one of the final value verification result, the stage verification result, the timeout verification result, and the stage switching verification result is unqualified, then the first risk classification label is risk.

[0151] In some embodiments, the first time-related feature includes the current torque curve, the current angle curve, and the current torque-angle coupling curve; the second verification result includes a first risk score, a second risk classification label, and / or the anomaly probability of each process stage; the identification unit 603 is used for: Using the similarity model, the torque curve is matched with the current torque curve according to the pre-configured torque reference curve, and the first curve similarity and the first deviation stage are output. Using the similarity model, the current angle curve is matched with a pre-configured angle reference curve to output the second curve similarity and the second deviation stage. Using the similarity model, the torque angle coupling reference curve is matched with the current torque angle coupling curve according to the pre-configured torque angle coupling reference curve, and the third curve similarity and third deviation stage are output. Based on the similarity of the first curve, the similarity of the second curve, and the similarity of the third curve, a first risk score and / or a second risk classification label are determined; and / or, Based on the first deviation stage, the second deviation stage, and the third deviation stage, the first abnormal probability of each process stage of the current workpiece tightening is determined.

[0152] In some embodiments, the process inspection device 600 for tightening workpieces further includes a generation unit (not shown in the figure), which is used for: Using torque curves from multiple standard samples in a pre-defined standard template library, a standardized torque curve is generated after feature extraction, curve normalization, and mean fitting. The standard samples are tightened workpiece samples with normal process inspection results. The standardized torque curve is configured as the torque reference curve of the similarity model. The scenario parameters of each standard sample used to generate the standardized torque curve are the same as the scenario parameters of the workpiece being tightened. The scenario parameters include at least one of the following: tightening gun number, workstation number, tightening program number, process version, product model, and production shift. And / or, using the angle curves of multiple standard samples in the preset standard template library, after feature extraction, curve normalization, and mean fitting, a standardized angle curve is generated. The standardized angle curve is configured as the angle reference curve of the similarity model. The scene parameters of each standard sample used to generate the standardized angle curve are the same as the scene parameters of the workpiece being tightened in this case. And / or, using the torque-angle coupling curves of multiple standard samples in a preset standard template library, a standardized torque-angle coupling curve is generated after feature extraction, curve normalization, and mean fitting. The standardized torque-angle coupling curve is configured as the torque-angle coupling reference curve of the similarity model. The scene parameters of each standard sample used to generate the standardized torque-angle coupling curve are the same as the scene parameters of the workpiece being tightened.

[0153] In some embodiments, the identification unit 603 is used for: Based on the first tightening process features and the first context process features, a combined encoding is performed to obtain the first multi-dimensional feature vector of the workpiece being tightened in this instance. The feature classification model is used to classify based on the first multi-dimensional feature vector and output the third verification result. The third verification result includes a second risk score, a third risk classification label, the probability and / or quality level of each candidate anomaly type, and the candidate anomaly type includes at least one of under-tightening, over-tightening, stripped thread, floating thread, missing thread, fit anomaly, and tool anomaly.

[0154] In some embodiments, the identification unit 603 is used for: Using the time-series deep model, anomaly detection is performed based on the first time-series features, and the fourth verification result is output. The fourth verification result includes the second anomaly probability, the third risk score, and the fourth risk classification label for each process stage.

[0155] In some embodiments, the identification unit 603 is used for: Based on the first tightening process features and the first context process features, a combined encoding is performed to obtain the first multi-dimensional feature vector of the workpiece being tightened in this instance. The unsupervised anomaly detection model performs anomaly detection based on the first multi-dimensional feature vector and the first temporal feature, and outputs the fifth verification result, which includes a fourth risk score and a fifth risk classification label.

[0156] In some embodiments, the process inspection result is used to indicate whether the workpiece being tightened is a compliant product, and whether the workpiece being tightened has at least one of the following: risk level, abnormal stage, abnormal attribution, and process optimization suggestion information; the second verification result includes a first risk score and a second risk classification label, the fourth verification result includes a third risk score and a fourth risk classification label, and the fifth verification result includes a fourth risk score and a fifth risk classification label; the decision unit 604 is used for: Based on the first verification result, a compliance test is performed on the workpiece being tightened. If the workpiece being tightened is a compliant product, then based on the second risk classification label, the third risk classification label, the fourth risk classification label, and the fifth risk classification label, it is determined whether the workpiece being tightened poses a risk. If the workpiece being tightened is a compliant product and there is a risk, then the risk level of the workpiece being tightened is determined based on the first risk score, the second risk score, the third risk score, and the fourth risk score. If the workpiece being tightened is a non-compliant product, or if the workpiece being tightened is a compliant product but poses a risk, then the second verification result, the third verification result, the fourth verification result, and / or the fifth verification result are combined to obtain the abnormal stage, abnormal attribution, and / or process optimization suggestion information for the workpiece being tightened. The process optimization suggestion information includes at least one of tightening parameter adjustment suggestions and process template optimization suggestions. The tightening parameter adjustment suggestions include at least one of torque final value window and angle final value window adjustment suggestions. The process template optimization suggestions include at least one of torque reference curve, angle reference curve, and torque-angle coupling reference curve adjustment suggestions.

[0157] In practice, each of the above units can be implemented as an independent entity or can be combined arbitrarily to be implemented as the same or several entities. For the specific implementation of each of the above units, please refer to the previous embodiment of the process inspection method for tightening workpieces, which will not be repeated here.

[0158] Those skilled in the art will understand that all or part of the steps in the above-described process inspection method for tightening workpieces can be completed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0159] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of computer programs that can be loaded by a processor to execute any of the process inspection methods for tightening workpieces provided in embodiments of this application. For example, the computer program can be loaded by a processor to perform the following steps: The process involves acquiring the first process context information and the first tightening process data of the workpiece to be tightened. The first process context information includes identification information, configuration information, and batch information. The first tightening process data includes the torque curve, angle curve, rotational speed curve, and stage markers for the entire tightening process of the workpiece. Based on the first process context information and the first tightening process data, feature construction is performed to obtain the first tightening feature set and the first temporal feature of the workpiece. Based on the first tightening feature set and the first temporal feature, the first verification result of the rule engine, the second verification result of the similarity model, and the feature classification model are obtained. The third verification result, the fourth verification result of the time-series deep model, and the fifth verification result of the unsupervised anomaly detection model are used to make a fusion decision based on the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result to obtain the process inspection result of the tightened workpiece. The fusion decision process includes making a compliance decision for the tightened workpiece based on the first verification result, and determining whether the tightened workpiece has at least one of the following information: risk, risk level, anomaly stage, anomaly attribution, and process optimization suggestion information based on the second verification result, the third verification result, the fourth verification result, and the fifth verification result.

[0160] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0161] In the above embodiments of the process inspection method for tightening workpieces, computer-readable storage media, and electronic devices, the descriptions of each embodiment have different focuses. For parts not described in detail in a particular embodiment, please refer to the relevant descriptions of other embodiments. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes and beneficial effects of the computer-readable storage media, electronic devices, and their corresponding units described above can be referred to the description of the process inspection method for tightening workpieces in the above embodiments, and will not be repeated here.

[0162] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.

[0163] The foregoing has provided a detailed description of a process inspection method, apparatus, electronic device, and computer-readable storage medium for tightening workpieces according to embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method of process verification of tightening of a workpiece, characterized by, The method includes: The first process context information and the first tightening process data of the workpiece to be tightened are obtained. The first process context information includes identification information, configuration information and batch information. The first tightening process data includes the torque curve, angle curve, speed curve and stage mark of the entire tightening process of the workpiece to be tightened. Based on the first process context information and the first tightening process data, feature construction is performed to obtain the first tightening feature set and the first temporal feature of the workpiece being tightened in this operation. Based on the first tightening feature set and the first temporal features, the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the temporal deep model, and the fifth verification result of the unsupervised anomaly detection model are obtained. Based on the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result, a fusion decision is made to obtain the process inspection result of the workpiece being tightened. The fusion decision process includes making a compliance decision on the workpiece being tightened based on the first verification result, and determining whether the workpiece being tightened has at least one of the following information: risk, risk level, abnormal stage, abnormal attribution, and process optimization suggestion information.

2. The process inspection method for tightening workpieces according to claim 1, characterized in that, The first tightening feature set includes a first basic tightening feature, a first time-related feature, a first tightening process feature, and a first contextual process feature; The feature construction based on the first process context information and the first tightening process data yields a first tightening feature set and a first temporal feature set for the workpiece being tightened, including: Based on the first tightening process data, the first basic tightening characteristics of the workpiece being tightened are determined. The first basic tightening characteristics include the final torque value and the final angle value, the duration of each process stage, the total process duration, and the stage switching characteristics. The stage switching characteristics include at least one of the stage switching physical quantity, the stage switching duration, and the stage switching sequence. Based on the first tightening process data, the first time-related features of the workpiece being tightened are determined, wherein the first time-related features include at least one of the current torque curve, current angle curve, current rotational speed curve, and current torque-angle coupling curve; Based on the first tightening process data, the first tightening process characteristics of the workpiece being tightened are determined, wherein the first tightening process characteristics include at least one of stiffness characteristics, ripple frequency domain characteristics, and torque-angle coupling characteristics. Based on the first process context information, the first context process feature of the workpiece being tightened is determined; Based on the first time-related features, equal-length serialization processing is performed to obtain the first time-series features.

3. The process inspection method for tightening workpieces according to claim 2, characterized in that, The first verification result is determined in the following way: The rule engine determines the final value verification result based on the pre-configured target final value window and the final torque and angle values ​​of the workpiece being tightened. The target final value window is a final value window that matches the process scenario parameters of the workpiece being tightened. The target final value window includes at least one of the torque final value window and the angle final value window. The rule engine determines the stage verification result based on the pre-configured target duration ranges for each stage and the duration values ​​of each process stage, wherein the target duration ranges for each stage are duration ranges that match the process scenario parameters. The timeout verification result is determined by the rule engine based on the pre-configured target total duration range and the total process duration value, wherein the target total duration range is the total duration range that matches the process scenario parameters. The rule engine determines the stage switching verification result based on the pre-configured target stage switching rule and the stage switching feature, wherein the target stage switching rule is a stage switching rule that matches the process scenario parameters. Based on the final value verification result, the stage verification result, the timeout verification result, and the stage switching verification result, the first verification result is determined, wherein the first verification result includes a first risk classification label. If at least one of the final value verification result, the stage verification result, the timeout verification result, and the stage switching verification result is unqualified, then the first risk classification label is risk.

4. The process inspection method for tightening workpieces according to claim 2, characterized in that, The first time-related features include the current torque curve, the current angle curve, and the current torque-angle coupling curve; the second verification result includes the first risk score, the second risk classification label, and / or the anomaly probability of each process stage. The second verification result is determined in the following way: Using the similarity model, the torque curve is matched with the current torque curve according to the pre-configured torque reference curve, and the first curve similarity and the first deviation stage are output. Using the similarity model, the current angle curve is matched with a pre-configured angle reference curve to output the second curve similarity and the second deviation stage. Using the similarity model, the torque angle coupling reference curve is matched with the current torque angle coupling curve according to the pre-configured torque angle coupling reference curve, and the third curve similarity and third deviation stage are output. Based on the first curve similarity, the second curve similarity, and the third curve similarity, a first risk score and / or a second risk classification label are determined; And / or, Based on the first deviation stage, the second deviation stage, and the third deviation stage, the first abnormal probability of each process stage of the current workpiece tightening is determined.

5. The process inspection method for tightening workpieces according to claim 4, characterized in that, The method further includes: Using torque curves from multiple standard samples in a pre-defined standard template library, a standardized torque curve is generated after feature extraction, curve normalization, and mean fitting. The standard samples are tightened workpiece samples with normal process inspection results. The standardized torque curve is configured as the torque reference curve of the similarity model. The scenario parameters of each standard sample used to generate the standardized torque curve are the same as the scenario parameters of the workpiece being tightened. The scenario parameters include at least one of the following: tightening gun number, workstation number, tightening program number, process version, product model, and production shift. And / or, using the angle curves of multiple standard samples in the preset standard template library, after feature extraction, curve normalization, and mean fitting, a standardized angle curve is generated. The standardized angle curve is configured as the angle reference curve of the similarity model. The scene parameters of each standard sample used to generate the standardized angle curve are the same as the scene parameters of the workpiece being tightened in this case. And / or, using the torque-angle coupling curves of multiple standard samples in a preset standard template library, a standardized torque-angle coupling curve is generated after feature extraction, curve normalization, and mean fitting. The standardized torque-angle coupling curve is configured as the torque-angle coupling reference curve of the similarity model. The scene parameters of each standard sample used to generate the standardized torque-angle coupling curve are the same as the scene parameters of the workpiece being tightened.

6. The process inspection method for tightening workpieces according to claim 2, characterized in that, The third verification result is determined in the following manner: Based on the first tightening process features and the first context process features, a combined encoding is performed to obtain the first multi-dimensional feature vector of the workpiece being tightened in this instance. The feature classification model is used to classify based on the first multi-dimensional feature vector and output the third verification result. The third verification result includes a second risk score, a third risk classification label, the probability and / or quality level of each candidate anomaly type, and the candidate anomaly type includes at least one of under-tightening, over-tightening, stripped thread, floating thread, missing thread, fit anomaly, and tool anomaly.

7. The process inspection method for tightening workpieces according to claim 2, characterized in that, The fourth verification result is determined in the following manner: Using the time-series deep model, anomaly detection is performed based on the first time-series features, and the fourth verification result is output. The fourth verification result includes the second anomaly probability, the third risk score, and the fourth risk classification label for each process stage.

8. The process inspection method for tightening workpieces according to claim 2, characterized in that, The fifth verification result is determined in the following manner: Based on the first tightening process features and the first context process features, a combined encoding is performed to obtain the first multi-dimensional feature vector of the workpiece being tightened in this instance. The unsupervised anomaly detection model performs anomaly detection based on the first multi-dimensional feature vector and the first temporal feature, and outputs the fifth verification result, which includes a fourth risk score and a fifth risk classification label.

9. The process inspection method for tightening workpieces according to claim 1, characterized in that, The process inspection results are used to indicate whether the workpiece being tightened is a compliant product, and whether the workpiece being tightened has at least one of the following: risk, risk level, abnormal stage, abnormal attribution, and process optimization suggestions; the second verification result includes a first risk score and a second risk classification label, the third verification result includes a second risk score and a third risk classification label, the fourth verification result includes a third risk score and a fourth risk classification label, and the fifth verification result includes a fourth risk score and a fifth risk classification label; The process inspection result of the tightened workpiece is obtained by fusing the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result, including: Based on the first verification result, a compliance test is performed on the workpiece being tightened. If the workpiece being tightened is a compliant product, then based on the second risk classification label, the third risk classification label, the fourth risk classification label, and the fifth risk classification label, it is determined whether the workpiece being tightened poses a risk. If the workpiece being tightened is a compliant product and there is a risk, then the risk level of the workpiece being tightened is determined based on the first risk score, the second risk score, the third risk score, and the fourth risk score. If the workpiece being tightened is a non-compliant product, or if the workpiece being tightened is a compliant product but poses a risk, then the second verification result, the third verification result, the fourth verification result, and / or the fifth verification result are combined to obtain the abnormal stage, abnormal attribution, and / or process optimization suggestion information for the workpiece being tightened. The process optimization suggestion information includes at least one of tightening parameter adjustment suggestions and process template optimization suggestions. The tightening parameter adjustment suggestions include at least one of torque final value window and angle final value window adjustment suggestions. The process template optimization suggestions include at least one of torque reference curve, angle reference curve, and torque-angle coupling reference curve adjustment suggestions.

10. A process inspection device for tightening workpieces, characterized in that, The process inspection device includes: The acquisition unit is used to acquire the first process context information and the first tightening process data of the workpiece being tightened. The first process context information includes identification information, configuration information and batch information. The first tightening process data includes the torque curve, angle curve, rotation speed curve and stage markers of the entire tightening process of the workpiece being tightened. The construction unit is used to construct features based on the first process context information and the first tightening process data to obtain the first tightening feature set and the first temporal feature of the workpiece being tightened this time. The identification unit is used to obtain the first verification result of the rule engine, the second verification result of the similarity model, the third verification result of the feature classification model, the fourth verification result of the temporal depth model, and the fifth verification result of the unsupervised anomaly detection model based on the first tightening feature set and the first temporal features. The decision-making unit is used to make a fusion decision based on the first verification result, the second verification result, the third verification result, the fourth verification result, and the fifth verification result to obtain the process inspection result of the workpiece being tightened. The fusion decision-making process includes making a compliance decision on the workpiece being tightened based on the first verification result, and determining whether the workpiece being tightened has at least one of the following information: risk, risk level, abnormal stage, abnormal attribution, and process optimization suggestion information.

11. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, it executes the process inspection method for tightening a workpiece as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, It stores a computer program, which is loaded by a processor to execute the process inspection method for tightening workpieces as described in any one of claims 1 to 9.

Images