An automated method and system for detecting the forming dimensions of automotive seat hardware stamping parts

Abstract

This application relates to the field of optical metrology technology, specifically to an automated inspection method and system for the forming dimensions of automotive seat hardware stamping parts. The method includes: acquiring a three-dimensional point cloud dataset and strain data of the workpiece during the automotive seat hardware stamping process; utilizing the abnormal deviation characteristics between the strain data, the three-dimensional point cloud dataset, and the standard point cloud dataset at each data acquisition moment to obtain a springback risk sequence and a dimensional deviation sequence; using the results of multiple linear regression of the springback risk sequence, its first-order difference sequence, and the dimensional deviation sequence, and combining this with an exponentially weighted moving average control chart to calculate a dynamic control upper limit for the springback risk sequence, determining the springback sensitivity index of the workpiece deformation at each data acquisition moment; and using the springback sensitivity index to implement an adaptive threshold compensation strategy to achieve automated workpiece inspection. This application aims to improve the reliability and efficiency of automated workpiece inspection during automotive seat hardware stamping.

Description

Technical Field

[0001] This application relates to the field of optical metrology technology, specifically to an automated detection method and system for the forming dimensions of automotive seat hardware stamping parts. Background Technology

[0002] In the automotive manufacturing industry, stamping is the first step in automobile production, and the quality and inspection of stamping have become primary goals for major automakers. In the early days, the main inspection method was the introduction of coordinate measuring machines (CMMs) to improve measurement accuracy. In recent years, with the rapid development of 3D machine vision, high-speed measurement of automotive seat stamping parts has been achieved, enabling online automated inspection in a non-contact manner. This allows for real-time monitoring of the full dimensional changes and fluctuations of the workpiece, enabling faster and more accurate detection of wear and dimensional deviations, thus ensuring the assembly accuracy of the seat assembly.

[0003] When using 3D machine vision to inspect the dimensions of automotive metal stamping parts, the residual stress in the workpiece after demolding undergoes nonlinear release, causing continuous deformation and springback of the workpiece's geometry. This deformation and springback further interferes with automated dimensional inspection, reducing the accuracy of static inspection methods. Current technologies primarily rely on closed-loop compensation systems to collect historical data and analyze springback patterns. However, when material dimensions change abruptly, the established model lacks prior knowledge of the new conditions, leading to model failure. This makes it difficult to effectively distinguish between acceptable springback (meeting design tolerances) and unacceptable springback (causing dimensional deviations or posing a risk of continued deformation) generated during residual stress release. This results in a risk of misjudging workpiece dimensional compliance, leading to the erroneous scrapping of acceptable parts and the retention of potentially defective parts, severely limiting the reliability and efficiency of automated inspection. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides an automated method and system for detecting the forming dimensions of automotive seat metal stamping parts. The specific technical solution adopted is as follows:

[0005] In a first aspect, one embodiment of this application provides an automated inspection method for the forming dimensions of automotive seat hardware stamping parts, the method comprising the following steps:

[0006] Collect 3D point cloud datasets and strain data sequences of the workpiece at each data acquisition moment during the stamping process of automotive seat hardware; and generate standard point cloud datasets of the workpiece based on the original CAD model.

[0007] By utilizing the dynamic mode decomposition features of the strain data sequence at each data acquisition moment and the abnormal deviation features between the 3D point cloud dataset and the standard point cloud dataset, the time-varying rebound risk level at each data acquisition moment is determined. Combined with the average deviation features at each data acquisition moment, the rebound risk sequence and size deviation sequence at each data acquisition moment are obtained.

[0008] The dynamic control upper limit is calculated based on the exponentially weighted moving average control chart for the springback risk sequence; multiple linear regression is performed with the springback risk sequence and its first-order difference sequence as independent variables and the size deviation sequence as the dependent variable, and the regression coefficients of the two independent variables are output. Combined with the dynamic control upper limit, the springback sensitivity index of the workpiece deformation at each data acquisition time is determined.

[0009] An adaptive threshold compensation strategy is executed based on the relationship between the springback sensitivity index and the pre-set two-level thresholds to achieve automated detection of the workpiece.

[0010] Preferably, the strain data sequence at each data acquisition moment is composed of strain data from all acquisition moments prior to each data acquisition moment, sorted in ascending chronological order.

[0011] Preferably, the method for determining the time-varying rebound risk at each data acquisition moment is as follows:

[0012] The dominant mode decay coefficient of the strain data sequence at each data acquisition moment is obtained by using a dynamic mode decomposition algorithm;

[0013] Obtain the abnormal deviation characteristics between the 3D point cloud dataset and the standard point cloud dataset at each data acquisition time;

[0014] Calculate the exponent of an exponential function with the negative of the dominant mode decay coefficient as the base of the natural constant, and multiply the result of the exponential function calculation with the abnormal deviation characteristic as the time-varying rebound risk degree at each data acquisition moment.

[0015] Preferably, the truncation order of the dynamic mode decomposition is 5.

[0016] Preferably, the method for obtaining the abnormal deviation characteristics is as follows:

[0017] For any point in the 3D point cloud dataset at each data acquisition time, the minimum 3D Euclidean distance between that point and a point in the standard point cloud dataset is used as an element in the deviation dataset.

[0018] The local outlier detection algorithm is used to obtain the local outlier factor of each element in the biased dataset;

[0019] The mean of the local outlier factors of all elements in the deviation dataset is used as the abnormal deviation feature.

[0020] Preferably, the method for obtaining the rebound risk sequence and size deviation sequence at each data acquisition moment includes:

[0021] The mean of all elements in the deviation dataset between the 3D point cloud dataset and the standard point cloud dataset at each data acquisition time is denoted as the first mean.

[0022] For each data acquisition moment, the time-varying rebound risk and the first mean calculated from all acquisition moments prior to each data acquisition moment are sorted in ascending order of time to obtain the rebound risk sequence and size deviation sequence for each data acquisition moment.

[0023] Preferably, the formula for calculating the springback sensitivity index of the workpiece deformation at each data acquisition moment is:

[0024]

[0025] Where B is the springback sensitivity index of the workpiece deformation at each acquisition time. and These are the regression coefficients of the two independent variables output by the multiple linear regression. This represents the upper limit of dynamic control output from the exponentially weighted moving average control chart, where e is the natural constant. It is the hyperbolic tangent function.

[0026] Preferably, the method for implementing an adaptive threshold compensation strategy based on the relationship between the rebound sensitivity index and a pre-set two-level threshold is as follows:

[0027] The two-level threshold includes a first threshold T1 and a second threshold T2, where 0 <T1<T2<1;

[0028] When B is less than or equal to the first threshold, maintain the original operation;

[0029] When B is greater than the first threshold and less than or equal to the second threshold, real-time monitoring and early warning will be activated.

[0030] When B is greater than the second threshold, the Yoshida-Uemori bounce prediction model is triggered to dynamically adjust the 3D point cloud dataset.

[0031] Preferably, the Yoshida-Uemori rebound prediction model outputs the predicted rebound amount of each measurement point in the X, Y, and Z directions, and superimposes the predicted rebound amount onto the coordinates of each point in the three-dimensional point cloud dataset.

[0032] Secondly, another embodiment of this application also provides an automated inspection system for the forming dimensions of automotive seat hardware stamping parts, implementing the aforementioned automated inspection method for the forming dimensions of automotive seat hardware stamping parts. The system includes:

[0033] The high-precision 3D scanning unit is a 3D line laser scanner installed at the end of the robotic arm;

[0034] The embedded stress monitoring unit is a piezoelectric sensor pre-embedded at the corner where the mold seat side plate connects to the horizontal slide rail.

[0035] A high-speed data acquisition unit is used to simultaneously acquire the output signals of the high-precision three-dimensional scanning unit and the embedded stress monitoring unit;

[0036] The central analysis unit is used to execute all the computational steps of the method described above.

[0037] It also includes a human-computer interaction interface for real-time display of point cloud deviation maps and strain curves, and for outputting warnings and data tracing based on the threshold step size strategy.

[0038] This application has at least the following beneficial effects:

[0039] 1. To address the springback problem of workpiece materials, a dynamic mode decomposition algorithm is used to extract the strain attenuation coefficient to characterize the stress release rate. Combined with the local outlier factor algorithm to analyze the local outlier factor of the workpiece point cloud deviation, the time-varying springback risk of the workpiece at each moment is calculated, quantifying the coupling risk between workpiece springback and residual stress attenuation, and solving the problem that the established model is difficult to capture the nonlinear springback of the workpiece.

[0040] 2. Due to the model failure caused by sudden changes in workpiece dimensions, the upper limit of springback risk control is calculated using an exponentially weighted moving average control chart. The influence of springback risk and its rate of change on dimensional deviation is fitted by multiple linear regression. Furthermore, the springback sensitivity index is calculated using both to characterize the sensitivity intensity of sudden changes in material properties and springback instability, thus eliminating the adaptive interference of historical models to sudden working conditions.

[0041] 3. A threshold response mechanism is designed based on the calculated springback sensitivity index. This mechanism ensures that the original compensation model continues to operate when the workpiece size is relatively insensitive to the risk of springback; when the workpiece size is relatively sensitive to the risk of springback, an audible and visual warning is activated on the human-machine interface; and when the workpiece size is extremely sensitive to the risk of springback, springback simulation is triggered. This ensures that the scrap rate caused by workpiece springback is reduced while maintaining continuous production line operation. Attached Figure Description

[0042] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the 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.

[0043] Figure 1 This is a flowchart illustrating an automated detection method for the forming dimensions of stamped automotive seat hardware parts, provided as an embodiment of this application. Detailed Implementation

[0044] Example 1

[0045] One embodiment of this application provides an automated inspection system for the forming dimensions of stamped automotive seat hardware parts. This system implements another embodiment of the automated inspection method for the forming dimensions of stamped automotive seat hardware parts. The system includes:

[0046] The high-precision 3D scanning unit is a 3D line laser scanner installed at the end of the robotic arm;

[0047] The embedded stress monitoring unit is a piezoelectric sensor pre-embedded at the corner where the mold seat side plate connects to the horizontal slide rail.

[0048] A high-speed data acquisition unit is used to simultaneously acquire the output signals of the high-precision three-dimensional scanning unit and the embedded stress monitoring unit;

[0049] The central analysis unit is used to perform all the calculation steps of an automated detection method for the forming dimensions of automotive seat hardware stamping parts provided in another embodiment below;

[0050] It also includes a human-computer interaction interface for real-time display of point cloud deviation maps and strain curves, and for outputting early warnings and data tracing based on a threshold step size strategy.

[0051] Example 2

[0052] One embodiment of this application provides an automated detection method for the forming dimensions of stamped automotive seat hardware parts. See details below. Figure 1 The method includes the following steps:

[0053] Step 1: Collect the 3D point cloud dataset and strain data sequence of the workpiece at each data acquisition moment during the stamping process of automotive seat hardware; and generate the standard point cloud dataset of the workpiece based on the original CAD model.

[0054] A piezoelectric sensor is embedded at the corner where the mold seat side plate connects to the horizontal slide rail to continuously collect the workpiece strain data, which is used to characterize the stress changes at key stress concentration points of the material; a three-dimensional line laser scanner is used to continuously acquire the dynamic point cloud of the workpiece to express the surface geometry of the workpiece at each acquisition moment.

[0055] The original CAD digital model of the automotive seat hardware stamping part is imported into the 3D point cloud generation software. After surface discretization and mesh optimization, a high-density point coordinate set is output, which is denoted as the standard point cloud dataset of the workpiece.

[0056] The data acquisition frequency of the piezoelectric sensor and the scanning frequency of the 3D line laser scanner are both set to 20Hz. For each data acquisition moment, the strain data of all acquisition moments before each data acquisition moment are sorted in ascending order of time to obtain the strain data sequence of each data acquisition moment. At the same time, the information of all points in the 3D point cloud data acquired at each data acquisition moment is recorded as the 3D point cloud dataset of each data acquisition moment.

[0057] Thus, the strain data sequence and three-dimensional point cloud dataset of the workpiece at each data acquisition moment during the stamping process of automotive seat hardware are obtained, and a standard point cloud dataset of the workpiece is generated based on the original CAD model.

[0058] Step 2: Utilize the dynamic mode decomposition characteristics of the strain data sequence at each data acquisition moment and the abnormal deviation characteristics between the 3D point cloud dataset and the standard point cloud dataset to determine the time-varying rebound risk at each data acquisition moment.

[0059] Based on the above analysis, the nonlinear release of residual stress after demolding will cause continuous deformation and springback of the workpiece assembly size. This will make it difficult for the transmission-based detection method to capture the dynamic springback characteristics of the workpiece, thus causing deviations in the detection results of the workpiece size.

[0060] Therefore, using the strain data sequence at each data acquisition moment as input, the dynamic mode decomposition algorithm is used, with a truncation order of 5, to output the dominant mode decay coefficient. The coefficient of elasticity represents the rate of energy decay of the rebound. The larger the value, the faster the residual stress is released and the faster the rebound stabilizes. The method of calculating the decay coefficient of the dominant mode using the dynamic mode decomposition algorithm is a well-known technique and will not be elaborated here.

[0061] Subsequently, the 3D point cloud dataset at the current acquisition time and the standard point cloud dataset of the workpiece are used as inputs. Each 3D point cloud data in the 3D point cloud dataset represents the point cloud data on the surface of the workpiece at that acquisition time, while the CAD model represents the ideal geometric shape. For each point in the 3D point cloud data at each data acquisition time, the 3D Euclidean distance relative to the nearest point in the CAD model is calculated. The 3D Euclidean distances of all points are combined into a deviation dataset, which characterizes the degree of deviation between the actual position and the design position of each point in the 3D point cloud data.

[0062] Furthermore, using the deviation dataset between the current 3D point cloud data and the standard point cloud data as input, the Local Outlier Detection (LOF) algorithm is used. With the number of neighborhood points k=20, the local outlier S of each element in the deviation dataset is output, reflecting the overall springback outlier degree of the workpiece. The larger the value, the more significantly the springback of a local area on the workpiece surface deviates from the overall trend. The LOF algorithm is a well-known technique and will not be described in detail here.

[0063] Based on the above analysis, the time-varying rebound risk A at each data acquisition moment is calculated using the following formula:

[0064]

[0065] in, The deviation data between the actual 3D point cloud of the workpiece and the standard point cloud at the current acquisition time is the mean of the local outlier factor of all elements in the dataset, also known as the abnormal deviation feature, which represents the spatial outlier of the workpiece's springback deformation and the degree of abnormality of the workpiece's overall geometric deviation. The dominant mode decay coefficient represents the dynamic stability of the dominant mode during residual stress release, characterizing the rate at which rebound energy decays or diverges over time, and e is a natural constant.

[0066] The time-varying springback risk level A dynamically quantifies the risk of overall springback of stamped parts. When the local springback of the workpiece is more outlier and the residual stress is released slowly, it indicates that the workpiece may have continuous time-varying deviations or even batch defects, thus realizing the abnormal risk weighted assessment under the constraint of decay rate.

[0067] Since the three-dimensional Euclidean distance between the three-dimensional point cloud dataset and the standard point cloud dataset at each data acquisition moment represents the deviation of the workpiece size at the selected acquisition moment from the standard size, the mean of all elements in the deviation dataset between the three-dimensional point cloud dataset and the standard point cloud dataset at each data acquisition moment is denoted as the first mean.

[0068] For each data acquisition moment, the time-varying rebound risk and the first mean calculated from all acquisition moments prior to each data acquisition moment are sorted in ascending order of time to obtain the rebound risk sequence and size deviation sequence for each data acquisition moment.

[0069] Step 3: Calculate the dynamic control upper limit for the springback risk sequence based on the exponentially weighted moving average control chart; perform multiple linear regression with the springback risk sequence and its first-order difference sequence as independent variables and the size deviation sequence as the dependent variable, output the regression coefficients of the two independent variables, and determine the springback sensitivity index of the workpiece deformation at each data acquisition time in combination with the dynamic control upper limit.

[0070] When automotive seat hardware stamping parts experience sudden changes in material properties (such as abnormal dimensional fluctuations), abnormal mold wear, or process jumps, the prior knowledge of the springback compensation model becomes invalid, leading to high scrap rates and production line downtime issues in traditional closed-loop compensation systems.

[0071] Therefore, using the rebound risk sequence at each data acquisition time as input, an exponentially weighted moving average control chart (EWMAChart) is used. To balance the sensitivity to recent fluctuations with the stability of historical data, a weighting factor of 0.2 is set, outputting a dynamic control upper limit (UCL). The UCL defines the upper limit of acceptable risk for the process under statistical control. A larger value indicates a looser control upper limit, lower sensitivity to process anomaly detection, and allows for a higher rebound risk. The exponentially weighted moving average control chart is a well-known technique and will not be elaborated further.

[0072] Subsequently, the rebound risk sequence is first-differenced and padded with zeros at the beginning to obtain a risk change rate sequence of the same length as the rebound risk sequence. Further, using the rebound risk sequence, risk change rate sequence, and size deviation sequence as inputs, and elements from the rebound risk sequence and risk change rate sequence as independent variables, and elements from the size deviation sequence as dependent variables, multiple linear regression is used. In this embodiment, recursive least squares is used to dynamically update the weights of the independent variables, outputting the regression coefficients of the two independent variables. and The regression coefficients quantify the influence of time-varying rebound risk and its rate of change on material size changes, i.e. the expected change in material size fluctuations, revealing the strength of the causal relationship between material properties and rebound dynamics.

[0073] Based on the above analysis, the springback sensitivity index B of the workpiece deformation at each acquisition moment is constructed, and the specific calculation formula is as follows:

[0074]

[0075] in, and The regression coefficients of the two independent variables output by the multiple linear regression are respectively the fitted regression coefficients of the time-varying rebound risk and its rate of change. They quantify the sensitivity of the material's size fluctuations to the current time-varying rebound risk and its rate of change regression coefficients. The product of the two represents the strength of the overall causal relationship between the material properties and the rebound dynamics. This represents the upper limit of dynamic control output from the exponentially weighted moving average control chart, where e is the natural constant. It is the hyperbolic tangent function.

[0076] The springback sensitivity index B quantifies the overall contribution of springback risk to abrupt changes in material dimensional properties, amplifies the coupling effect of springback risk and its rate of change on material dimensional properties, and inversely weights the degree of relaxation of the EWMA control upper limit, reflecting the degree of influence of workpiece springback on dimensional changes. The larger the B value, the more sensitive the current workpiece size is to small disturbances, and the greater the probability of dimensional deviation and scrap rate.

[0077] Step 4: Execute an adaptive threshold compensation strategy based on the relationship between the springback sensitivity index and the pre-set two-level thresholds to achieve automated detection of the workpiece.

[0078] To avoid the need for traditional methods to reconstruct models from massive amounts of data to address the failure of the prior model in the closed-loop compensation system caused by abrupt changes in material properties and springback, an adaptive compensation strategy needs to be designed based on the springback sensitivity index. The specific adaptive compensation strategy design is as follows:

[0079] Based on the calculated rebound sensitivity index, two threshold levels are preset. In this embodiment, the first threshold T1 is set to 0.4, and the second threshold T2 is set to 0.7. Wherein, 0... <T1<T2<1。

[0080] Furthermore, the relationship between B and the two threshold levels is determined: when B is less than or equal to the first threshold T1, the original operation is maintained; when B is greater than the first threshold T1 and less than or equal to the second threshold T2, real-time monitoring and early warning are initiated; when B is greater than the second threshold T2, the workpiece springback prediction model is activated to avoid downtime and waiting.

[0081] Specifically, when the springback sensitivity index is less than or equal to the first threshold, it means that the workpiece size attribute is less sensitive to workpiece springback. The system only needs to maintain the existing closed-loop compensation model and continue to perform the current detection without triggering an additional compensation mechanism.

[0082] When the springback sensitivity index is greater than the first threshold and less than or equal to the second threshold, it means that the workpiece's size attributes are more sensitive to the workpiece's springback. At this time, the change curve of the springback sensitivity index is highlighted on the human-machine interface, and an audible and visual warning is activated to remind the staff. Historical data changes are also recorded in the background. It should be noted that the system's compensation is not intervened at this time.

[0083] When the springback sensitivity index exceeds the second threshold, it indicates that the workpiece's dimensional properties are extremely sensitive to springback, and the system's prior model is at risk of failure and collapse. In this case, the Yoshida-Uemori model (YU model) is used to simulate springback, outputting the predicted springback amount for each point in three directions. The springback amount for each point is then superimposed with the coordinates of each point in the 3D point cloud dataset acquired at the current acquisition time, dynamically adjusting the 3D point cloud dataset. This ensures that subsequent detection uses the adjusted 3D point cloud dataset as a benchmark, eliminating the systematic error between the static detection benchmark and dynamic springback deformation, tracking workpiece springback deformation in real time, and resolving false detection problems caused by deformation. The Yoshida-Uemori model (YU model) is a well-known technology and will not be elaborated upon in this embodiment.

[0084] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not invented in this application.

[0085] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. An automated method for detecting the forming dimensions of stamped automotive seat hardware parts, characterized in that, The method includes the following steps: Collect 3D point cloud datasets and strain data sequences of the workpiece at each data acquisition moment during the stamping process of automotive seat hardware; and generate standard point cloud datasets of the workpiece based on the original CAD model. By utilizing the dynamic mode decomposition features of the strain data sequence at each data acquisition moment and the abnormal deviation features between the 3D point cloud dataset and the standard point cloud dataset, the time-varying rebound risk level at each data acquisition moment is determined. Combined with the average deviation features at each data acquisition moment, the rebound risk sequence and size deviation sequence at each data acquisition moment are obtained. The dynamic control upper limit is calculated based on the exponentially weighted moving average control chart for the springback risk sequence; multiple linear regression is performed with the springback risk sequence and its first-order difference sequence as independent variables and the size deviation sequence as the dependent variable, and the regression coefficients of the two independent variables are output. Combined with the dynamic control upper limit, the springback sensitivity index of the workpiece deformation at each data acquisition time is determined. An adaptive threshold compensation strategy is executed based on the relationship between the springback sensitivity index and the preset two-level thresholds to achieve automated detection of the workpiece. The method for implementing an adaptive threshold compensation strategy based on the relationship between the rebound sensitivity index and a pre-set two-level threshold is as follows: The two-level threshold includes a first threshold T1 and a second threshold T2, where 0 <T1<T2<1; When the rebound sensitivity index is less than or equal to the first threshold, maintain the original operation; Real-time monitoring and early warning are activated when the rebound sensitivity index is greater than the first threshold and less than or equal to the second threshold. When the rebound sensitivity index is greater than the second threshold, the Yoshida-Uemori rebound prediction model is triggered to dynamically adjust the 3D point cloud dataset. The Yoshida-Uemori rebound prediction model outputs the predicted rebound amount of each measurement point in the X, Y, and Z directions, and superimposes the predicted rebound amount onto the coordinates of each point in the 3D point cloud dataset.

2. The automated detection method for the forming dimensions of automotive seat hardware stamping parts as described in claim 1, characterized in that, The strain data sequence at each data acquisition moment is composed of strain data from all acquisition moments prior to each data acquisition moment, sorted in ascending chronological order.

3. The automated detection method for the forming dimensions of automotive seat hardware stamping parts as described in claim 1, characterized in that, The method for determining the time-varying rebound risk at each data acquisition moment is as follows: The dominant mode decay coefficient of the strain data sequence at each data acquisition moment is obtained by using a dynamic mode decomposition algorithm; Obtain the abnormal deviation characteristics between the 3D point cloud dataset and the standard point cloud dataset at each data acquisition time; Calculate the exponent of an exponential function with the negative of the dominant mode decay coefficient as the base of the natural constant, and multiply the result of the exponential function calculation with the abnormal deviation characteristic as the time-varying rebound risk degree at each data acquisition moment.

4. The automated detection method for the forming dimensions of automotive seat hardware stamping parts as described in claim 3, characterized in that, The truncation order of the dynamic mode decomposition is 5.

5. The automated detection method for the forming dimensions of automotive seat hardware stamping parts as described in claim 3, characterized in that, The method for obtaining the abnormal deviation characteristics is as follows: For any point in the 3D point cloud dataset at each data acquisition time, the minimum 3D Euclidean distance between that point and a point in the standard point cloud dataset is used as an element in the deviation dataset. The local outlier detection algorithm is used to obtain the local outlier factor of each element in the biased dataset; The mean of the local outlier factors of all elements in the deviation dataset is used as the abnormal deviation feature.

6. The automated detection method for the forming dimensions of automotive seat hardware stamping parts as described in claim 5, characterized in that, The methods for obtaining the rebound risk sequence and size deviation sequence at each data acquisition time include: The mean of all elements in the deviation dataset between the 3D point cloud dataset and the standard point cloud dataset at each data acquisition time is denoted as the first mean. For each data acquisition moment, the time-varying rebound risk and the first mean calculated from all acquisition moments prior to each data acquisition moment are sorted in ascending order of time to obtain the rebound risk sequence and size deviation sequence for each data acquisition moment.

7. The automated detection method for the forming dimensions of automotive seat hardware stamping parts as described in claim 4, characterized in that, The formula for calculating the springback sensitivity index of the workpiece deformation at each data acquisition moment is as follows: Where B is the springback sensitivity index of the workpiece deformation at each acquisition time. and These are the regression coefficients of the two independent variables output by the multiple linear regression. This represents the upper limit of dynamic control output from the exponentially weighted moving average control chart, where e is the natural constant. It is the hyperbolic tangent function.

8. An automated inspection system for the forming dimensions of stamped automotive seat hardware parts, characterized in that, The system, which implements an automated inspection method for the forming dimensions of automotive seat hardware stamping parts as described in any one of claims 1-7, comprises: The high-precision 3D scanning unit is a 3D line laser scanner installed at the end of the robotic arm; The embedded stress monitoring unit is a piezoelectric sensor pre-embedded at the corner where the mold seat side plate connects to the horizontal slide rail. A high-speed data acquisition unit is used to simultaneously acquire the output signals of the high-precision three-dimensional scanning unit and the embedded stress monitoring unit; The central analysis unit is used to execute all the computational steps of the method according to any one of claims 1-7; It also includes a human-computer interaction interface for displaying point cloud deviation maps and strain curves in real time, and outputting early warnings and data tracing based on the threshold compensation strategy.

Images