Solder point quality analysis method, device, terminal equipment and medium

Abstract

The present invention discloses a solder joint quality analysis method, device, terminal equipment and medium, wherein the method comprises: obtaining solder joint data to be analyzed, wherein the solder joint data to be analyzed comprises solder joint data of the solder joint to be analyzed; calling a solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtaining an analysis quality result of the solder joint to be analyzed; wherein the solder joint quality scoring card model is obtained by pre-modeling and training the historical solder joint data using an outlier quantile algorithm, and the analysis quality result is used to indicate the quality of the welding of the solder joint to be analyzed. The present invention can solve the technical problems in the prior art such as the inability to quantitatively analyze solder joint quality, high analysis cost or low analysis efficiency.

Description

Technical Field

[0001] The present invention relates to the field of terminal technology, and in particular to a solder joint quality analysis method, device, terminal equipment and medium. Background Art

[0002] Welding is a common process in modern machinery manufacturing and is widely used in automobile manufacturing. The welding process uses a welding gun as a tool to connect two or more metal materials of the same or different types into one through the combination and diffusion of atoms or molecules in the form of welds. The quality of welds will have a great impact on the overall quality of the car. Common weld quality problems include: too much current in thin plate welding, easy to bite and weld through, gas shielded welding mainly has porosity, incomplete welding, spatter during welding, etc.

[0003] At present, in the field of industrial manufacturing, due to the lack of collection and storage of enough solder joint data in the past and insufficient knowledge of artificial intelligence (AI), the analysis and research of solder joint quality in the existing technology is more of a qualitative study, not a quantitative study. Or, complex physical devices are added to observe the welding process in real time, and the current, voltage, resistance and other information during welding are transmitted to the computer for analysis and research through the CAN card, and the spatter images of the solder joints are taken for analysis and research, etc. This will lead to an increase in research costs and low research and analysis efficiency. Summary of the invention

[0004] The embodiments of the present application provide a solder joint quality analysis method, thereby solving the technical problems in the prior art such as the inability to quantitatively analyze solder joint quality, high analysis cost or low analysis efficiency.

[0005] On the one hand, the present application provides a solder joint quality analysis method through an embodiment of the present application, the method comprising:

[0006] Acquire the solder joint data to be analyzed, wherein the solder joint data to be analyzed includes solder joint data of the solder joint to be analyzed;

[0007] Calling a solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtaining analysis quality results of the solder joints to be analyzed;

[0008] The solder joint quality scorecard model is obtained by pre-building and training the historical solder joint data using an outlier quantile algorithm, and the analysis quality result is used to indicate the quality of the solder joint to be analyzed.

[0009] Optionally, before calling the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, the method further includes:

[0010] Acquire historical solder joint data, wherein the historical solder joint data includes solder joint data of at least one sample solder joint, and the historical solder joint data is stored in a storage structure associated with a data index file and a solder joint data file;

[0011] Performing data feature engineering processing on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data, wherein the data feature engineering processing at least includes cross-features and derivative calculations;

[0012] The outlier quantile algorithm is used to mark the abnormal quality of solder joints on the characteristic calculation variables, and the solder joint quality scorecard model is constructed based on the marked characteristic calculation variables.

[0013] Optionally, the data feature engineering processing further includes timestamp processing and data segmentation processing, and the data feature engineering processing is performed on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data, including:

[0014] Performing time stamp processing on the historical solder joint data to obtain characteristic calculation variables of a portion corresponding to the historical solder joint data;

[0015] Performing data segmentation processing on the historical welding point data to obtain a plurality of segmentation moments corresponding to the historical welding point data;

[0016] According to the multiple segmented moments, cross-feature and derivative calculations are performed on the power characterization data in the historical weld data to obtain feature calculation variables corresponding to each of the multiple segmented moments; wherein the power characterization parameters include at least one of the following: weld current, weld voltage, weld resistance and weld power.

[0017] Optionally, the acquiring historical welding point data includes:

[0018] Acquire initial welding point data, wherein the initial welding point data includes initial data of at least one sample welding point;

[0019] The initial solder joint data is subjected to data preprocessing to obtain the historical solder joint data; wherein the data preprocessing includes at least one of the following: deleting solder joint data whose time is not within a preset time range, deleting solder joint data whose preset power characterization parameter is zero in a preset middle position, filling the empty values ​​in the solder joint data with the previous row of solder joint data, deleting multiple rows of zero data values ​​except the last row of zero data values ​​in the initial solder joint data, and adding a sampling time corresponding to the solder joint data and filtering processing to the solder joint data.

[0020] Optionally, the obtaining of initial welding point data includes:

[0021] Reading the initial solder joint data once in a preset block size manner;

[0022] The initial solder joint data are explored using preset software basic statements to obtain the initial solder joint data including basic information, wherein the basic information includes basic data of the initial solder joint data, detected missing values ​​and abnormal values.

[0023] Optionally, constructing the solder joint quality scorecard model based on the marked feature calculation variables includes:

[0024] Calculate variables according to the marked features, and count solder joint quality score information, wherein the solder joint quality score information includes statistically marked normal solder joints, abnormal solder joints, and probability of abnormal solder joint quality;

[0025] Build an initial scorecard model;

[0026] The initial scoring card model is trained using the solder joint quality scoring information to obtain the solder joint quality scoring card model.

[0027] Optionally, the method further comprises:

[0028] The solder joint quality score card model is evaluated by using the model evaluation index AUC method;

[0029] If the model evaluation does not meet the standards, the model parameters in the solder joint quality scorecard model are updated.

[0030] On the other hand, the present application provides a solder joint quality analysis device through an embodiment of the present application, the device comprising an acquisition module and an analysis module, wherein:

[0031] The acquisition module is used to acquire the solder joint data to be analyzed, wherein the solder joint data to be analyzed includes solder joint data of the solder joint to be analyzed;

[0032] The analysis module is used to call the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtain the analysis quality results of the solder joints to be analyzed;

[0033] The solder joint quality scorecard model is obtained by pre-building and training the historical solder joint data using an outlier quantile algorithm, and the analysis quality result is used to indicate the quality of the solder joint to be analyzed.

[0034] For the contents not introduced or elaborated in the embodiments of the present application, please refer to the relevant introduction in the aforementioned method embodiments, which will not be repeated here.

[0035] On the other hand, the present application provides a terminal device through an embodiment of the present application, and the terminal device includes: a processor, a memory, a communication interface and a bus; the processor, the memory and the communication interface are connected through the bus and communicate with each other; the memory stores executable program code; the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to execute the solder joint quality analysis method as described above.

[0036] On the other hand, the present application provides a computer-readable storage medium through an embodiment of the present application, wherein the computer-readable storage medium stores a program, and when the program is run on a terminal device, the solder joint quality analysis method as described above is executed.

[0037] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: the present application obtains the solder joint data to be analyzed, calls the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtains the analysis quality results of the solder joints to be analyzed. In this way, the solder joint quality can be automatically and quickly analyzed based on the solder joint quality scoring card model, thereby achieving the purpose of increasing efficiency, saving energy, and reducing costs. At the same time, it also solves the technical problems in the prior art such as the inability to quantitatively analyze the solder joint quality, high analysis cost, or low analysis efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 It is a flow chart of a solder joint quality analysis method provided in an embodiment of the present application.

[0040] Figure 2 It is a flow chart of another solder joint quality analysis method provided in an embodiment of the present application.

[0041] Figure 3 It is a schematic diagram of a storage structure of welding data provided in an embodiment of the present application.

[0042] Figure 4 It is a schematic diagram of a solder joint resistance-time curve provided in an embodiment of the present application.

[0043] Figure 5 It is a schematic diagram of multiple segments on a solder joint resistance-time curve provided in an embodiment of the present application.

[0044] Figure 6 It is a schematic diagram for determining the splash point of a weld curve provided in an embodiment of the present application.

[0045] Figure 7 It is a schematic diagram of a working characteristic ROC curve provided in an embodiment of the present application.

[0046] Figure 8 It is a structural schematic diagram of a solder joint quality analysis device provided in an embodiment of the present application.

[0047] Fig. 9 It is a structural diagram of a terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] The embodiments of the present application provide a solder joint quality analysis method, thereby solving the technical problems existing in the prior art such as the inability to quantitatively analyze solder joint quality, high analysis cost or low analysis efficiency.

[0049] The technical solution of the embodiment of the present application is to solve the above technical problems, and the overall idea is as follows: obtaining solder joint data to be analyzed, wherein the solder joint data to be analyzed includes solder joint data of the solder joint to be analyzed; calling a solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtaining analysis quality results of the solder joint to be analyzed; wherein the solder joint quality scoring card model is obtained by pre-modeling and training historical solder joint data using an outlier quantile algorithm, and the analysis quality results are used to indicate the quality of the welding quality of the solder joint to be analyzed.

[0050] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0051] First of all, the term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0052] In the process of filing this application, the applicant also discovered that the current automated manufacturing plants collect and store all process data of the welding process, providing valuable original data for solder joint quality research. This application uses artificial intelligence (AI) means, takes the collected welding process data as the research object, and establishes a solder joint quality scoring card model through mathematical and statistical methods, from which various factors and feature importance affecting the quality of solder joints are obtained, and targeted improvement plans are proposed for factors with higher importance, thereby improving the welding quality of solder joints.

[0053] The scoring card model is one of the core management technologies for banks, credit card companies, personal consumer credit companies, telecommunications companies, water and electricity service companies, insurance companies and other corporate entities. It is widely used in credit card life cycle management, auto loan management, housing loan management, personal loan management, other consumer loan management and other fields, and plays a very important role in marketing, credit approval, risk management, account management, customer relationship management and other aspects. However, its application in the field of industrial manufacturing is almost blank.

[0054] This application proposes a solder joint quality scoring card idea to provide an effective management tool for solder joint quality control. This application takes the welding process data of the welding robot as the research object, and uses artificial intelligence AI to preprocess the original welding process data, feature engineering, linear modeling, formulate scoring rules, etc., and finally establishes a solder joint quality scoring card model to provide effective control means for subsequent further analysis of solder joint quality and improvement of welding process.

[0055] See also Figure 1 , is a flow chart of a solder joint quality analysis method provided in an embodiment of the present application. Figure 1 The method shown includes the following implementation steps:

[0056] S101, obtaining solder joint data to be analyzed, wherein the solder joint data to be analyzed includes solder joint data of the solder joint to be analyzed.

[0057] The present application can obtain the weld data to be analyzed, wherein the weld data to be analyzed includes one or more weld data of the welds to be analyzed, wherein the weld data of each weld to be analyzed can be stored in an associated storage structure of a data index file and a weld master file. The weld data includes but is not limited to data information such as welding process, welding time, welding current, welding voltage, welding resistance and welding power.

[0058] S102, calling a solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtaining analysis quality results of the solder joints to be analyzed; wherein the solder joint quality scoring card model is obtained by pre-building and training a model on historical solder joint data using an outlier quantile algorithm, and the analysis quality results are used to indicate the quality of the welding quality of the solder joints to be analyzed.

[0059] The present application can call a pre-stored solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtain an analysis quality result for each solder joint to be analyzed. The analysis quality result is used to indicate the quality of the welding quality of the solder joint to be analyzed, for example, whether it is a normal quality solder joint or an abnormal quality solder joint, etc.

[0060] By implementing the present application, the present application obtains the solder joint data to be analyzed, calls the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtains the analysis quality results of the solder joints to be analyzed. In this way, the solder joint quality can be automatically and quickly analyzed based on the solder joint quality scoring card model, thereby achieving the purpose of increasing efficiency, saving energy, and reducing costs. At the same time, it also solves the technical problems in the prior art such as the inability to quantitatively analyze solder joint quality, high analysis cost, or low analysis efficiency.

[0061] The following is an example of creating a solder joint quality scorecard model. Figure 2 , is a flow chart of another solder joint quality analysis method provided in an embodiment of the present application. Figure 2 The method shown includes the following implementation steps:

[0062] S201, obtaining initial welding point data.

[0063] The initial solder joint data described in this application is stored in the storage structure of the data index file and the solder joint master file. Figure 3 A schematic diagram showing a possible solder joint data storage structure is shown in FIG. Figure 3 Each solder point index file corresponds to a solder point folder, and the solder point folder includes at least one solder point master file.

[0064] It should be noted that the solder joint data that needs to be counted in this application is a large amount of welding process data of the welding robot that has been automatically collected in real time (for example, half a year, 500G, etc.). The sampling period of the solder joint data is a preset duration, for example, 1 millisecond (ms). Each solder joint data (also referred to as a solder joint file) includes approximately 400 solder joint records. To facilitate data sorting and storage, this application uses a data storage structure associated with a data index file and a solder joint master file for storage.

[0065] Due to the huge amount of data, this application can use a preset block size (python chunksize) to read the initial solder joint data into a data frame (DataFrame) in blocks during statistical analysis to avoid problems such as a long time spent reading data at one time.

[0066] S202: Perform data exploration on the initial solder joint data to obtain basic information of the initial solder joint data.

[0067] This application can use the basic statements of the preset software to perform data exploration (exploratory data analysis, EDA) on the acquired initial solder joint data to obtain the basic information of the initial solder joint data, and input the basic data of the initial solder joint data, the detected missing values ​​and abnormal values ​​and other data information. Among them, the basic statements of the preset software (such as Python software) used for data exploration are shown in Table 1 below:

[0068] Table 1

[0069]

[0070] S203, preprocessing the initial welding point data to obtain historical welding point data. The historical welding point data includes welding point data of at least one sample welding point.

[0071] In the process of processing industrial data by big data statistical analysis methods, based on the results of big data statistical analysis, the value of industrial manufacturing big data is deeply mined, and data preprocessing is the first step of data mining. The data preprocessing includes but is not limited to data cleaning, data filtering, etc.

[0072] In data cleaning, any one or more combinations of the following operations can be performed, and this application does not limit it. For example, based on prior knowledge, the solder joint data whose time is not within the preset time range is deleted, such as the solder joint data file whose deletion time exceeds the first preset time length (for example, greater than or equal to 400ms) or is shorter than the second preset time length (for example, less than or equal to 200ms), etc. Such data should be solder joint data of failed welding. Delete the solder joint data whose preset power characterization parameters (such as current, voltage, resistance or power parameters) in the preset middle position of the solder joint data file are all 0. Such data should be the welding gun power failure caused by an emergency during the welding process. Fill the empty values ​​in the solder joint data with the solder joint data value of the previous row, and default the current solder joint state to keep the solder joint state of the previous moment. Delete the data marked as 0 in the last multiple rows on the solder joint resistance curve, and retain the solder joint data of 0 in the first row, where the extra rows of 0 data values ​​are considered to be the end of welding and the completion of the welding gun action, but the instruction of the data transmission host computer is later than the instruction to complete the welding gun action. Or according to the collection time of the solder joint data, a corresponding time data column is added to the solder joint data (or solder joint data file) so that the sampling time of the solder joint data is transmitted in real time and added to the solder joint data.

[0073] In the filtering process, the present application found that there are many burrs on the curve by drawing the solder joint resistance-time curve, that is, there are many noises. The present application further performs filtering on the solder joint data, for example, using the sliding average filtering method to filter the solder joint resistance-time curve. Figure 4 A possible solder joint resistance-time curve diagram is shown in FIG. Figure 4 The right side is a partial enlarged schematic diagram of the section indicated by the black line frame on the left. Figure 4 In FIG. 1 , curve 1 represents the original solder joint resistance-time curve, and curve 2 represents the solder joint resistance-time curve after filtering.

[0074] S204: Perform data feature engineering processing on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data.

[0075] The feature engineering of this application is the process of converting data attributes into data features, and the attributes represent the data dimensions. When modeling data, if the original data attributes are mechanically learned, the potential trend of the data cannot be found well. Feature engineering can reduce the interference of noise, so as to better find the potential trend of the data. The feature engineering process includes but is not limited to any one or more of the following: timestamp processing, discrete variable processing, segmentation processing (also known as binning or partitioning processing), cross-feature and derivative calculations, feature scaling and other data feature engineering processes for feature extraction.

[0076] The following description of the feature engineering process including timestamp processing, segmentation processing, cross-feature processing and derivative calculation is taken as an example to describe the relevant contents in detail. Specifically:

[0077] In timestamp processing, timestamps are time multi-scales, which separate timestamps into multiple dimensions, such as timestamp classifications of year, month, day, hour, minute, second, quarter, working day, etc. This application can perform timestamp processing on historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data. Specifically, feature calculation variables (also called new feature variables) can be added to each solder joint data, such as feature variable fields such as year, month, day, weekday, hour, minute, second, and season, where weekday is any positive integer from 1 to 7.

[0078] In the segmentation processing, the present application can perform segmentation processing on the historical solder joint data to obtain multiple segmentation moments corresponding to the historical solder joint data. Among them, the segmentation processing can convert continuous variables into categories, and at the same time enable the algorithm to reduce the interference of noise and divide a certain range of values ​​into a certain number of partitions / blocks. Specifically:

[0079] This application can analyze the solder joint resistance-time curve corresponding to the historical solder joint data to find the relatively low and high points of the solder joint resistance that conform to the solder joint law, and divide the time into three sections (t1, t2, t3). Figure 5A schematic diagram of multiple segmented moments on a solder joint resistance-time curve is shown. Among them, 0-t1 is stage 1, indicating the stage of dense metal bonding; t1-t2 is stage 2, indicating the stage of metal heating and melting; t3- is stage 3, indicating that the weld nugget is formed and the welding is completed. Among them, t1, t2 and t3 are three segmented moments. Based on the three segmented moments t1, t2 and t3, this application can further find other extreme point moments. Specifically, Figure 5 In the equation, A is the first derivative of the resistance R with respect to time t, and the time corresponding point where the first derivative is less than 0 until it is equal to 0 is found. That is, A:

[0080] B is to find the first-order derivative of resistance R with respect to time t, and find the time corresponding point where the first-order derivative is greater than 0 until it is equal to 0. That is, B:

[0081] C is the time corresponding to the first current being 0 in the solder joint resistance-time curve.

[0082] D is the splash point. Find the second and third derivatives of the resistance R with respect to time t, and find the time corresponding to the second derivative being 0 and the absolute value of the third derivative being the maximum. That is, D:

[0083] Among cross-features and derivative calculations, cross-feature calculation is one of the most important methods in feature engineering. It combines two or more categorical attributes (such as current, voltage, resistance, and power) into a feature calculation variable. Mathematically, it is to cross-multiply all values ​​of the categorical features, such as cross-multiplying the values ​​of current and voltage. Derivative calculation refers to the newly added feature variables (also called feature calculation variables) obtained through mathematical calculations, such as sum, maximum, minimum, average, first-order derivative, second-order derivative, and integral.

[0084] The newly added feature variables obtained by cross-feature and derivative calculation in this application are specifically shown in Table 2 below:

[0085] Table 2

[0086]

[0087]

[0088]

[0089] In Table 2 above, t1 is the time corresponding to the minimum resistance within 4-10ms (the first-order derivative of resistance with respect to time is 0), t2 is the time corresponding to the first maximum resistance with an interval greater than 2ms after t1 (the first-order derivative of resistance with respect to time is 0), and t3 is the time when the current is 0 after t1 and t2. Ele_consump_T2 is the power integration in the t2 stage, obtained by the Simpson algorithm. Ele_consump_T3 is the power integration in the t3 stage, obtained by the Simpson algorithm. Ele_consump is the power integration in the t1, t2 and t3 stages, obtained by the Simpson algorithm.

[0090] S205, using an outlier quantile algorithm to develop a model for the characteristic calculation variables to construct a solder joint quality score card model.

[0091] This application uses an outlier quantile algorithm of feature calculation variables to mark solder joint quality anomalies, and constructs a solder joint quality scorecard model based on the feature calculation variables after anomaly marking. In the anomaly marking, quantile statistics are performed on the cross-features and the feature calculation variables obtained by derivative dispersion, and three quantile intervals (i.e., the upper and lower thresholds for anomaly judgment) are obtained. Further, according to the intervals, the feature calculation variables are marked for solder joint quality, that is, normal or abnormal judgment. For example, normal can be marked as 0, and abnormal can be marked as 1. The cumulative score (Score) of the abnormal marking of all feature calculation variables is further counted, and the cumulative probability P of the abnormal score is calculated. The abnormal cumulative probability score P is compared with the threshold value (P thrd ) for comparison. If P is less than or equal to P thrd , then the solder joint quality is normal and the sample solder joint is marked as a normal solder joint; otherwise, the solder joint quality is abnormal and the sample solder joint is marked as an abnormal solder joint. Among them, the quality of solder joints is a relative concept, and the probability threshold of solder joint quality abnormality P thrd The setting of is the key, which can be determined according to the strictness of the solder joint quality. From this, the solder joint quality scoring information can be obtained, which can also be called the scoring card model scoring table.

[0092] The constructed initial scoring card model (such as a logistic regression model) can be further used to develop and construct a solder joint quality scoring card model. Specifically, according to the solder joint quality scoring card model scoring table, the information data can be divided into a training set, a test set, and a prediction set, and then the initial scoring card model can be built. For example, the Python toolkit sklearn.linear_modelLogisticRegression() is called to train the initial scoring card model through the training set (solder joint quality scoring information), thereby obtaining the solder joint quality scoring card model.

[0093] The following is a detailed introduction to quantile statistics, solder joint marking, abnormal score accumulation, abnormal probability calculation, solder joint quality scoring information (scoring table) and initial scoring card model construction.

[0094] In the quantile statistics of feature calculation variables, quantiles are used to divide a batch of data by probability. First, the data (specifically, the feature calculation variables) can be sorted, and then the probability distribution range of a random variable (i.e., discrete quantile algorithm) is used to divide it into several equally divided numerical points, among which the commonly used quantiles are median (i.e., 2 quantiles, 50%), quartiles, percentiles, etc. Taking the quantile of ±2.5 quantile as an example, this application can mark each feature calculation variable as normal within the ±2.5 quantile range, otherwise it can be marked as abnormal.

[0095] In the abnormal marking, the present application can perform quantile statistics on the 47 feature calculation variables calculated in the above step S205, and each feature calculation variable is divided into three intervals: [0, 2.5), [2.5, 97.5] and (97.5, 100). Among them, [0, 2.5) and (97.5, 100] are abnormal distinctions. When the feature calculation variable is in this interval, the feature calculation variable can be marked as abnormal, for example, marked as 1, and it is considered to be an outlier data. [2.5, 97.5] is a normal interval. When the feature calculation variable is in this interval, it can be marked as normal, for example, 0.

[0096] It should be noted that there are 12 normal and abnormal marks of characteristic calculation variables in this application, which are used as references for subsequent model evaluation, because these characteristic calculation variables obtained by derivative calculation of historical solder joint data are obtained by given threshold judgment, such as the number of spatter points in the t2 and t3 stages. Only when the spatter reference threshold is given can it be determined whether other spatters are determined as spatters. For example, see Figure 6 A schematic diagram showing a possible spatter point on the solder joint resistance-time curve. Figure 6 In the figure, there are splash point 1 (inflect1) and splash point 2 (inflect2). The splash degree of splash point 1 is higher than that of splash point 2. Whether splash point 2 is a splash point is determined based on a pre-given splash reference threshold. Therefore, the characteristic calculation variables derived from the statistics of the number of splash points can only be used as a reference for the solder joint quality scorecard model. See Table 3 below for details:

[0097] Table 3

[0098] Serial number Added new feature variables (feature calculation variables) Variable Description 17 expt_num_T2 Number of inflection points in stage t2 18 inflect_num_T2 Number of splash points in stage t2 19 Grad_sum_T2 The gradient and 20 expt_num_T3 Number of inflection points in stage t3 21 inflect_num_T3 Number of splash points in stage t3 22 Grad_sum_T3 t3 stage gradient and 39 expt_num_T2_P Number of power inflection points in stage t2 40 inflect_num_T2_P Number of power splash points in stage t2 41 Grad_sum_T2_P The power gradient and 42 expt_num_T3_P Number of power inflection points in stage t3 43 inflect_num_T3_P Number of power splash points in stage t3 44 Grad_sum_T3_P The power gradient and

[0099] The 13 characteristic calculation variables shown in Table 3 above are used for model evaluation of the subsequent solder joint quality scorecard model.

[0100] In the abnormal score accumulation (Score), 35 (47-12) normal or abnormal marked feature calculation variables are accumulated for abnormal mark, and the abnormal mark accumulation score S is iterated. Then, based on the score, the probability of solder joint quality abnormality P = Score / 35 is calculated. If P ≤ P thrd (abnormal solder joint quality probability threshold), the solder joint quality is finally judged to be normal, y=0; otherwise, the solder joint quality is judged to be abnormal, y=1.

[0101] In the solder joint quality scoring card model statistics table (solder joint quality scoring information), the present application can save the solder joint data corresponding to the feature calculation variables marked as normal or abnormal, including basic solder joint information (such as solder joint ID, time scale field, solder joint file path, etc.), final solder joint abnormality probability and other information, as a solder joint quality scoring model statistics table. Specifically, the welding robot of the present application collects n solder joint data in real time, calculates and processes r feature calculation variables corresponding to each solder joint data, and in this example, r=35; then, based on the 35 feature calculation variables, the corresponding result data, i.e., the solder joint quality abnormality probability, is calculated and obtained, and then based on this information, the corresponding solder joint quality scoring card model statistics table is constructed, as shown in Table 4 below:

[0102]

[0103] In the construction of the solder joint quality scorecard model, this application uses a Logistic regression model to calculate the probability of solder joint quality abnormality (y=1), referred to as the solder joint quality abnormality probability P, and the specific calculation is shown in the following formula (1):

[0104]

[0105] Among them, β0, β1, β r are all model parameters. β0 is the intercept term, a constant.

[0106] In practical applications, this application uses the statistical table of the solder joint quality score card model as input data, and divides the input data into a prediction set, a training set, and a test set according to a preset ratio (for example, 2:7:1). This application uses 70% of the data as a training set for training. 20% of the data is used for verification, and 10% of the data is used for Logistic regression model prediction.

[0107] In the specific implementation, this application can call the Python tool package sklearn.linear_model, import LogisticRegression; use the train_test_split() function to segment the input data; use logisticRegression() for training, and finally obtain the solder joint quality score card logistic regression model (i.e. solder joint quality score card model). Then, the predict_proba() function of the logistic regression model is used to predict the probability P of abnormal solder joint quality of the sample; finally, by referencing the binary classification sigmoid() function, it is determined whether the solder joint quality is normal or abnormal, that is, the solder joint quality analysis result is obtained.

[0108] S206: performing model evaluation on the solder joint quality score card model.

[0109] The model evaluation involved in this application is used to evaluate the distinguishing ability, predictive ability, and stability of the model, and to form a model evaluation report to draw a conclusion on whether the model can be used. Specifically, the present application can use logistic regression AUC to perform model evaluation on the solder joint quality scorecard model. Among them, AUC (Area Under Curve) is defined as the area under the ROC curve and the coordinate axis X, that is, the integral of the ROC curve over time, and the value will not be greater than 1. The ROC curve is generally above the straight line y=x, so the value range of AUC is between 0.5 and 1. The closer the AUC is to 1.0, the higher the authenticity of the detection method; when it is equal to 0.5, the authenticity is the lowest and has no application value, that is, the model is not available. The full name of the ROC curve is the receiver operating characteristic curve, which is a curve drawn based on a series of different binary classification methods (cutoff values ​​or decision thresholds), with the true positive rate (sensitivity) as the vertical coordinate and the false positive rate (1-specificity) as the horizontal coordinate. For example, please refer to Figure 7 A possible ROC curve schematic diagram is shown, and the figure shows three ROC curves by way of example.

[0110] In specific implementation, the present application can directly call the roc_auc_score() function to calculate the AUC value corresponding to the solder joint quality scoring card model. If the calculated AUC value is greater than the preset threshold, it is determined that the solder joint quality scoring card model has strong discrimination ability and prediction ability, and the solder joint quality scoring card model meets the standard; otherwise, it is determined that the discrimination ability and prediction energy of the solder joint quality scoring card model are weak, and the model does not meet the standard.

[0111] S207: When the model evaluation meets the standard, output the solder joint quality score card model.

[0112] When the model evaluates that the solder joint quality scoring card model meets the standards, the present application can directly output the solder joint quality scoring card model; otherwise, continue to use the solder joint quality scoring card model statistics table to update the model parameters in the solder joint quality scoring card model to output the solder joint quality scoring card model with better prediction ability.

[0113] According to the embodiment of the present application, a big data statistical method based on artificial intelligence is used as a means to perform data exploration, data preprocessing, and data engineering feature processing on solder joint data, and an outlier quantile algorithm is used to build a solder joint quality scoring card model for solder joint data. Through this model, direct analysis and evaluation of the solder joint quality of the solder joint data can be achieved, and a score can be obtained. At the same time, quality defects can be quickly found according to the solder joint quality evaluation table, and the main reasons affecting quality abnormalities can be found, thereby realizing predictive maintenance of quality inspection, process, etc., so as to achieve the purpose of increasing efficiency, saving energy and reducing costs.

[0114] Based on the same inventive concept, another embodiment of the present application provides an apparatus and a terminal device for implementing the solder joint quality analysis method described in the embodiment of the present application.

[0115] See also Figure 8 , is a schematic diagram of the structure of a solder joint quality analysis device provided in an embodiment of the present application. Figure 8 The device shown includes: an acquisition module 801 and an analysis module 802, wherein:

[0116] The acquisition module 801 is used to acquire the solder joint data to be analyzed, wherein the solder joint data to be analyzed includes the solder joint data of the solder joint to be analyzed;

[0117] The analysis module 802 is used to call the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtain the analysis quality results of the solder joints to be analyzed;

[0118] The solder joint quality scorecard model is obtained by pre-building and training the historical solder joint data using an outlier quantile algorithm, and the analysis quality result is used to indicate the quality of the solder joint to be analyzed.

[0119] Optionally, the device further includes a processing module 803 and a construction module 804, wherein:

[0120] The acquisition module 801 is further used to acquire historical solder joint data, wherein the historical solder joint data includes solder joint data of at least one sample solder joint, and the historical solder joint data is stored in a storage structure associated with a data index file and a solder joint data file;

[0121] The processing module 803 is used to perform data feature engineering processing on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data, wherein the data feature engineering processing at least includes cross-features and derivative calculations;

[0122] The construction module 804 is used to use an outlier quantile algorithm to mark the characteristic calculation variables as abnormal solder joint quality, and to construct the solder joint quality scorecard model based on the marked characteristic calculation variables.

[0123] Optionally, the data feature engineering process further includes timestamp processing and data segmentation processing, and the processing module 803 is specifically used for:

[0124] Performing time stamp processing on the historical solder joint data to obtain characteristic calculation variables of a portion corresponding to the historical solder joint data;

[0125] Performing data segmentation processing on the historical welding point data to obtain a plurality of segmentation moments corresponding to the historical welding point data;

[0126] According to the multiple segmented moments, cross-feature and derivative calculations are performed on the power characterization data in the historical weld data to obtain feature calculation variables corresponding to each of the multiple segmented moments; wherein the power characterization parameters include at least one of the following: weld current, weld voltage, weld resistance and weld power.

[0127] Optionally, the acquisition module 801 is specifically used for:

[0128] Acquire initial welding point data, wherein the initial welding point data includes initial data of at least one of the sample welding points;

[0129] The initial solder joint data is subjected to data preprocessing to obtain the historical solder joint data; wherein the data preprocessing includes at least one of the following: deleting solder joint data whose time is not within a preset time range, deleting solder joint data whose preset power characterization parameter is zero in a preset middle position, filling the empty values ​​in the initial solder joint data with the previous row of solder joint data, deleting multiple rows of data values ​​being zero and retaining the solder joint data with the last row of zero data values, adding a sampling time corresponding to the initial solder joint data to the initial solder joint data, and data filtering processing.

[0130] Optionally, the acquisition module 801 is specifically used for:

[0131] Reading the initial welding point data in blocks by using a preset block size;

[0132] The initial solder joint data are explored using preset software basic statements to obtain the initial solder joint data including basic information, wherein the basic information includes basic data of the initial solder joint data, detected missing values ​​and abnormal values.

[0133] Optionally, the construction module 804 is specifically used for:

[0134] Calculate variables according to the marked features, and obtain solder joint quality score information statistically, wherein the solder joint quality score information includes statistically marked normal solder joints, abnormal solder joints, and probability of abnormal solder joint quality;

[0135] Build an initial scorecard model;

[0136] The initial scoring card model is trained using the solder joint quality scoring information to obtain the solder joint quality scoring card model.

[0137] Optionally, the processing module 803 is further used to:

[0138] The solder joint quality score card model is evaluated by using the model evaluation index AUC method;

[0139] If the model evaluation does not meet the standards, the model parameters in the solder joint quality scorecard model are updated.

[0140] Please refer to 9, which is a schematic diagram of the structure of a terminal device provided in an embodiment of the present application. Fig. 9 The terminal device 90 shown includes: at least one processor 901, a communication interface 902, a user interface 903 and a memory 904. The processor 901, the communication interface 902, the user interface 903 and the memory 904 can be connected via a bus or other means. The embodiment of the present invention takes the connection via a bus 905 as an example.

[0141] The processor 901 may be a general-purpose processor, such as a central processing unit (CPU).

[0142] The communication interface 902 may be a wired interface (eg, an Ethernet interface) or a wireless interface (eg, a cellular network interface or a wireless local area network interface) for communicating with other terminals or websites. In the embodiment of the present invention, the communication interface 902 is specifically used to obtain welding point data.

[0143] The user interface 903 may be a touch panel, including a touch screen and a touch screen, for detecting operation instructions on the touch panel, or a physical button or a mouse. The user interface 903 may also be a display screen for outputting and displaying images or data.

[0144] The memory 904 may include a volatile memory, such as a random access memory (RAM); the memory may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD); the memory 904 may also include a combination of the above-mentioned types of memory. The memory 904 is used to store a set of program codes, and the processor 901 is used to call the program codes stored in the memory 904 to perform the following operations:

[0145] Acquire the solder joint data to be analyzed, wherein the solder joint data to be analyzed includes solder joint data of the solder joint to be analyzed;

[0146] Calling a solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtaining analysis quality results of the solder joints to be analyzed;

[0147] The solder joint quality scorecard model is obtained by pre-building and training the historical solder joint data using an outlier quantile algorithm, and the analysis quality result is used to indicate the quality of the solder joint to be analyzed.

[0148] Optionally, before calling the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, the processor 901 is further configured to:

[0149] Acquire historical solder joint data, wherein the historical solder joint data includes solder joint data of at least one sample solder joint, and the historical solder joint data is stored in a storage structure associated with a data index file and a solder joint data file;

[0150] Performing data feature engineering processing on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data, wherein the data feature engineering processing at least includes cross-features and derivative calculations;

[0151] The outlier quantile algorithm is used to mark the abnormal quality of solder joints on the characteristic calculation variables, and the solder joint quality scorecard model is constructed based on the marked characteristic calculation variables.

[0152] Optionally, the data feature engineering processing further includes timestamp processing and data segmentation processing, and the data feature engineering processing is performed on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data, including:

[0153] Performing time stamp processing on the historical solder joint data to obtain characteristic calculation variables of a portion corresponding to the historical solder joint data;

[0154] Performing data segmentation processing on the historical welding point data to obtain a plurality of segmentation moments corresponding to the historical welding point data;

[0155] According to the multiple segmented moments, cross-feature and derivative calculations are performed on the power characterization data in the historical weld data to obtain feature calculation variables corresponding to each of the multiple segmented moments; wherein the power characterization parameters include at least one of the following: weld current, weld voltage, weld resistance and weld power.

[0156] Optionally, the acquiring historical welding point data includes:

[0157] Acquire initial welding point data, wherein the initial welding point data includes initial data of at least one sample welding point;

[0158] The initial solder joint data is subjected to data preprocessing to obtain the historical solder joint data; wherein the data preprocessing includes at least one of the following: deleting solder joint data whose time is not within a preset time range, deleting solder joint data whose preset power characterization parameter is zero in a preset middle position, filling the empty values ​​in the solder joint data with the previous row of solder joint data, deleting multiple rows of data values ​​of zero and retaining the solder joint data of the last row of zero data values, and adding the sampling time corresponding to the solder joint data and filtering processing to the solder joint data.

[0159] Optionally, the obtaining of initial welding point data includes:

[0160] Reading the initial solder joint data once in a preset block size manner;

[0161] The initial solder joint data are explored using preset software basic statements to obtain the initial solder joint data including basic information, wherein the basic information includes basic data of the initial solder joint data, detected missing values ​​and abnormal values.

[0162] Optionally, constructing the solder joint quality scorecard model based on the marked feature calculation variables includes:

[0163] Calculate variables according to the marked features, and count solder joint quality score information, wherein the solder joint quality score information includes statistically marked normal solder joints, abnormal solder joints, and probability of abnormal solder joint quality;

[0164] Build an initial scorecard model;

[0165] The initial scoring card model is trained using the solder joint quality scoring information to obtain the solder joint quality scoring card model.

[0166] Optionally, the processor 901 is further configured to:

[0167] The solder joint quality score card model is evaluated by using the model evaluation index AUC method;

[0168] If the model evaluation does not meet the standards, the model parameters in the solder joint quality scorecard model are updated.

[0169] Since the terminal device introduced in this embodiment is a terminal device used to implement the solder joint quality analysis method in the embodiment of this application, based on the solder joint quality analysis method introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the terminal device of this embodiment and its various variations, so how the terminal device implements the method in the embodiment of this application is not described in detail here. As long as the terminal device used by the technical personnel of this field to implement the information processing method in the embodiment of this application belongs to the scope of protection of this application.

[0170] The technical solution in the above-mentioned embodiment of the present application has at least the following technical effects or advantages: the present application obtains the solder joint data to be analyzed, calls the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtains the analysis quality results of the solder joints to be analyzed. In this way, the solder joint quality can be automatically and quickly analyzed based on the solder joint quality scoring card model, thereby achieving the purpose of increasing efficiency, saving energy, and reducing costs. At the same time, it also solves the technical problems in the prior art such as the inability to quantitatively analyze the solder joint quality, high analysis cost, or low analysis efficiency.

[0171] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0172] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0173] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0174] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0175] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0176] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A solder joint quality analysis method, characterized in that: The method comprises: Acquire the welding point data to be analyzed, wherein the welding point data to be analyzed includes welding point data of the welding point to be analyzed; the welding point data includes: welding process, welding time, welding current, welding voltage, welding resistance and welding power; Calling a solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtaining analysis quality results of the solder joints to be analyzed; The solder joint quality scorecard model is obtained by pre-building and training the historical solder joint data using the outlier quantile algorithm, and the analysis quality result is used to indicate the quality of the solder joint to be analyzed; Before calling the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, the method further includes: Acquire historical solder joint data, wherein the historical solder joint data includes solder joint data of at least one sample solder joint, and the historical solder joint data is stored in a storage structure associated with a data index file and a solder joint data file; Performing data feature engineering processing on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data, wherein the data feature engineering processing at least includes cross-features and derivative calculations; Using an outlier quantile algorithm to mark the characteristic calculation variables as abnormal solder joint quality, and constructing the solder joint quality scorecard model based on the marked characteristic calculation variables; The constructing the solder joint quality scorecard model based on the marked feature calculation variables comprises: Calculate variables according to the marked features, and obtain solder joint quality score information statistically, wherein the solder joint quality score information includes statistically marked normal solder joints, abnormal solder joints, and probability of abnormal solder joint quality; Build an initial scorecard model; The initial scoring card model is trained using the solder joint quality scoring information to obtain the solder joint quality scoring card model.

2. The method according to claim 1, characterized in that The data feature engineering processing also includes timestamp processing and data segmentation processing. The data feature engineering processing is performed on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data, including: Performing time stamp processing on the historical solder joint data to obtain characteristic calculation variables of a portion corresponding to the historical solder joint data; Performing data segmentation processing on the historical welding point data to obtain a plurality of segmentation moments corresponding to the historical welding point data; According to the multiple segmented moments, cross-feature and derivative calculations are performed on the power characterization data in the historical weld point data to obtain feature calculation variables corresponding to each of the multiple segmented moments; wherein the power characterization data includes at least one of the following: weld point current, weld point voltage, weld point resistance and weld point power.

3. The method according to claim 1, characterized in that The obtaining of historical solder joint data comprises: Acquire initial welding point data, wherein the initial welding point data includes initial data of at least one of the sample welding points; Performing data preprocessing on the initial solder joint data to obtain the historical solder joint data; wherein the data preprocessing includes at least one of the following: Delete the welding point data whose time is not within the preset time range; Deleting the soldering point data with a preset power characterization parameter of zero in the preset middle position; Filling the empty values ​​in the initial solder joint data with the solder joint data of the previous row; Deleting zero data values ​​in other rows except the last row in the initial solder joint data; Adding a sampling time corresponding to the initial solder point data to the initial solder point data; Data filtering processing.

4. The method according to claim 3, characterized in that The obtaining of initial solder joint data comprises: Reading the initial welding point data in blocks by using a preset block size; The initial solder joint data are explored using preset software basic statements to obtain the initial solder joint data including basic information, wherein the basic information includes basic data of the initial solder joint data, detected missing values ​​and abnormal values.

5. The method according to claim 1, characterized in that The method further comprises: The solder joint quality score card model is evaluated by using the model evaluation index AUC method; the AUC is defined as the area under the ROC curve and the coordinate axis X; If the model evaluation does not meet the standards, the model parameters in the solder joint quality scorecard model are updated.

6. A solder joint quality analysis device, characterized in that: The device comprises an acquisition module and an analysis module, wherein: The acquisition module is used to acquire the welding point data to be analyzed, wherein the welding point data to be analyzed includes welding point data of the welding point to be analyzed; the welding point data includes: welding process, welding time, welding current, welding voltage, welding resistance and welding power; The analysis module is used to call the solder joint quality scoring card model to perform solder joint quality analysis on the solder joint data to be analyzed, and obtain the analysis quality results of the solder joints to be analyzed; The solder joint quality scorecard model is obtained by pre-building and training the historical solder joint data using the outlier quantile algorithm, and the analysis quality result is used to indicate the quality of the solder joint to be analyzed; The acquisition module is further used to acquire historical solder joint data, wherein the historical solder joint data includes solder joint data of at least one sample solder joint, and the historical solder joint data is stored in a storage structure associated with a data index file and a solder joint data file; The device also includes a processing module and a building module, The processing module is used to perform data feature engineering processing on the historical solder joint data to obtain feature calculation variables corresponding to the historical solder joint data, wherein the data feature engineering processing at least includes cross-features and derivative calculations; The construction module is used to use an outlier quantile algorithm to mark the characteristic calculation variables as abnormal solder joint quality, and to construct the solder joint quality scorecard model based on the marked characteristic calculation variables; The construction module is also used to: calculate variables according to the marked features, and obtain solder joint quality score information statistically, wherein the solder joint quality score information includes statistically marked normal solder joints, abnormal solder joints, and solder joint quality abnormality probability; Build an initial scorecard model; The initial scoring card model is trained using the solder joint quality scoring information to obtain the solder joint quality scoring card model.

7. A terminal device, characterized in that: include: Processors, memories, communication interfaces and buses; The processor, the memory and the communication interface are connected via the bus and communicate with each other; The memory stores executable program code; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to execute the solder joint quality analysis method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, and when the program is run on a terminal device, the solder joint quality analysis method as described in any one of claims 1 to 5 is executed.

Images