Method for establishing resistance spot welding quality monitoring system based on integrated algorithm

Abstract

The invention relates to a method for establishing a resistance spot welding quality monitoring system based on an integrated algorithm, and belongs to the field of quality monitoring. Welding parameters and ultrasonic detection information of welding spots in the actual welding production process are collected, a long-short-term memory network algorithm and a random forest algorithm are introduced to establish a basic learning device and an auxiliary learning device, cross validation and weighted output are completed through an integrated algorithm, and resistance spot welding quality monitoring is achieved. According to the method, process information in actual welding production is extracted to the maximum extent, and the influence caused by the difference between test piece welding and actual production welding parts is remarkably reduced. Therefore, the established model shows excellent adaptive capacity and effectiveness. And integrated learning and reinforcement learning technologies are introduced, so that the precision of the model is further improved, a prediction result is closer to a real value, the prediction capability of the model is enhanced, and a more reliable basis is provided for evaluation of the welding quality.

Description

Technical Field

[0001] The present invention relates to the field of quality monitoring, and in particular to a resistance spot welding quality monitoring method based on an integrated algorithm and collecting actual welding production process information to establish a quality assessment model, and more particularly to a method for establishing a resistance spot welding quality monitoring system based on an integrated algorithm. Background Art

[0002] Resistance spot welding technology, with its significant advantages such as high efficiency, low cost, easy operation and high degree of automation, occupies a pivotal position in the traditional manufacturing industry, especially in the manufacturing of automobiles and rail vehicles. However, in the welding process of rail vehicle bodies, many interfering factors such as electrode wear, assembly gap and working surface contamination are often present. These factors not only increase the instability of the welding process, but also may cause a series of welding quality problems such as cold welding and spattering, thus threatening the safety performance and service life of the vehicle body. In order to ensure the quality of the welds, real-time monitoring of the weld quality has become an indispensable measure. Through real-time monitoring, problems in the welding process can be discovered and solved in a timely manner, thereby ensuring the stability and reliability of the welding quality, and thus improving the overall performance and life of the vehicle body.

[0003] In today's industrial field, the monitoring technology of welding quality mainly relies on the collection of parameters of the test piece during welding and the comprehensive testing of the mechanical properties after welding. Through these data, a quality assessment model is constructed to monitor the welding quality. However, this technology faces significant problems in practical application. There are essential differences between test piece welding and actual body welding, which limits its effectiveness in actual production environments. In addition, the number of test pieces is limited, which further weakens the monitoring accuracy of the technology and its adaptability to different welding machines. Summary of the invention

[0004] The purpose of the present invention is to provide a method for establishing a resistance spot welding quality monitoring system based on an integrated algorithm, which solves the problems of low monitoring accuracy and poor adaptability in the prior art, and gets rid of the dependence on a large amount of welding test piece data. The present invention collects key parameters in the actual welding process in real time, such as current and voltage, and performs non-destructive monitoring after welding is completed, especially collecting ultrasonic monitoring signals. Based on these detailed actual welding process information, a quality assessment model is established, thereby constructing an accurate resistance spot welding quality monitoring system.

[0005] The above-mentioned object of the present invention is achieved by the following technical solutions: The method of establishing a resistance spot welding quality monitoring system based on an integrated algorithm uses the welding parameter information and ultrasonic detection information in the actual production process as quality assessment parameters, extracts dynamic resistance curves and ultrasonic detection images; constructs a basic learner and an auxiliary learner, and passes the output of the basic learner and the auxiliary learner as input to the integrated algorithm unit, and determines the appropriate weights in the reserved validation set or using cross-validation to balance the contribution of different models, and constructs a resistance spot welding quality assessment model. The model predicts and evaluates the quality of the weld, and determines whether the weld is a qualified weld by comparing it with the industry's requirements for weld quality, thereby realizing the quality monitoring of resistance spot welding.

[0006] Establish a quality assessment model. During the modeling process, input the dynamic resistance characteristic value, introduce the long short-term memory network algorithm to make multi-task regression prediction outputs for the weld nugget size and shear strength, and build a basic learner through multiple iterative training; input the ultrasonic detection weld nugget size characteristic value, introduce the random forest algorithm, make regression predictions for shear strength, and build an auxiliary learner through multiple iterative training. The outputs of the basic learner and the auxiliary learner are passed as inputs to the integrated algorithm unit, and the appropriate weights are determined in the reserved validation set or by cross-validation to balance the contributions of different models to complete the establishment of the quality assessment model.

[0007] The quality assessment model established based on the integrated algorithm has the function of reinforcement learning. The learning method is: the prediction results of the basic learner and the auxiliary learner are weighted averaged to output the tensile shear strength prediction result, and the output result and the characteristic value of the weld nugget size are transmitted back to the basic learner, the basic learner data is updated, and the basic learner is reinforced.

[0008] The method for obtaining the dynamic resistance characteristic value is as follows: the welding current and welding voltage in the actual welding production process are collected, the dynamic resistance characteristic value is extracted in chronological order after mean filtering and effective value calculation for each 1ms data, and the slope and mean value of the dynamic resistance curve are extracted in sequence from characteristic value 1 to characteristic value 9.

[0009] The integrated algorithm described in the present invention is:

[0010] in, r 真实 In order to verify the true value of concentrated shear strength, r 输出 is the final prediction output of the tensile shear strength after model integration. h t(1) , h t(2) They are the predicted values ​​of tensile shear strength and nugget size of the basic learner, y (n) , l(n) are the input value of the nugget size and the predicted value of tensile shear strength of the auxiliary learner, respectively. is the integration parameter, and λ is the integration weight, which will be continuously modified according to each new data input.

[0011] The method for updating the basic learner is: .

[0012] The beneficial effects of the present invention are: 1. The present invention not only extracts the process information in actual welding production to the maximum extent, but also significantly reduces the impact caused by the difference between the test piece welding and the actual production welded parts. Therefore, the model established by the method of the present invention shows excellent adaptability and effectiveness.

[0013] 2. The model breaks through the limitation of the number of test pieces. By collecting the signal characteristics in actual welding production in real time, it theoretically has almost unlimited learning samples. This innovative method greatly enriches the data basis of the model and provides it with a broader learning space.

[0014] 3. By introducing ensemble learning and reinforcement learning technology, the accuracy of the model is further improved, making the prediction results closer to the true value. This integration of technology not only enhances the prediction ability of the model, but also provides a more reliable basis for the evaluation of welding quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide further understanding of the present invention and constitute a part of this application. The illustrative examples of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0016] Figure 1 A flow chart of establishing a monitoring system of the present invention; Figure 2 It is a dynamic resistance characteristic value extraction diagram of the present invention; Figure 3 It is a characteristic value extraction diagram of the size of the weld nugget monitored by ultrasonic wave according to the present invention; Figure 4 It is a long short-term memory network algorithm diagram of the present invention; Figure 5 It is the random forest algorithm diagram of the present invention; Figure 6 It is the prediction effect diagram of the present invention. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0018] See also Figures 1 to 6 As shown, the method of establishing a resistance spot welding quality monitoring system based on an integrated algorithm of the present invention collects welding parameters in the actual welding production process and ultrasonic detection information of the welding point, introduces a long short-term memory network algorithm and a random forest algorithm to establish a basic learner and an auxiliary learner, and completes cross-validation and weighted output through an integrated algorithm to achieve resistance spot welding quality monitoring. The monitoring system includes six parts: an input unit, a data processing unit, a basic learning period, an auxiliary learner, an integrated algorithm unit, and an output unit.

[0019] Input unit: collects welding parameters during the actual welding process, uses twisted pair wires to collect welding voltage, uses current sensors to collect welding current, and inputs welding current and voltage data; performs ultrasonic testing on welding points and inputs ultrasonic testing information.

[0020] Data processing unit: obtain the dynamic resistance effective value curve through filtering, effective value calculation, etc.; extract the dynamic resistance characteristic value, extract the dynamic resistance characteristic value in time sequence, and extract the slope and mean of the dynamic resistance curve from characteristic value 1 to characteristic value 9 in turn; process the ultrasonic detection information, extract the weld nugget size characteristics, extract the weld nugget diameter, weld nugget area, and weld nugget circumference characteristic values, take the average value of the distance between the two farthest points of a line in the four directions in the image as the weld nugget diameter, and the weld nugget circumference and area are calculated according to the circumference and area formula of the circle to extract the characteristic values.

[0021] Basic learning period: Introduce the long short-term memory network algorithm, input the dynamic resistance characteristic value, make multi-task regression prediction output for the weld nugget size and tensile shear strength, and build a basic learner through multiple iterative training. The long short-term memory network algorithm consists of three units: forget gate, input gate and output gate. These units work together to ensure that the network can effectively process and learn complex sequence data.

[0022] Forget gate: used to keep the information in the memory unit discarded or retained

[0023] in f t is the output of the forget gate, xt is the input feature data, h t-1 is the output of the previous unit, W if , W hf As parameters, b if , b hf is the weight; Input gate: determines the information to update the memory unit, including sigmoid and tanh functions, both of which include the current input x t Hidden state at the previous moment h t-1 :

[0024] in i t and g t They are the outputs of the sigmoid and tanh functions, c t is the new state output, W ii , W hi , W ig , W hg As parameters, b ii , b hi , b ig , b hg is the weight; Output gate: Its function is to read the neural network state that has just been updated and output the memory unit:

[0025] in, o t is the output gate output, h t is the predicted output of the unit and the oldest state is passed to the next unit. If the unit is the final unit, then h t is the final prediction output, W io , W ho As parameters, b io , bho is the weight.

[0026] Finally, the predicted value of tensile shear strength is obtained h t(1) And the size of the weld nugget h t(2) . Auxiliary learner: The random forest algorithm is introduced, and the characteristic value of the weld nugget size is input to make a regression prediction of the tensile shear strength. Through multiple iterative training, an auxiliary learner is constructed. y n Input for the nugget size characteristic value, l n Output for tensile shear strength prediction.

[0027] Integrated algorithm unit: The output of the basic learning period and the auxiliary learner is passed into the integrated algorithm unit as input, and the appropriate weights are determined in the reserved validation set or by cross-validation to balance the contributions of different models. The prediction results of the long short-term memory network algorithm and the random forest algorithm are weighted averaged to output the tensile shear strength prediction results. The integrated algorithm is as follows:

[0028] in, r 真实 In order to verify the true value of concentrated shear strength, r 输出 is the final prediction output of the tensile shear strength after model integration. h t(1) , h t(2) They are the predicted values ​​of tensile shear strength and nugget size of the basic learner, y (n) , l (n) are the input value of the nugget size and the predicted value of tensile shear strength of the auxiliary learner, respectively. is the integration parameter, and λ is the integration weight, which will be continuously modified according to each new data input; At the same time, the output result and the characteristic value of the weld nugget size are transmitted back to the basic learner, the basic learner data is updated, and reinforcement learning is performed on the basic learner;

[0029] Output unit: The integrated tensile shear strength results are output to the model, and at the same time, the tensile shear strength regulations of the welds in the industry are compared to determine whether the welds are qualified welds, thereby completing real-time monitoring of the spot welding quality.

[0030] See also Figures 1 to 6As shown, the method of establishing a resistance spot welding quality monitoring system based on an integrated algorithm and collecting information from the actual welding production process of the present invention, the system collects the current and voltage signals in the actual welding process, and extracts the dynamic resistance characteristic value after filtering and effective value calculation processing. Subsequently, the long short-term memory network algorithm is introduced, and the basic learner is constructed using these characteristic values ​​to achieve multi-task regression prediction of shear strength and nugget size. At the same time, the system also collects ultrasonic detection information of the weld, extracts nugget size information, and uses the random forest algorithm to construct an auxiliary learner to perform regression prediction on shear strength. Finally, the training, enhancement and update of the model and the accurate prediction of weld quality are achieved through the integrated algorithm of multiple models.

[0031] Step 1: Data collection and processing: See also Figure 2 As shown in the figure, the system first collects the welding parameters in the actual welding process, uses twisted pair wires to collect welding voltage, and uses current sensors to collect welding current. The collected signals are filtered and the effective value is calculated. The sampling frequency of current and voltage is 10kHz, and the number of sampling points in every 1ms is 10. Therefore, the effective values ​​of resistance, voltage and current are calculated once every 1ms as follows:

[0032] Step 2: Perform ultrasonic testing or other non-destructive testing on the weld to extract the nugget size information and nugget size features, including the nugget diameter, circumference and area. Figure 3 As shown in the figure, the shape of the weld nugget is approximately circular. When measuring the diameter of the weld nugget, the average value of the distance between the two farthest points of a line in four directions in the image is taken as the diameter of the weld nugget, and the circumference and area of ​​the weld nugget are calculated according to the circumference and area formula of a circle to extract the characteristic value.

[0033] Step 3: Introduce the long short-term memory network algorithm. As a special recurrent neural network, the long short-term memory network algorithm shows excellent performance in processing sequence data, especially in capturing long-term dependencies in time series. The dynamic resistance curve of resistance spot welding, as a typical time series data, records the change of resistance value over time during the welding process. The characteristic of this data is that the data points follow a strict time order, and each data point is closely related to a specific time point or time period. Therefore, the characteristic value of dynamic resistance is extracted in time order, see Figure 2 , extract the mean and slope of the dynamic resistance curve from eigenvalue 1 to eigenvalue 9 in turn. Input the eigenvalues ​​into the long short-term memory network algorithm, and build a basic learner through multiple training iterations to achieve the prediction output of tensile shear strength and weld nugget size. Figure 4,The long short-term memory network algorithm consists of three units: the forget gate, the input gate, and the output gate. These units work together to ensure that the network can effectively process and learn complex sequence data.

[0034] Forget gate: used to discard or retain the information in the memory unit;

[0035] in f t is the output of the forget gate, x t is the input feature data, h t-1 is the output of the previous unit, W if , W hf As parameters, b if , b hf is the weight; Input gate: determines the information to update the memory unit, including sigmoid and tanh functions, both of which include the current input x t Hidden state at the previous moment h t-1 .

[0036]

[0037] in i t and g t They are the outputs of the sigmoid and tanh functions, c t is the new state output, W ii , W hi , W ig , W hg As parameters, b ii , b hi , b ig , b hg is the weight; Output gate: Its function is to read the neural network state that has just been updated and output the memory unit

[0038]

[0039] in, o t is the output gate output, h t is the predicted output of the unit and the oldest state is passed to the next unit. If the unit is the final unit, then h t is the final prediction output, W io , W ho As parameters, b io , b ho is the weight.

[0040] Finally, the predicted value of tensile shear strength is obtained h t(1) And the size of the weld nugget h t(2) .

[0041] Step 4: Introduce the random forest algorithm, input the eigenvalue of the nugget size to build an auxiliary learner, and make a regression prediction of the tensile shear strength; see Figure 5 , y n Input for the nugget size characteristic value, l n Output for tensile shear strength prediction.

[0042] Step 5: Use an ensemble learning algorithm to combine the basic learner with the auxiliary learner to achieve more accurate predictions. In this process, use the retained validation set or cross-validation technology to evaluate the performance of each model. Balance the contributions of different models by setting appropriate weights. Finally, output the prediction results of these models in a weighted average to achieve the best prediction effect. The ensemble algorithm is as follows:

[0043] in, r 真实 In order to verify the true value of concentrated shear strength, r 输出 is the final prediction output of the tensile shear strength after model integration. h t(1) , h t(2) They are the predicted values ​​of tensile shear strength and nugget size of the basic learner, y (n) , l (n) are the input value of the nugget size and the predicted value of tensile shear strength of the auxiliary learner, respectively. is the integration parameter, and λ is the integration weight, which will be continuously modified according to each new data input; Step 6: r 真实 and y (n) Input back to the basic learner to update the data and strengthen the basic learner;

[0044] Step 7. The quality assessment model is constructed through the above steps. In the actual welding production process, the welding parameters of the welding points are collected in real time. The welding current and voltage are filtered and the effective value is calculated through the data processing unit. The dynamic resistance characteristic value is input and the result is evaluated by the model. r 真实 Output, compare with the industry's regulations on the shear strength and nugget size of the weld, and determine whether the weld is a qualified weld. This completes real-time monitoring of the weld quality. Example

[0045] This embodiment provides a quality monitoring method based on actual welding process information to evaluate the quality of double-pulse DC welding of SUS301L, 2+0.6mm stainless steel plates. The spot welding process parameters are 5KA for the primary pulse current, 7KA, 7.5KA, 8KA, 8.5KA, 9KA for the secondary pulse current, and 6000N for the electrode pressure. Five groups of tests were conducted, with 8 welding spots in each group. The dynamic resistance characteristic value was extracted by collecting welding parameters, see Figure 2 As shown; collect ultrasonic detection information and extract the characteristic value of the weld nugget size, see Figure 3 shown.

[0046] The dynamic resistance characteristic value and the nugget size characteristic value are input into the model to construct the basic learning period and auxiliary learner. After verification by the integrated algorithm, the weighted tensile shear strength prediction value is output, and the prediction accuracy can reach 97.47%. Figure 6 shown.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made to the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for establishing a resistance spot welding quality monitoring system based on an integrated algorithm, characterized in that: The welding parameter information and ultrasonic detection information in the actual production process are used as quality assessment parameters to extract dynamic resistance curves and ultrasonic detection images; a basic learner and an auxiliary learner are constructed, and the outputs of the basic learner and the auxiliary learner are passed as inputs to the integrated algorithm unit. The contribution of different models is balanced in the reserved validation set or by using cross-validation to determine the weights, and a resistance spot welding quality assessment model is constructed; A quality assessment model was established. During the modeling process, the dynamic resistance characteristic value was input, and the long short-term memory network algorithm was introduced to make multi-task regression prediction outputs for the weld nugget size and tensile shear strength. Through multiple iterative training, a basic learner was constructed. The ultrasonic detection weld nugget size characteristic value was input, and the random forest algorithm was introduced to make regression predictions for tensile shear strength. Through multiple iterative training, an auxiliary learner was constructed. The outputs of the basic learner and the auxiliary learner are passed into the integrated algorithm unit as input, and the weights are determined in the reserved validation set or by cross-validation to balance the contributions of different models, thus completing the establishment of the quality assessment model.

2. The method for establishing a resistance spot welding quality monitoring system based on an integrated algorithm according to claim 1, characterized in that: The quality assessment model established based on the integrated algorithm has the function of reinforcement learning. The learning method is: the prediction results of the basic learner and the auxiliary learner are weighted averaged to output the tensile shear strength prediction result, and the output result and the characteristic value of the weld nugget size are transmitted back to the basic learner, the basic learner data is updated, and the basic learner is reinforced.

3. The method for establishing a resistance spot welding quality monitoring system based on an integrated algorithm according to claim 1, characterized in that: The method for obtaining the dynamic resistance characteristic value is as follows: the welding current and welding voltage in the actual welding production process are collected, the dynamic resistance characteristic value is extracted in chronological order after mean filtering and effective value calculation for each 1ms data, and the slope and mean value of the dynamic resistance curve are extracted in sequence from characteristic value 1 to characteristic value 9.

4. The method for establishing a resistance spot welding quality monitoring system based on an integrated algorithm according to claim 1, characterized in that: The integrated algorithm is: ； in, r 真实 In order to verify the true value of concentrated shear strength, r 输出 is the final prediction output of the tensile shear strength after model integration. h t(1) , h t(2) They are the predicted values ​​of tensile shear strength and nugget size of the basic learner, y (n) , l (n) are the input value of the nugget size and the predicted value of tensile shear strength of the auxiliary learner, respectively. is the integration parameter, and λ is the integration weight, which will be continuously modified according to each new data input.

5. The method for establishing a resistance spot welding quality monitoring system based on an integrated algorithm according to claim 2, characterized in that: The method for updating the basic learner is: 。

Images