Intelligent mechanical equipment design method based on case reasoning

Abstract

The invention provides a case reasoning-based mechanical equipment intelligent design method. The method comprises the following steps: constructing a mechanical equipment design case library; calculating subjective and objective weights of design features, optimizing subject and guest weight distribution, and calculating a mixed weight; taking the new problem as a retrieval condition, and identifying candidate cases based on weighted similarity; normalizing problem attributes of the candidate cases according to column maxima to form a similarity matrix; solving utility values of candidate case problem attributes to form utility vectors; normalizing problem parts of the candidate cases according to column maxima to form a similarity matrix; combining the similarity matrix and the grey incidence relation matrix, and obtaining a weighted average matrix after normalization; and solving the value of the solution feature of the new problem by using a weighted average method according to the elements of the weighted average matrix so as to obtain a solution. According to the invention, the efficiency and the intelligent level of mechanical equipment design can be improved.

Description

Technical Field

[0001] The invention belongs to the field of mechanical equipment design, and in particular to an intelligent design method of mechanical equipment based on case reasoning. Background Art

[0002] Modern machinery companies are facing increasingly fierce competition. In order to survive and develop in the fierce market competition, companies must adopt advanced design methods, shorten product design cycles, and improve product design success rates. Only in this way can they adapt to the ever-changing market demands and maintain a strong position in the market competition.

[0003] With the rapid development of computer technology and intelligent technology, case-based reasoning (CBR) technology has been widely used in the field of computer-aided parametric mechanical design. CBR systems are data-driven and generate new solutions by using historical cases. They follow the four core principles of retrieval, reuse, revision and retention, highlighting the importance of historical experience in the new design process. However, this technology still faces two major challenges in practical applications. The first is the accuracy of case retrieval. Due to the difference in weights of mechanical design parameters, the reasonable allocation of these weights has a direct impact on the retrieval results. Secondly, facing K similar cases, how to effectively adapt cases has become the second difficult problem. In most cases, the solutions of old cases are not always directly applicable to new problems. Therefore, exploring how to mine implicit knowledge from the solutions of similar cases to improve the accuracy of case adaptation is also a challenging task. These two problems highlight the complexity of the application of case-based reasoning technology in the field of mechanical design, and also point out the development direction of this technology, that is, how to improve the design efficiency and accuracy of CBR systems by improving case retrieval and adaptation strategies.

[0004] In the process of reviewing relevant domestic and foreign literature and patents, no literature or patents that are exactly the same as the method proposed in this invention were found. Researchers mainly apply case-based reasoning methods to fault diagnosis. Patent CN201680013547.2 uses CBR to determine the relationship between building automation and metadata or enterprise data. In addition, in research involving case-based reasoning, the most similar cases are often selected as candidate cases, and these candidate cases are rewritten based on predefined rules. Patent CN115905469A introduces a case retrieval method based on similarity and a case retrieval method based on fuzzy nearest neighbors to identify a single most similar candidate case for decision support. These methods do not take into account the possible contributions of other similar cases and rely too much on defined rules, resulting in a low degree of automation and intelligence. Summary of the invention

[0005] The purpose of the present invention is to provide a method for intelligent design of mechanical equipment based on case-based reasoning, aiming to improve the efficiency and intelligence level of mechanical equipment design.

[0006] The technical solution to achieve the purpose of the present invention is:

[0007] A method for intelligent design of mechanical equipment based on case-based reasoning, comprising:

[0008] Step S1: construct a mechanical equipment design case library;

[0009] The case library consists of multiple past design cases, each of which is represented as a vector containing m problem attributes and n solution attributes;

[0010] Step S2: first obtain the subjective and objective weights of the design problem, then optimize the distribution of subjective and objective weights to obtain the mixed weight;

[0011] Firstly, the subjective weight and objective weight of the mechanical equipment design parameter problem attributes are obtained, and the importance of the mechanical equipment design parameters is determined by minimizing the sum of the deviations of the subjective and objective weights and using the gradient descent algorithm to obtain the mixed weight.

[0012] Step S3: using the new question as the search condition, identifying candidate cases based on weighted similarity;

[0013] Step S4: normalize the problem attributes of the candidate cases according to the column maximum values ​​to form a similarity matrix;

[0014] Step S5: Obtain the utility values ​​of the problem attributes of the candidate cases to form a utility vector;

[0015] Step S6: Calculate the grey correlation between the problem attribute and the solution attribute of the candidate case, normalize the utility vector ratio to obtain the utility weight, and use the utility weight to correct the grey correlation to obtain the grey correlation matrix;

[0016] Step S7: combining the similarity matrix and the grey correlation matrix, and obtaining a weighted average matrix after normalization;

[0017] Step S8: Using the weighted average method according to the elements of the weighted average matrix, the value of the solution feature of the new problem is obtained to obtain a solution.

[0018] Compared with the prior art, the present invention has the following significant advantages:

[0019] First, the case retrieval part based on hybrid weights integrates subjective and objective weights to prioritize different design parameters to identify multiple similar candidate cases in the case library. Secondly, the case adaptation part based on hybrid weighted average decouples the implicit knowledge between candidate cases and covers three aspects: the similarity between the new case problem and the candidate case problem, the applicability of each candidate case, and the gray relationship between each candidate case problem and the solution. Finally, by integrating these three types of implicit knowledge, a solution can be obtained. This method makes full use of historical cases in the case library and mines the implicit knowledge between candidate cases, realizing the intelligent design of mechanical equipment under multiple similar cases, and improving the design efficiency, design accuracy and intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a functional module diagram of the intelligent design method of mechanical equipment based on case reasoning of the present invention.

[0021] Figure 2 It is a flow chart of the intelligent design method of mechanical equipment based on case reasoning of the present invention.

[0022] Figure 3 It is a schematic diagram of a mechanical press workbench designed as an object in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0024] Combination Figure 1 , Figure 2 , a method for intelligent design of mechanical equipment based on case-based reasoning, the steps are as follows:

[0025] Step S1: construct a mechanical equipment design case library;

[0026] The mechanical equipment design case library is constructed through knowledge representation and case library construction functions. The case library consists of multiple past design cases. Each case is represented as a vector containing m problem attributes and n solution attributes. The zth case in the case library is represented as:

[0027] C z ={p z1 ,p z2 ,...,p zi ,...,p zm ,s z1 ,s z2 ,...,s zj ...,s zn}1≤i≤m;1≤j≤n

[0028] Among them, P zirepresents the i-th problem attribute of the z-th case in the case library; S zj represents the j-th solution attribute of the z-th case in the case base, m is the number of problem attributes of each case in the case base, and n is the number of solution attributes of each case in the case base.

[0029] The new design requirements are expressed as:

[0030] C0={p 01 ,p 02 ,...,p 0i ,...,p 0m ,s 01 ,s 02 ,...,s 0j ...,s 0n}1≤i≤m;1≤j≤n

[0031] Among them, p 0i represents the known i-th problem attribute of the new design problem, S 0j It represents the j-th solution attribute to be solved for the new design problem. m and n represent the number of new design problem attributes and solution attributes, which are numerically the same as the number of problem attributes and solution attributes of the cases in the case library.

[0032] Step S2: first obtain the subjective and objective weights of the design problem, then optimize the distribution of subjective and objective weights to obtain the mixed weight;

[0033] First, the subjective weight and objective weight of the mechanical equipment design parameter problem attributes are obtained, and then the mixed weight is obtained to identify the importance of mechanical equipment design parameters and ensure the accuracy of case retrieval. Among them, the subjective weight is obtained by the hierarchical analysis method and expressed as:

[0034] W s =(w s1 ,w s2 ,...,w si ,...,w sm )

[0035] Among them, w si is the subjective weight of the attribute of the i-th question;

[0036] The objective weight is obtained by the entropy weight method and is expressed as:

[0037] W o =(w o1 ,w o2 ,...,w oi ,...,w om )

[0038] Among them, w oi is the objective weight of the i-th question attribute;

[0039] The mixed weight takes into account both subjective and objective weights, and is obtained by the gradient descent algorithm by minimizing the sum of the deviations of subjective and objective weights. Suppose the mixed weight is:

[0040] W=(w h1 ,w h2 ,...,w hi ,...,w hm )

[0041] Among them, w hi is the mixed weight of the i-th problem attribute;

[0042] In order to consider both subjective and objective weights, the deviation function is introduced:

[0043]

[0044] Among them, d z represents the deviation of the decision made by the zth case under subjective weight and mixed weight, h z represents the deviation of the decision made by the zth case under the objective weight and the mixed weight. In order to obtain the mixed weight and minimize the deviation, the target programming problem is constructed as follows:

[0045]

[0046] As shown in the above formula, l is the total number of cases in the case library. Given the subjective weight and objective weight, the mixed weight can be obtained using the gradient descent method.

[0047] Step S3: using the new question as the search condition, identifying candidate cases based on weighted similarity;

[0048] The identification of candidate cases is achieved through the similar case identification function. First, the similarity of each case is calculated according to the following formula:

[0049]

[0050] Among them, sim z is the similarity of the zth case, p zi is the i-th problem attribute of the z-th case.

[0051] Then, K cases with the highest similarity ranking are identified to construct a candidate case set, and the sth candidate case C s It is expressed as:

[0052]

[0053] in, is the i-th problem attribute in the s-th candidate case; is the j-th solution attribute in the s-th candidate case.

[0054] Step S4: normalize the problem attributes of the candidate cases according to the column maximum values ​​to form a similarity matrix;

[0055] The similarity matrix reflects the similarity between the new design problem attributes and the candidate case problem attributes. It is composed of the problem attributes in the candidate case normalized by the column maximum value, including K similar cases and m problem attributes. The representation method of the similarity matrix is ​​as follows:

[0056]

[0057] Among them, S is the similarity matrix, is the similarity value of the kth row and mth column.

[0058] Step S5: Obtain the utility values ​​of the problem attributes of the candidate cases to form a utility vector;

[0059] The utility vector is obtained through utility analysis to measure the preference for possible results. The utility vector is expressed as follows:

[0060] U=(u1,u2,…,u s ,…u K ) T

[0061]

[0062] In the formula, u s is the utility value of the sth candidate case.

[0063] Step S6: Calculate the grey correlation between the problem attribute and the solution attribute of the candidate case, normalize the utility vector ratio to obtain the utility weight, and use the utility weight to correct the grey correlation to obtain the grey correlation matrix;

[0064] Normalize the weights of the elements in the utility vector U to obtain the utility weights,

[0065]

[0066] In the formula, is the normalized utility weight value of the sth candidate case.

[0067] In order to reduce the impact of different data dimensions on the element values ​​in the candidate case set, the elements in the candidate case set are all subjected to maximum value normalization.

[0068]

[0069] In the formula, and are the element values ​​in the normalized candidate case set. and The grey correlation coefficient between can be calculated as:

[0070]

[0071] In the formula, ε is the resolution coefficient, and then the utility weight is used to correct the calculation of the grey relational degree.

[0072]

[0073] In the formula, r ij for and The grey correlation degree between them can be obtained, and the grey correlation matrix can be recorded as:

[0074]

[0075] Step S7: Combine the two matrices S and R and obtain a weighted average matrix after normalization;

[0076]

[0077] Among them, WM is the weighted average matrix, w sj is the element in the sth row and jth column of the weighted average matrix.

[0078] Step S8: Use the weighted average method to find a solution to the new problem.

[0079] The solution characteristics of the new problem are obtained by outputting the solution function. First, the weighted average matrix is ​​normalized and the calculation method is as follows:

[0080]

[0081] Among them, HWM is the weight matrix after normalization. is the element in the sth row and jth column of the weight matrix after weight normalization, w sj is the element in the sth row and jth column of the weighted average matrix;

[0082] The solution method for the solution attribute corresponding to the new design problem is:

[0083]

[0084] Among them, s 0j is the value of the j-th solution feature to be found in the new design problem, from which the solution corresponding to the new design problem can be obtained.

[0085] Example 1

[0086] Combination Figure 3In this embodiment, the design target to be determined is the outer dimensions of the upper and lower working tables of the mechanical press. The steps of a mechanical equipment intelligent design method based on case reasoning include:

[0087] Step S1: construct a mechanical equipment design case library;

[0088] Combination Figure 3 The problem attributes in the case library constructed in this embodiment include the nominal force F g , Nominal force stroke S g , slider stroke S, slider stroke times n, maximum die height H and die height adjustment ΔH, which reflect the working capacity of the press, the size range of machinable parts and related productivity indicators. The solution attributes include the length L, width W, thickness T of the lower template and the width W of the upper template s Length L s ,Right now:

[0089] C k ={F g ,S g ,S,n,H,ΔH,L,W,T,W s ,L s}

[0090] The case library contains 128 data, and the representative data are shown in Table 1:

[0091] Table 1 Mechanical press case library

[0092]

[0093]

[0094] The problem attributes of the design objectives are shown in Table 2:

[0095] Table 2 Newly designed question attributes

[0096]

[0097] Step S2: Obtain the subjective and objective weights of the design features, optimize the distribution of subjective and objective weights, and obtain the mixed weights;

[0098] Invite experts in the field to score the problem attributes and use the AHP method to obtain the subjective weights, which are expressed as:

[0099] W s =[0.3735 0.2462 0.1462 0.11753 0.06691 0.04779]

[0100] Extract the problem part of the design case in the case library, and use the entropy weight method to obtain the objective weight of the corresponding problem attribute, which is expressed as:

[0101] W o =[0.23084 0.1424 0.19108 0.23537 0.1422 0.05811]

[0102] The idea of ​​comprehensive evaluation is introduced, and the gradient descent method is used to minimize the sum of the deviations between the mixed weight and the subjective weight and between the mixed weight and the objective weight. The optimized mixed weight is expressed as:

[0103] W=[0.3029 0.1898 0.1769 0.173 0.1046 0.0528]

[0104] Step S3: using the new question as the search condition, identifying candidate cases based on weighted similarity;

[0105] The similarity of each case is calculated according to the mixed weight and weighted similarity calculation formula obtained in step S2, and the first four cases with the largest similarity are extracted to form a candidate case set. The candidate case set is listed in Table 3:

[0106] Table 3 Candidate case set

[0107]

[0108] Step S4: Use the question parts of the candidate cases to form a similarity matrix;

[0109] The problem parts in the candidate cases are normalized according to the column maximum value to form a similarity matrix, which is expressed as:

[0110]

[0111] Step S5: Obtain the utility value of the question part of the candidate case to form a utility vector;

[0112] The utility of a candidate case is the comprehensive score of each case under the mixed weight, and the utility vector is expressed as:

[0113]

[0114] Step S6: Calculate the grey correlation degree between the candidate case problem and the solution part, and use the utility vector to correct the grey correlation degree to obtain the grey correlation relationship matrix;

[0115] Normalize the weights of the elements in the utility vector U to obtain the utility weights,

[0116]

[0117] In the formula, is the normalized utility weight value of the sth candidate case.

[0118] In order to reduce the impact of different data dimensions on the element values ​​in the candidate case set, the elements in the candidate case set are all subjected to maximum value normalization.

[0119]

[0120] In the formula, and are the element values ​​in the normalized candidate case set. and The grey correlation coefficient between can be calculated as:

[0121]

[0122] In the formula, ε is the resolution coefficient, and then the utility weight is used to correct the calculation of the grey relational degree.

[0123]

[0124] In the formula, r ij for and The grey correlation degree between them can be obtained, and the grey correlation matrix can be recorded as:

[0125]

[0126] In this embodiment, the grey correlation matrix is ​​calculated and expressed as follows:

[0127]

[0128] Step S7: combining the two matrices and normalizing the weights to obtain a weighted average matrix;

[0129] Combining the two matrices yields WM, calculated as:

[0130]

[0131] After normalizing WM according to column weights, we get the weighted average matrix HWM. The weighted average matrix is ​​calculated as:

[0132]

[0133] Step S8: using the weighted average method to find a solution to the new problem;

[0134] The solution to the new problem can be obtained by weighted averaging the candidate case problem part and the corresponding position elements in HWM, which is expressed as:

[0135] s 01=1100×0.2573+740×0.2394+840×0.2517+840×0.2517=882.95

[0136] s 02 =2250×0.2569+1300×0.2386+2250×0.2522+1300×0.2522=2130.47

[0137] s 03 =170×0.2583+170×0.2369+170×0.2524+170×0.2524)=165.26

[0138] s 04 =1000×0.2572+580×0.2390+650×0.2519+650×0.2519=745.96

[0139] s 05 =2300×0.2557+770×0.2406+1800×0.2518+1900×0.2518=1730.37

[0140] Therefore, the final design results are: lower template length L = 882.95mm, width W = 2130.47mm, thickness T = 165.26mm; upper template width W s =745.96mm, length L s =1730.37mm.

Claims

1. A method for intelligent design of mechanical equipment based on case-based reasoning, characterized in that: include: Step S1: construct a mechanical equipment design case library; The case library consists of multiple past design cases, each of which is represented as a vector containing m problem attributes and n solution attributes; Step S2: first obtain the subjective and objective weights of the design problem, then optimize the distribution of subjective and objective weights to obtain the mixed weight; Firstly, the subjective weight and objective weight of the mechanical equipment design parameter problem attributes are obtained, and the importance of the mechanical equipment design parameters is determined by minimizing the sum of the deviations of the subjective and objective weights and using the gradient descent algorithm to obtain the mixed weight. Step S3: using the new question as the search condition, identifying candidate cases based on weighted similarity; Step S4: normalize the problem attributes of the candidate cases according to the column maximum values ​​to form a similarity matrix; Step S5: Obtain the utility values ​​of the problem attributes of the candidate cases to form a utility vector; Step S6: Calculate the grey correlation between the problem attribute and the solution attribute of the candidate case, normalize the utility vector ratio to obtain the utility weight, and use the utility weight to correct the grey correlation to obtain the grey correlation matrix; Step S7: combining the similarity matrix and the grey correlation matrix, and obtaining a weighted average matrix after normalization; Step S8: Using the weighted average method according to the elements of the weighted average matrix, the value of the solution feature of the new problem is obtained to obtain a solution.

2. The method for intelligent design of mechanical equipment based on case reasoning according to claim 1 is characterized in that: The construction of mechanical equipment design case library is expressed as: C z ={p z1 ,p z2 ,...,p zi ,...,p zm ,s z1 ,s z2 ,...,s zj ...,s zn } 1≤i≤m;1≤j≤n Among them, P zi represents the i-th problem attribute of the z-th case in the case library; S zj The j-th solution attribute of the z-th case in the case library.

3. The method for intelligent design of mechanical equipment based on case reasoning according to claim 1, characterized in that: The process of obtaining the mixed weight is: Introduce the deviation function: Among them, d z represents the deviation of the decision made by the zth case under subjective weight and mixed weight, h z represents the deviation of the decision made in the zth case under the objective weight and the mixed weight; w si is the subjective weight of the attribute of the i-th question; w oi is the objective weight of the i-th problem attribute; w hi is the mixed weight of the i-th problem attribute; P zi represents the i-th problem attribute of the z-th case in the case base; The goal programming problem is constructed as follows: l is the total number of cases in the case library. Given the subjective weight and objective weight, the mixed weight can be obtained using the gradient descent method.

4. The method for intelligent design of mechanical equipment based on case-based reasoning according to claim 1, characterized in that: The identification of candidate cases is achieved through the similar case identification function. First, the similarity of each case is calculated: Among them, sim z is the similarity of the zth case, p zi is the i-th problem attribute of the z-th case, p 0i represents the known i-th problem attribute of the new design problem; w hi is the mixed weight of the i-th problem attribute; Then, K cases with the highest similarity ranking are identified to construct a candidate case set, and the sth candidate case C s It is expressed as: in, is the i-th problem attribute in the s-th candidate case; is the j-th solution attribute in the s-th candidate case.

5. The method for intelligent design of mechanical equipment based on case reasoning according to claim 4 is characterized in that: The utility vector is expressed as follows: U=(u1,u2,…,u s ,…u K ) T In the formula, u s is the utility value of the sth candidate case.

6. The method for intelligent design of mechanical equipment based on case reasoning according to claim 4 is characterized in that: The grey incidence matrix is ​​recorded as: where r ij for and The grey correlation between for and The grey correlation coefficient between and They are and The element value in the candidate case set after normalization, is the normalized utility weight value of the sth candidate case.

7. The method for intelligent design of mechanical equipment based on case-based reasoning according to claim 4 is characterized in that: To obtain the solution characteristic value of the new problem, first normalize the weighted average matrix. The calculation method is as follows: Among them, HWM is the weight matrix after normalization. is the element in the sth row and jth column of the weight matrix after weight normalization, w sj is the element in the sth row and jth column of the weighted average matrix; K is the number of candidate case sets; The solution method for the solution attribute corresponding to the new design problem is: Among them, s 0j is the value of the j-th solution feature to be found in the new design problem, from which the solution corresponding to the new design problem can be obtained.

Images