Optimization method of motor housing processing parameters based on machine learning

Abstract

The present invention discloses a method for optimizing the process parameters of a motor housing based on machine learning, and relates to the field of motor housing processing. The technical solution of the present invention obtains the evaluation parameters of the process quality of the node to be optimized and the data values ​​of the influencing parameters, constructs an evaluation model of the process quality, calculates the quantitative evaluation value of the node process quality, performs a grey correlation analysis between the obtained quantitative evaluation value and the influencing parameters, obtains the correlation between the quantitative evaluation value of the node process quality and each influencing parameter, sets a correlation threshold to filter out the influencing parameters that have a strong correlation with the quantitative evaluation value, takes the obtained strongly correlated influencing parameter data as input, takes the quantitative evaluation value of the process quality of the node to be optimized as output for pre-training, constructs a neural network model, obtains the mapping relationship between the strongly correlated influencing parameters and the quantitative evaluation value of the node process quality, and determines the optimization parameters of each strongly correlated influencing parameter according to the obtained mapping relationship.

Description

Technical Field

[0001] The present invention relates to the field of motor casing processing, and in particular to a method for optimizing process parameters of motor casing processing based on machine learning. Background Art

[0002] Machine learning is a common research hotspot in the field of artificial intelligence and pattern recognition. Its theories and methods have been widely used to solve complex problems in engineering applications and scientific fields. With the introduction of the concept of intelligent manufacturing, machine learning technology has been increasingly applied to the mechanical manufacturing industry. Generally, the processing of motor housings requires multiple processes including blanking, punching, stretching, and forming. According to the simulation analysis of the motor housing processing process, the final process quality of the motor housing depends on the quality of each processing node, which is largely affected by the forming parameters. There are many factors that affect the process quality at the processing nodes, including friction coefficient, stamping speed, blank holding force, mold geometry parameters, etc.

[0003] The existing processing flow analyzes the factors affecting the process quality of the processing nodes, verifies through numerical simulation, and optimizes the processing parameters based on the analysis and verification results. Since the factors affecting the processing technology are diversified, and there is no clear linear mapping relationship between the factors and the process quality, the existing simulation verification method cannot establish a processing parameter optimization mechanism under the joint action of multiple parameters, and can often only perform adaptive optimization for a specific processing parameter. To this end, we propose a motor housing processing parameter optimization method based on machine learning. Summary of the invention

[0004] The main purpose of the present invention is to provide a method for optimizing the processing parameters of a motor casing based on machine learning, which can effectively solve the problems in the background technology.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] The motor housing processing parameter optimization method based on machine learning includes:

[0007] Obtain the processing process flow, and mark the processing nodes as node p1, node p2, ..., node p1 in the order of the process flow. n , get the node p to be optimized m Evaluation parameters of process quality u mi And the influencing parameter v mj The data value of mi Represented as node p m The i-th evaluation parameter of process quality, v mj Represented as node p mThe jth influencing parameter of process quality, m≤n; i=1,2,...; j=1,2,...;

[0008] Construct the node p to be optimized m The evaluation model of process quality is used to calculate the node p according to the model. m Quantitative evaluation value of process quality U m , the expression of the evaluation model is:

[0009]

[0010] Where U m is the node p to be optimized m Quantitative evaluation value of process quality; ω mi For node p m The weight of the i-th evaluation parameter of process quality, the weight of the evaluation parameter ω mi Satisfies the relationship: and is proportional to the node p m The absolute value of the correlation coefficient between the quantitative evaluation value of process quality and the evaluation parameter, node p m The calculation formula of the correlation coefficient between the quantitative evaluation value of process quality and the evaluation parameters is:

[0011]

[0012] In the formula, r mj Represented as node p m The correlation coefficient between the quantitative evaluation value of process quality and the jth evaluation parameter; U mt For node p m The tth data value of the quantitative evaluation value of process quality; For node p m The mean of the quantitative evaluation value of process quality; u mjt Represented as node p m The tth data value of the jth evaluation parameter of process quality; Represented as node p m The mean value of the jth evaluation parameter of process quality; T is the total amount of data;

[0013] The quantitative evaluation value and influencing parameter v obtained mj Perform grey correlation analysis between them to obtain node p m Quantitative evaluation value of process quality and various influencing parameters v mj The correlation between mj , the calculation formula is as follows:

[0014]

[0015] Among them, Δ min and Δmax They are the minimum difference between the two levels and the maximum difference between the two levels of the quantitative evaluation value and the standardized data of the influencing parameters, and the calculation formulas are:

[0016] Δ min =min(min(|X0(k)-X j (k)|));

[0017] Δ max =max(max(|X0(k)-X j (k)|));

[0018] X0(k) is the node p m The kth data of the standardized value of the quantitative evaluation value of the process quality; X j (k) is the node p m The kth data of the standardized value of the jth influencing parameter of the quantitative evaluation value of process quality; min(min(·)) means first taking the minimum value of all factors in (·) and then taking the minimum value of all minimum values; max(max(·)) means first taking the maximum value of all factors in (·) and then taking the maximum value of all maximum values; λ is the data volume; ρ is the constant coefficient, and the value range is (0,1). The value range of the correlation threshold is set to [0.7,1] to screen out the influencing parameters that have a strong correlation with the quantitative evaluation value;

[0019] The obtained strongly correlated influencing parameter data is used as input to optimize the node p m The quantitative evaluation value of the process quality is pre-trained for output, and a neural network model is constructed. The neural network model is an input layer-hidden layer-output layer structure, wherein the number of input layer nodes p, the number of intermediate layer nodes s, and the number of output layer nodes q satisfy the following relationship:

[0020]

[0021] Where a is an integer from 1 to 9. In this model, the number of output layer nodes q = 1; the number of input layer nodes p is equal to the number of influencing parameters that are strongly correlated with the quantitative evaluation value. According to the model, the strongly correlated influencing parameters and node p are obtained. m The mapping relationship between the quantitative evaluation values ​​of process quality is obtained, and the optimization parameters of various strongly correlated influencing parameters are determined based on the obtained mapping relationship.

[0022] The present invention has the following beneficial effects:

[0023] Compared with the prior art, the technical solution of the present invention obtains the processing flow, marks the processing nodes in sequence according to the sequence of the process flow, and obtains the node to be optimized p m Evaluation parameters of process quality u mi And the influencing parameter vmj The data value of the node to be optimized is constructed m The evaluation model of process quality is used to calculate the node p according to the model. m Quantitative evaluation value of process quality U m , the obtained quantitative evaluation value and influencing parameter v mj Perform grey correlation analysis between them to obtain node p m0 Quantitative evaluation value of process quality and various influencing parameters v mj The correlation between mj , set the correlation threshold to filter out the influencing parameters that have a strong correlation with the quantitative evaluation value, and use the obtained strongly correlated influencing parameter data as input to optimize the node p m The quantitative evaluation value of process quality is pre-trained as the output, and a neural network model is constructed. According to the model, the strongly correlated influencing parameters and node p are obtained. m The mapping relationship between the quantitative evaluation values ​​of process quality is obtained, and the optimization parameters of each strongly correlated influencing parameter are determined based on the obtained mapping relationship. By constructing a process quality evaluation model and a neural network model, a processing parameter optimization mechanism under the joint action of multiple influencing factors is established, which effectively solves the technical problem of only being able to adaptively optimize a specific processing parameter due to the diversification of processing technology influencing factors and the non-clear linear mapping relationship between influencing factors and process quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flow chart of the method for optimizing the process parameters of a motor housing processing based on machine learning of the present invention;

[0025] Figure 2 This is a schematic diagram of the decomposition of the motor housing processing process nodes;

[0026] Figure 3 Schematic diagram of the neural network model structure. DETAILED DESCRIPTION

[0027] The present invention will be further described below in conjunction with specific implementation methods, wherein the accompanying drawings are only used for exemplary descriptions and represent only schematic diagrams rather than actual drawings, and should not be understood as limiting the present invention. In order to better illustrate the specific implementation methods of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product.

[0028] Example 1

[0029] like Figure 1-3 As shown, the specific implementation process of the technical solution of the present invention includes the following steps:

[0030] Step 1: Obtain the processing process flow, and mark the processing nodes as node p1, node p2, ..., node p in the order of the process flow. n ,like Figure 2 As shown, get the node p to be optimized m Evaluation parameters of process quality u mi And the influencing parameter v mj The data value of mi Represented as node p m The i-th evaluation parameter of process quality, v mj Represented as node p m The jth influencing parameter of process quality, m≤n; i=1,2,...; j=1,2,...;

[0031] In this embodiment, the punching process in the motor housing processing process is taken as an example, assuming that the punching process is the node p to be optimized. m The evaluation parameters that affect the process quality of the punching process include but are not limited to the aperture accuracy u m1 、Position accuracy u m2 , roundness u m3 , surface finish u m4 The influencing parameters include but are not limited to the punching speed v m1 , Punching force v m2 and punch pressure v m3 wait;

[0032] Step 2: Construct the node p to be optimized m The evaluation model of process quality is used to calculate the node p according to the model. m Quantitative evaluation value of process quality U m , the expression of the evaluation model is:

[0033]

[0034] Where U m is the node p to be optimized m Quantitative evaluation value of process quality; ω mi For node p m The weight of the i-th evaluation parameter of process quality, the weight of the evaluation parameter ω mi Satisfies the relationship: and is proportional to the node p m The absolute value of the correlation coefficient between the quantitative evaluation value of process quality and the evaluation parameter, node p m The calculation formula of the correlation coefficient between the quantitative evaluation value of process quality and the evaluation parameters is:

[0035]

[0036] In the formula, rmj Represented as node p m The correlation coefficient between the quantitative evaluation value of process quality and the jth evaluation parameter; U mt For node p m The tth data value of the quantitative evaluation value of process quality; For node p m The mean of the quantitative evaluation value of process quality; u mjt Represented as node p m The tth data value of the jth evaluation parameter of process quality; Represented as node p m The mean value of the jth evaluation parameter of process quality; T is the total amount of data;

[0037] In this embodiment, we take the weights of various evaluation parameters as equal for calculation description, and there is ω mi =0.25, then the expression of the evaluation model is:

[0038] U m =0.25(u m1 +u m2 +u m3 +u m4 )

[0039] Substituting the obtained process quality evaluation parameters into the above formula, the node p can be solved m Quantitative evaluation value of process quality;

[0040] Step 3: Compare the obtained quantitative evaluation value and influencing parameter v mj The grey correlation analysis is carried out between them. The specific steps are as follows:

[0041] Take node p m The data sequence is constructed based on the quantitative evaluation value of process quality and the historical data of each influencing parameter, where the quantitative evaluation value is the parent sequence and the value of each influencing parameter is the child sequence. The data is sorted to obtain the following table:

[0042] Table 1 Data sequence table

[0043]

[0044] The data in the above table are standardized, and the standardization formula is:

[0045]

[0046] Where, X i (k) is the standardized data, and the node p is obtained by calculation according to the formula m Quantitative evaluation value of process quality and various influencing parameters v mj The correlation between mj, the calculation formula is as follows:

[0047]

[0048] Among them, Δ min and Δ max They are the minimum difference between the two levels and the maximum difference between the two levels of the quantitative evaluation value and the standardized data of the influencing parameters, and the calculation formulas are:

[0049] Δ min =min(min(|X0(k)-X j (k)|));

[0050] Δ max =max(max(|X0(k)-X j (k)|));

[0051] X0(k) is the node p m The kth data of the standardized value of the quantitative evaluation value of the process quality; X j (k) is the node p m The kth data of the standardized value of the jth influencing parameter of the quantitative evaluation value of process quality; min (min (·)) means first taking the minimum value of all factors in (·), and then taking the minimum value of all minimum values; max (max (·)) means first taking the maximum value of all factors in (·), and then taking the maximum value of all maximum values; λ is the data volume; ρ is a constant coefficient, and the value range is (0,1). After solving the correlation value, the influencing parameters with strong correlation with the quantitative evaluation value are screened out according to the set correlation threshold value range [0.7,1]. In this embodiment, the four influencing parameters listed are all influencing parameters with strong correlation;

[0052] Step 4: Take the obtained strong correlation influencing parameter data as input to optimize the node p m The quantitative evaluation value of process quality is pre-trained as the output to build a neural network model. It should be noted that the input data and output data of the neural network model need to be normalized. In order to train and test the established neural network prediction model, a part of the actual monitored data is selected as training samples and a part as test samples, and then the variables are normalized to the [0, 1] interval. The normalization formula is:

[0053] R=(x-Xmin) / (Xmax-Xmin)

[0054] Among them, R is the normalized data value; x is the unnormalized data value; Xmax and Xmin are the maximum and minimum values ​​in the data sample respectively. The neural network model is an input layer-hidden layer-output layer structure, in which the number of input layer nodes p, the number of intermediate layer nodes s and the number of output layer nodes q satisfy the following relationship:

[0055]

[0056] Where a is an integer from 1 to 9. In this model, the number of output layer nodes q = 1; the number of input layer nodes p is equal to the number of influencing parameters that have a strong correlation with the quantitative evaluation value, and p = 4. In this embodiment, a = 4 is taken, and the number of intermediate layer nodes is If we round up to the nearest integer, the number of nodes in the middle layer is s=5, and the neural network model is a structure of 4-node input layer-5-node hidden layer-1-node output layer, as shown in Figure 3 As shown;

[0057] Step 5: Obtain strong correlation influencing parameters and node p according to the model m The mapping relationship between the quantitative evaluation values ​​of process quality is obtained, and the optimization parameters of various strongly correlated influencing parameters are determined based on the obtained mapping relationship.

[0058] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for optimizing process parameters of motor housing processing based on machine learning, characterized in that: include: Obtain the processing process flow, and mark the processing nodes as node p1, node p2, ..., node p1 in the order of the process flow. n , get the node p to be optimized m Evaluation parameters of process quality u mi And the influencing parameter v mj The data value of mi Represented as node p m The i-th evaluation parameter of process quality, v mj Represented as node p m The jth influencing parameter of process quality, m≤n; i=1,2,...; j=1,2,...; Construct the node p to be optimized m The evaluation model of process quality is used to calculate the node p according to the model. m Quantitative evaluation value of process quality U m ; The quantitative evaluation value and influencing parameter v obtained mj Perform grey correlation analysis between them to obtain node p m Quantitative evaluation value of process quality and various influencing parameters v mj The correlation between mj , setting the correlation threshold to filter out the influencing parameters that have a strong correlation with the quantitative evaluation value; The obtained strongly correlated influencing parameter data is used as input to optimize the node p m The quantitative evaluation value of process quality is pre-trained as the output, and a neural network model is constructed. According to the model, the strongly correlated influencing parameters and node p are obtained. m The mapping relationship between the quantitative evaluation values ​​of process quality is obtained, and the optimization parameters of various strongly correlated influencing parameters are determined based on the obtained mapping relationship.

2. The method for optimizing the processing parameters of a motor housing based on machine learning according to claim 1 is characterized in that: The expression of the evaluation model is: Where U m is the node p to be optimized m Quantitative evaluation value of process quality; For node p m The weight of the i-th evaluation parameter of process quality.

3. The machine learning-based motor housing processing parameters according to claim 1 The optimization method is characterized in that Correlation mj The calculation formula is as follows: Among them, Δ min and Δ max They are the minimum difference between the two levels and the maximum difference between the two levels of the quantitative evaluation value and the standardized data of the influencing parameters, and the calculation formulas are: D min =min(min(X0(k)-X j (k))); Δ max =max(max(X0(k)-X j (k))); X0(k) is the node p m The kth data of the standardized value of the quantitative evaluation value of the process quality; X j (k) is the node p m The kth data of the jth item of the quantitative evaluation value of process quality affecting the standardized value of the parameter; min(min(·)) means first taking the minimum value of all factors in (·), and then taking the minimum value of all minimum values; max(max(·)) means first taking the maximum value of all factors in (·), and then taking the maximum value of all maximum values; λ is the data volume; ρ is the constant coefficient, and its value range is (0,1).

4. The method for optimizing the processing parameters of a motor housing based on machine learning according to claim 1, characterized in that: The neural network model is an input layer-hidden layer-output layer structure, wherein the number of input layer nodes p, the number of intermediate layer nodes s and the number of output layer nodes q satisfy the following relationship: Wherein, a is an integer from 1 to 9. In this model, the number of output layer nodes q=1; the number of input layer nodes p is equal to the number of influencing parameters that are strongly correlated with the quantitative evaluation value.

5. The method for optimizing the processing parameters of a motor housing based on machine learning according to claim 1, characterized in that: The value range of the correlation threshold is [0.7, 1].

6. The method for optimizing the processing parameters of a motor housing based on machine learning according to claim 2, characterized in that: Weight of evaluation parameters Satisfies the relationship: and is proportional to the node p m The absolute value of the correlation coefficient between the quantitative evaluation value of process quality and the evaluation parameter.

7. The method for optimizing the processing parameters of a motor housing based on machine learning according to claim 6 is characterized in that: Node p m The calculation formula of the correlation coefficient between the quantitative evaluation value of process quality and the evaluation parameters is: In the formula, r mj Represented as node p m The correlation coefficient between the quantitative evaluation value of process quality and the jth evaluation parameter; U mt For node p m The tth data value of the quantitative evaluation value of process quality; For node p m The mean of the quantitative evaluation value of process quality; u mjt Represented as node p m The tth data value of the jth evaluation parameter of process quality; Represented as node p m The mean value of the jth evaluation parameter of process quality; T is the total amount of data.

Images