Transportation equipment capability assessment method based on improved neural network

Abstract

The invention discloses a transportation equipment capability assessment method based on an improved neural network. The transport equipment capability assessment method based on the improved neural network comprises the following steps: firstly, constructing an index system influencing the transport capability, and determining an index importance degree by using an improved neural network algorithm; secondly, determining a membership function and an evaluation result based on a fuzzy theory; and finally, verifying the effectiveness of the evaluation method through experimental simulation and comparative analysis. The method provided by the invention plays an important role in improving the capacity of the transportation equipment and guiding the construction and development of the transportation equipment.

Description

Technical Field

[0001] The present invention relates to the field of transport equipment capability assessment, and more specifically, to a transport equipment capability assessment method based on an improved neural network. Background Art

[0002] The transport equipment capability assessment is an important basis for testing the transport equipment capability and promoting the development of transport equipment construction. At present, there are many capability assessment methods, such as fuzzy hierarchical analysis method, entropy weight method, fuzzy comprehensive evaluation method, evidence reasoning method, principal component analysis method and machine learning method.

[0003] The current capacity assessment methods still have the following problems: first, the accuracy of the assessment results is not high enough to accurately reflect the ability level; second, the assessment method is not objective enough and is affected by personal subjectivity and experience during the assessment process. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the deficiencies in the prior art and provide a method for evaluating the capability of transport equipment based on an improved neural network.

[0005] The present invention provides a method for evaluating the transport equipment capability based on an improved neural network, which is performed according to the following steps:

[0006] Step 1: Construction of transport equipment capability indicators

[0007] First, we should build an index system that affects transport capacity, and analyze the factors affecting transport equipment capacity from four aspects: operational capacity, mobility, survivability and general capacity. The operational capacity includes carrying capacity, loading and unloading capacity, and evacuation capacity. The mobility includes maneuverability, climbing ability, and the ability to cross obstacles or trenches. Survivability includes the equipment's protection, reliability, maintainability, testability, environmental adaptability, and safety. General capacity includes the level of three modernizations, the level of informatization and intelligence, and technical compatibility.

[0008] Step 2: Determine the importance of indicators using improved neural networks

[0009] Based on 16 sets of scoring data, each set of data is a score of the importance of 20 indicators corresponding to operational capability, mobility, survivability and general capabilities. The specific data are as follows:

[0010] The first set of scoring data = [0.4 0.4 0.1 0.1 0.4 0.4 0.2 0.4 0.3 0.3 0.1 0.30.30.1 0.1 0.1 0.25 0.25 0.25 0.25];

[0011] The second set of scoring data = [0.3 0.4 0.15 0.15 0.4 0.4 0.2 0.2 0.4 0.4 0.1 0.20.20.2 0.2 0.1 0.5 0.3 0.1 0.1];

[0012] The third set of scoring data = [0.3 0.3 0.2 0.2 0.5 0.25 0.25 0.3 0.3 0.4 0.1 0.30.10.05 0.4 0.05 0.3 0.3 0.2 0.2];

[0013] The fourth set of scoring data = [0.3 0.3 0.3 0.1 0.4 0.3 0.3 0.4 0.4 0.2 0.1 0.30.10.1 0.3 0.1 0.4 0.2 0.2 0.2];

[0014] The fifth set of scoring data = [0.5 0.3 0.15 0.05 0.4 0.3 0.3 0.3 0.35 0.35 0.10.20.2 0.1 0.2 0.2 0.25 0.25 0.25 0.25];

[0015] The sixth group of scoring data = [0.3 0.3 0.25 0.15 0.4 0.3 0.3 0.3 0.35 0.35 0.10.10.1 0.1 0.3 0.3 0.5 0.15 0.15 0.2];

[0016] The seventh group of scoring data = [0.35 0.35 0.15 0.15 0.4 0.4 0.2 0.2 0.4 0.4 0.10.30.3 0.1 0.1 0.1 0.3 0.3 0.2 0.2];

[0017] The eighth group of scoring data = [0.45 0.35 0.1 0.1 0.45 0.45 0.1 0.3 0.35 0.35 0.20.30.2 0.2 0.05 0.05 0.25 0.25 0.25 0.25];

[0018] The ninth group of scoring data = [0.3 0.4 0.15 0.15 0.3 0.3 0.4 0.2 0.4 0.4 0.1 0.20.20.2 0.2 0.1 0.3 0.3 0.2 0.2];

[0019] The tenth group of scoring data = [0.4 0.3 0.2 0.1 0.4 0.3 0.3 0.4 0.3 0.4 0.2 0.30.10.1 0.2 0.1 0.25 0.25 0.25 0.25];

[0020] The eleventh group of scoring data = [0.2 0.2 0.4 0.2 0.4 0.3 0.3 0.2 0.4 0.4 0.1 0.30.20.1 0.2 0.1 0.5 0.25 0.17 0.1];

[0021] The twelfth group of scoring data = [0.3 0.3 0.2 0.2 0.4 0.35 0.25 0.45 0.35 0.2 0.20.40.2 0.05 0.1 0.05 0.3 0.3 0.2 0.2];

[0022] The thirteenth group of scoring data = [0.5 0.25 0.15 0.1 0.4 0.4 0.2 0.2 0.4 0.4 0.20.20.2 0.1 0.2 0.1 0.4 0.3 0.2 0.1];

[0023] The fourteenth group of scoring data = [0.4 0.3 0.2 0.1 0.3 0.3 0.4 0.3 0.3 0.4 0.2 0.30.10.1 0.1 0.1 0.4 0.3 0.2 0.1];

[0024] The fifteenth group of scoring data = [0.35 0.35 0.2 0.1 0.4 0.4 0.2 0.2 0.4 0.4 0.10.20.1 0.1 0.3 0.2 0.45 0.25 0.2 0.1];

[0025] The sixteenth group of scoring data = [0.45 0.4 0.1 0.05 0.4 0.3 0.3 0.3 0.4 0.3 0.10.30.1 0.1 0.2 0.2 0.5 0.2 0.2 0.1];

[0026] The particle swarm optimization (PSO) algorithm is used to update the weights and thresholds in the BP neural network and then predict the importance of indicators. This can not only reduce the impact of subjective factors on the importance of indicators, but also effectively improve the prediction accuracy. The specific process is as follows:

[0027] (1) Establish BP neural network and determine the topological structure

[0028] It adopts a three-layer structure. The first layer is the input layer, which is used to input expert knowledge data. The number of neurons is consistent with the number of experts. The second layer is the hidden layer. The number of neurons in this layer is determined by the empirical formula Determine, where h, m and n represent the number of neurons in the hidden layer, input layer and output layer respectively, and λ is a constant between [1,10]. The third layer is the output layer, which is used to output the importance of the capability index, and the number of neurons is 1.

[0029] (2) Use the PSO algorithm to optimize and output the optimal weight and threshold

[0030] The fitness function of the PSO algorithm is set as:

[0031]

[0032] Among them, y and t represent the actual output and predicted output values ​​respectively, n1 is the number of capability indicators, and f is the fitness of the particle;

[0033] The particle reaches the new position by continuously updating its own speed, and moves closer to the optimal area according to the fitness, and finally obtains the optimal weight and threshold. The particle updates its position according to the following formula:

[0034]

[0035] x ij (k+1)=x ij (k)+v ij (k+1) (3)

[0036] Where, 1≤i≤n, 1≤j≤m, n=In num ×Hid num +Hid num +Hid num ×Out num +Out num is the number of optimized variables, where In num 、Hid num and Out num are the number of neurons in the input layer, hidden layer and output layer respectively, m is the number of populations; k is the number of iterations; rand1 and rand2 are two [0,1] independent random numbers; the inertia factor w is the degree of retention of the original velocity of the particle; the memory factors c1 and c2 are the step sizes for adjusting the particle to move towards the individual optimal and global optimal positions;

[0037] (3) Update the weights and thresholds of the BP neural network and output the indicator importance prediction results

[0038] The weight and threshold update formula of BP neural network is:

[0039] w1=PSO_w1 (4)

[0040] b1=PSO_b1 (5)

[0041] w2=PSO_w2 (6)

[0042] b2=PSO_b2 (7)

[0043] Among them, w1 and b1 are the connection weight and threshold from the input layer to the hidden layer respectively; w2 and b2 are the connection weight and threshold from the hidden layer to the output layer respectively. PSO_w1, PSO_b1, PSO_w2, PSO_b2 are the optimal weights and thresholds output after the PSO algorithm is optimized;

[0044] According to the above formula, the weights and thresholds of the neural network are updated to obtain the indicator importance prediction results;

[0045] Step 3: Determine membership function and evaluation results based on fuzzy theory

[0046] Before using fuzzy theory to determine the membership function and evaluation results, it is necessary to first establish a comment set and map the transport equipment capability index value to the comment set. The comment set is constructed as follows:

[0047] V={V1,V2,V3,V4,V5}={very weak, weak, medium, strong, very strong}

[0048] In order to solve the problem that the transport equipment capability index is highly ambiguous, the evaluation result is determined based on fuzzy theory, as shown in formula (8):

[0049]

[0050] Among them, i is the level number of capability indicators; B i and W i are the membership vector and weight vector of the i-th layer capability index respectively; R i Representation capability index u ij The evaluation membership matrix relative to the review set V; r ij Indicates that the i-th indicator corresponds to the comment set V j The matching degree of (1≤j≤5);

[0051] According to the trigonometric function calculation in fuzzy theory, the evaluation membership matrix can be obtained, as shown in formula (9), which can fuzzify the midpoint of the interval and avoid the problem of jumps between adjacent intervals;

[0052]

[0053] The comprehensive evaluation score Where B0 = [B1B2B3B4] T , is the comment set vector corresponding to the comment set V;

[0054] Step 4: Experimental simulation and analysis

[0055] Taking the capability of X transport equipment as the evaluation object, the correctness and effectiveness of the method are verified through experimental simulation and comparative analysis.

[0056] Through the above three-layer neural network structure, the 16 groups of scoring data in step 2 are collected. The scores of the indicator importance of any 15 groups of scoring data are used as input, and the other group of scoring data is used as output and imported into the training model; assuming that the learning rate of the network is 0.02, the maximum number of iterations is 200, and the minimum error is 0.001; the number of hidden layer nodes is selected to be 5, that is, the neural network topology is 15-5-1; the improved and traditional neural networks are used to predict the indicator importance respectively, and the prediction error of the improved neural network is 0.1368, and the prediction error of the traditional neural network is 1.0802. Obviously, the improved neural network has a smaller prediction error and higher accuracy.

[0057] Assign values ​​to the comment set according to the interval range of the membership function, and get the comment set vector Vr = [55 65 75 8595] T If the capability index score is known, then according to formulas (8) and (9), we can get:

[0058]

[0059] Among them, B1 is the membership vector of the carrying, loading and unloading, and racking and removal capability indicators; W1 is the importance of the above three capability indicators predicted by the improved neural network; R1 is the evaluation membership matrix of the above three capability indicators relative to the comment set V; then the operation capability score

[0060] Similarly, the mobility score Survivability Score General Ability Score

[0061] The comprehensive score of transport equipment capability P0 is:

[0062]

[0063] According to the principle of maximum membership, the capability of the transport equipment is between "medium" and "strong". In order to improve the capability of the transport equipment, its components are further analyzed: operational capability, mobility, survivability and general capability. According to the above evaluation results, P1>P2>P3>P4, that is, the operational capability is the strongest, the mobility is second, the survivability is third, and the general capability is the weakest; and only the operational capability is between "strong" and "very strong", the mobility is between "medium" and "strong", and the survivability and general capability are between "weak" and "medium". Therefore, if a certain capability is strong, the impact on the overall capability of the transport equipment is relatively limited. The various capabilities of the transport equipment should be developed in a balanced manner to make up for the shortcomings of survivability and general capability, and then improve the overall capability.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] The present invention combines the improved BP neural network with fuzzy theory. On the basis of constructing the transport equipment capability index system, the improved neural network algorithm is used to replace the expert scoring to determine the index importance; and the membership function is constructed based on the fuzzy theory to obtain the evaluation result; in addition, a certain transport equipment can be taken as an example to conduct experimental simulation and result analysis, and put forward countermeasures and suggestions to provide support for improving the transport equipment capability. Through the technical solution of the present invention, the problems of strong subjectivity and low accuracy in capability evaluation can be effectively solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a framework diagram of the transport equipment capability index system of the present invention;

[0067] Figure 2 Comparison chart of actual and predicted values ​​of traditional neural network samples;

[0068] Figure 3 Improved the comparison chart of actual and predicted values ​​of neural network samples;

[0069] Figure 4 Comparison of prediction errors between improved and traditional neural networks. DETAILED DESCRIPTION

[0070] In order to be able to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other. In the following description, many specific details are set forth in order to fully understand the present invention, and the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the art without making creative work are within the scope of protection of the present invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0071] The present invention provides a method for evaluating the transport equipment capability based on an improved neural network, which is performed according to the following steps:

[0072] Step 1: Construction of transport equipment capability indicators

[0073] To evaluate the transport equipment capability, we should first build an indicator system that affects the transport capability. Figure 1 As shown in the figure, the factors affecting the transport equipment capability are analyzed mainly from four aspects: operational capability, mobility, survivability and general capability. The operational capability is determined by the capabilities of carrying, loading and unloading, and evacuation; the mobility is determined by the capabilities of maneuverability, climbing, and crossing obstacles / trenchways; the survivability is determined by the equipment's protection, reliability, maintainability, testability, environmental adaptability, and safety; the general capability is determined by the level of modernization, informationization, and intelligence, and technical compatibility.

[0074] Step 2: Determine the importance of indicators using improved neural networks

[0075] Based on 16 sets of scoring data, each set of data is a score of the importance of 20 indicators corresponding to operational capability, mobility, survivability and general capabilities. The specific data are as follows:

[0076] The first set of scoring data = [0.4 0.4 0.1 0.1 0.4 0.4 0.2 0.4 0.3 0.3 0.1 0.30.30.1 0.1 0.1 0.25 0.25 0.25 0.25];

[0077] The second set of scoring data = [0.3 0.4 0.15 0.15 0.4 0.4 0.2 0.2 0.4 0.4 0.1 0.20.20.2 0.2 0.1 0.5 0.3 0.1 0.1];

[0078] The third set of scoring data = [0.3 0.3 0.2 0.2 0.5 0.25 0.25 0.3 0.3 0.4 0.1 0.30.10.05 0.4 0.05 0.3 0.3 0.2 0.2];

[0079] The fourth set of scoring data = [0.3 0.3 0.3 0.1 0.4 0.3 0.3 0.4 0.4 0.2 0.1 0.30.10.1 0.3 0.1 0.4 0.2 0.2 0.2];

[0080] The fifth set of scoring data = [0.5 0.3 0.15 0.05 0.4 0.3 0.3 0.3 0.35 0.35 0.10.20.2 0.1 0.2 0.2 0.25 0.25 0.25 0.25];

[0081] The sixth group of scoring data = [0.3 0.3 0.25 0.15 0.4 0.3 0.3 0.3 0.35 0.35 0.10.10.1 0.1 0.3 0.3 0.5 0.15 0.15 0.2];

[0082] The seventh group of scoring data = [0.35 0.35 0.15 0.15 0.4 0.4 0.2 0.2 0.4 0.4 0.10.30.3 0.1 0.1 0.1 0.3 0.3 0.2 0.2];

[0083] The eighth group of scoring data = [0.45 0.35 0.1 0.1 0.45 0.45 0.1 0.3 0.35 0.35 0.20.30.2 0.2 0.05 0.05 0.25 0.25 0.25 0.25];

[0084] The ninth group of scoring data = [0.3 0.4 0.15 0.15 0.3 0.3 0.4 0.2 0.4 0.4 0.1 0.20.20.2 0.2 0.1 0.3 0.3 0.2 0.2];

[0085] The tenth group of scoring data = [0.4 0.3 0.2 0.1 0.4 0.3 0.3 0.4 0.3 0.4 0.2 0.30.10.1 0.2 0.1 0.25 0.25 0.25 0.25];

[0086] The eleventh group of scoring data = [0.2 0.2 0.4 0.2 0.4 0.3 0.3 0.2 0.4 0.4 0.1 0.30.20.1 0.2 0.1 0.5 0.25 0.17 0.1];

[0087] The twelfth group of scoring data = [0.3 0.3 0.2 0.2 0.4 0.35 0.25 0.45 0.35 0.2 0.20.40.2 0.05 0.1 0.05 0.3 0.3 0.2 0.2];

[0088] The thirteenth group of scoring data = [0.5 0.25 0.15 0.1 0.4 0.4 0.2 0.2 0.4 0.4 0.20.20.2 0.1 0.2 0.1 0.4 0.3 0.2 0.1];

[0089] The fourteenth group of scoring data = [0.4 0.3 0.2 0.1 0.3 0.3 0.4 0.3 0.3 0.4 0.2 0.30.10.1 0.1 0.1 0.4 0.3 0.2 0.1];

[0090] The fifteenth group of scoring data = [0.35 0.35 0.2 0.1 0.4 0.4 0.2 0.2 0.4 0.4 0.10.20.1 0.1 0.3 0.2 0.45 0.25 0.2 0.1];

[0091] The sixteenth group of scoring data = [0.45 0.4 0.1 0.05 0.4 0.3 0.3 0.3 0.4 0.3 0.10.30.1 0.1 0.2 0.2 0.5 0.2 0.2 0.1];

[0092] The particle swarm optimization (PSO) algorithm is used to update the weights and thresholds in the BP neural network and then predict the importance of indicators. This can not only reduce the impact of subjective factors on the importance of indicators, but also effectively improve the prediction accuracy. The specific process is as follows:

[0093] (1) Establish BP neural network and determine the topological structure

[0094] It adopts a three-layer structure. The first layer is the input layer, which is used to input expert knowledge data. The number of neurons is consistent with the number of experts. The second layer is the hidden layer. The number of neurons in this layer is determined by the empirical formula (where h, m and n represent the number of neurons in the hidden layer, input layer and output layer respectively, and λ is a constant between [1,10]); the third layer is the output layer, which is used to output the importance of the ability index, and the number of neurons is 1.

[0095] (2) Use the PSO algorithm to optimize and output the optimal weight and threshold

[0096] The fitness function of the PSO algorithm is set as:

[0097]

[0098] Among them, y and t represent the actual output and predicted output values ​​respectively, n1 is the number of capability indicators, and f is the fitness of the particle.

[0099] The particle reaches the new position by continuously updating its own speed, and moves closer to the optimal area according to the fitness, and finally obtains the optimal weight and threshold. The particle updates its position according to the following formula:

[0100]

[0101] x ij (k+1)=x ij (k)+v ij (k+1) (3)

[0102] Where, 1≤i≤n, 1≤j≤m, n=In num ×Hid num +Hid num +Hid num ×Out num +Out num (In num 、Hid num and Out num are the number of neurons in the input layer, hidden layer and output layer respectively) is the number of optimization variables, m is the number of populations; k is the number of iterations; rand1 and rand2 are two [0,1] independent random numbers; the inertia factor w is the degree of retention of the original speed of the particle; the memory factors c1 and c2 are the step sizes for adjusting the particles to move towards the individual optimal and global optimal positions.

[0103] (3) Update the weights and thresholds of the BP neural network and output the indicator importance prediction results

[0104] The weight and threshold update formula of BP neural network is:

[0105] w1=PSO_w1 (4)

[0106] b1=PSO_b1 (5)

[0107] w2=PSO_w2 (6)

[0108] b2=PSO_b2 (7)

[0109] Among them, w1 and b1 are the connection weight and threshold from the input layer to the hidden layer, respectively; w2 and b2 are the connection weight and threshold from the hidden layer to the output layer, respectively. PSO_w1, PSO_b1, PSO_w2, and PSO_b2 are the optimal weights and thresholds output after the PSO algorithm is optimized.

[0110] By updating the weights and thresholds of the neural network according to the above formula, the indicator importance prediction results can be obtained.

[0111] Step 3: Determine membership function and evaluation results based on fuzzy theory

[0112] Before using fuzzy theory to determine the membership function and evaluation results, it is necessary to first establish a comment set and map the transport equipment capability index value to the comment set. The comment set is constructed as follows:

[0113] V={V1,V2,V3,V4,V5}={very weak, weak, medium, strong, very strong}

[0114] In order to solve the problem that the transport equipment capability index is highly ambiguous, the present invention determines the evaluation result based on fuzzy theory, as shown in formula (8).

[0115]

[0116] Among them, i is the level number of capability indicators; B i and W i are the membership vector and weight vector of the i-th layer capability index respectively; R i Representation capability index u ij The evaluation membership matrix relative to the review set V; r ij Indicates that the i-th indicator corresponds to the comment set V j The matching degree of (1≤j≤5).

[0117] The evaluation membership matrix can be obtained by calculating the trigonometric functions in fuzzy theory, as shown in formula (9), which can fuzzify the midpoint of the interval and avoid the problem of jumps between adjacent intervals.

[0118]

[0119] The comprehensive evaluation score Where B0 = [B1B2B3B4] T , is the comment set vector corresponding to the comment set V.

[0120] Step 4: Experimental simulation and analysis

[0121] Taking the capability of X transport equipment as the evaluation object, the correctness and effectiveness of the method are verified through experimental simulation and comparative analysis.

[0122] Through the above three-layer neural network structure, the 16 groups of scoring data in step 2 are collected. The scores of the indicator importance of any 15 groups of scoring data are used as input, and the other group of scoring data is used as output and imported into the training model. Assume that the learning rate of the network is 0.02, the maximum number of iterations is 200, and the minimum error is 0.001; the number of hidden layer nodes is selected to be 5, that is, the neural network topology is 15-5-1. Use the improved and traditional neural networks to predict the indicator importance (such as Figure 2 and Figure 3 As shown in the figure, the prediction error of the improved neural network is 0.1368, and the prediction error of the traditional neural network is 1.0802. Obviously, the prediction error of the improved neural network is smaller and the accuracy is higher. The error comparison chart is as follows Figure 4 shown.

[0123] Assign values ​​to the comment set according to the interval range of the membership function, and get the comment set vector Vr = [55 65 75 8595] T If the capability index score is known, then according to formulas (8) and (9), we can get:

[0124]

[0125] Among them, B1 is the membership vector of the carrying, loading and unloading, and racking and removal capability indicators; W1 is the importance of the above three capability indicators predicted by the improved neural network; R1 is the evaluation membership matrix of the above three capability indicators relative to the comment set V; then the operation capability score

[0126] Similarly, the mobility score Survivability Score General Ability Score

[0127] The comprehensive score of transport equipment capability P0 is:

[0128]

[0129] According to the principle of maximum membership, the capability of the transport equipment is between "medium" and "strong". In order to improve the capability of the transport equipment, its components are further analyzed: operational capability, mobility, survivability and general capability. According to the above evaluation results, P1>P2>P3>P4, that is, the operational capability is the strongest, the mobility is second, the survivability is third, and the general capability is the weakest; and only the operational capability is between "strong" and "very strong", the mobility is between "medium" and "strong", and the survivability and general capability are between "weak" and "medium". Therefore, if a certain capability is strong, the impact on the overall capability of the transport equipment is relatively limited. The various capabilities of the transport equipment should be developed in a balanced manner to make up for the shortcomings of survivability and general capability, and then improve the overall capability.

[0130] The above is only a preferred embodiment of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for evaluating the capability of transport equipment based on an improved neural network, characterized in that: Follow the steps below: Step 1: Construction of transport equipment capability indicators First, we should build an index system that affects transport capacity, and analyze the factors that affect transport equipment capacity from four aspects: operational capacity, mobility, survivability, and general capacity. The operational capability includes carrying capacity, loading and unloading capacity, and erection and withdrawal capacity; the mobility includes maneuverability, climbing capacity, and obstacle or trench crossing capacity; the survivability includes the equipment's protection, reliability, maintainability, testability, environmental adaptability, and safety; the general capability includes the three-in-one level, information and intelligent level, and technical compatibility; Step 2: Determine the importance of indicators using improved neural networks Based on 16 sets of scoring data, each set of data is a score of the importance of 20 indicators corresponding to operational capability, mobility, survivability and general capabilities. The specific data are as follows: The first set of scoring data = [0.4 0.4 0.1 0.1 0.4 0.4 0.2 0.4 0.3 0.3 0.1 0.30.3 0.10.1 0.1 0.25 0.25 0.25 0.25]; The second set of scoring data = [0.3 0.4 0.15 0.15 0.4 0.4 0.2 0.2 0.4 0.4 0.1 0.20.2 0.20.2 0.1 0.5 0.3 0.1 0.1]; The third set of scoring data = [0.3 0.3 0.2 0.2 0.5 0.25 0.25 0.3 0.3 0.4 0.1 0.30.10.05 0.4 0.05 0.3 0.3 0.2 0.2]; The fourth set of scoring data = [0.3 0.3 0.3 0.1 0.4 0.3 0.3 0.4 0.4 0.2 0.1 0.30.1 0.10.3 0.1 0.4 0.2 0.2 0.2]; The fifth set of scoring data = [0.5 0.3 0.15 0.05 0.4 0.3 0.3 0.3 0.35 0.35 0.10.2 0.20.1 0.2 0.2 0.25 0.25 0.25 0.25]; The sixth group of scoring data = [0.3 0.3 0.25 0.15 0.4 0.3 0.3 0.3 0.35 0.35 0.10.1 0.10.1 0.3 0.3 0.5 0.15 0.15 0.2]; The seventh group of scoring data = [0.35 0.35 0.15 0.15 0.4 0.4 0.2 0.2 0.4 0.4 0.10.3 0.30.1 0.1 0.1 0.3 0.3 0.2 0.2]; The eighth set of scoring data = [0.45 0.35 0.1 0.1 0.45 0.45 0.1 0.3 0.35 0.35 0.20.3 0.20.2 0.05 0.05 0.25 0.25 0.25 0.25]; The ninth group of scoring data = [0.3 0.4 0.15 0.15 0.3 0.3 0.4 0.2 0.4 0.4 0.1 0.20.2 0.20.2 0.1 0.3 0.3 0.2 0.2]; The tenth group of scoring data = [0.4 0.3 0.2 0.1 0.4 0.3 0.3 0.4 0.3 0.4 0.2 0.30.1 0.10.2 0.1 0.25 0.25 0.25 0.25]; The eleventh group of scoring data = [0.2 0.2 0.4 0.2 0.4 0.3 0.3 0.2 0.4 0.4 0.1 0.30.2 0.10.2 0.1 0.5 0.25 0.17 0.1]; The twelfth group of scoring data = [0.3 0.3 0.2 0.2 0.4 0.35 0.25 0.45 0.35 0.2 0.20.4 0.20.05 0.1 0.05 0.3 0.3 0.2 0.2]; The thirteenth group of scoring data = [0.5 0.25 0.15 0.1 0.4 0.4 0.2 0.2 0.4 0.4 0.20.2 0.20.1 0.2 0.1 0.4 0.3 0.2 0.1]; The fourteenth group of scoring data = [0.4 0.3 0.2 0.1 0.3 0.3 0.4 0.3 0.3 0.4 0.2 0.30.1 0.10.1 0.1 0.4 0.3 0.2 0.1]; The fifteenth group of scoring data = [0.35 0.35 0.2 0.1 0.4 0.4 0.2 0.2 0.4 0.4 0.10.2 0.10.1 0.3 0.2 0.45 0.25 0.2 0.1]; The sixteenth group of scoring data = [0.45 0.4 0.1 0.05 0.4 0.3 0.3 0.3 0.4 0.3 0.10.3 0.10.1 0.2 0.2 0.5 0.2 0.2 0.1]; The weights and thresholds in the BP neural network are updated through the particle swarm PSO algorithm to predict the importance of indicators. The specific process is as follows: (1) Establish BP neural network and determine the topological structure It adopts a three-layer structure. The first layer is the input layer, which is used to input expert knowledge data. The number of neurons is consistent with the number of experts. The second layer is the hidden layer. The number of neurons in this layer is determined by the empirical formula Determine, where h, m and n represent the number of neurons in the hidden layer, input layer and output layer respectively, and λ is a constant between [1,10]. The third layer is the output layer, which is used to output the importance of the capability index, and the number of neurons is 1. (2) Use the PSO algorithm to optimize and output the optimal weight and threshold The fitness function of the PSO algorithm is set as: Among them, y and t represent the actual output and predicted output values ​​respectively, n1 is the number of capability indicators, and f is the fitness of the particle; The particle reaches the new position by continuously updating its own speed, and moves closer to the optimal area according to the fitness, and finally obtains the optimal weight and threshold. The particle updates its position according to the following formula: x ij (k+1)=x ij (k)+v ij (k+1) (3) Where, 1≤i≤n, 1≤j≤m, n=In num ×Hid num +Hid num +Hid num ×Out num +Out num is the number of optimized variables, where In num 、Hid num and Out num are the number of neurons in the input layer, hidden layer and output layer respectively, m is the number of populations; k is the number of iterations; rand1 and rand2 are two [0,1] independent random numbers; the inertia factor w is the degree of retention of the original velocity of the particle; the memory factors c1 and c2 are the step sizes for adjusting the particle to move towards the individual optimal and global optimal positions; (3) Update the weights and thresholds of the BP neural network and output the indicator importance prediction results The weight and threshold update formula of BP neural network is: w1=PSO_w1 (4) b1=PSO_b1 (5) w2=PSO_w2 (6) b2=PSO_b2 (7) Among them, w1 and b1 are the connection weight and threshold from the input layer to the hidden layer respectively; w2 and b2 are the connection weight and threshold from the hidden layer to the output layer respectively. PSO_w1, PSO_b1, PSO_w2, PSO_b2 are the optimal weights and thresholds output after the PSO algorithm is optimized; According to the above formula, the weights and thresholds of the neural network are updated to obtain the indicator importance prediction results; Step 3: Determine membership function and evaluation results based on fuzzy theory Before using fuzzy theory to determine the membership function and evaluation results, it is necessary to first establish a comment set and map the transport equipment capability index value to the comment set. The comment set is constructed as follows: V={V1,V2,V3,V4,V5}={very weak, weak, medium, strong, very strong} In order to solve the problem that the transport equipment capability index is highly ambiguous, the evaluation result is determined based on fuzzy theory, as shown in formula (8): Among them, i is the level number of capability indicators; B i and W i are the membership vector and weight vector of the i-th layer capability index respectively; R i Representation capability index u ij The evaluation membership matrix relative to the review set V; r ij Indicates that the i-th indicator corresponds to the comment set V j (1≤j≤5) matching degree; According to the trigonometric function calculation in fuzzy theory, the evaluation membership matrix can be obtained, as shown in formula (9), which can fuzzify the midpoint of the interval and avoid the problem of jumps between adjacent intervals; The comprehensive evaluation score Where B0 = [B1B2B3B4] T , is the comment set vector corresponding to the comment set V; Step 4: Experimental simulation and analysis Through the above three-layer neural network structure, the 16 groups of scoring data in step 2 are collected; the scores of the indicator importance of any 15 groups of scoring data in the data set are used as input, and the other group of scoring data is used as output and imported into the training model; it is assumed that the learning rate of the network is 0.02, the maximum number of iterations is 200, and the minimum error is 0.001; the number of hidden layer nodes is selected to be 5, that is, the neural network topology structure is 15-5-1; the improved and traditional neural networks are used to predict the indicator importance respectively, the prediction error of the improved neural network is 0.1368, and the prediction error of the traditional neural network is 1.0802. Obviously, the improved neural network has a smaller prediction error and higher accuracy.

2. The method for evaluating the transport equipment capability based on an improved neural network according to claim 1 is characterized in that: In the fourth step, the comment set is assigned a value according to the interval range of the membership function to obtain the comment set vector If the capability index score is known, then according to formulas (8) and (9), we can get: Among them, B1 is the membership vector of the carrying, loading and unloading, and racking and removal capability indicators; W1 is the importance of the above three capability indicators predicted by the improved neural network; R1 is the evaluation membership matrix of the above three capability indicators relative to the comment set V; then the operation capability score Similarly, the mobility score Survivability Score General Ability Score The comprehensive score of transport equipment capability P0 is: According to the maximum membership principle, the transport equipment capability is between "medium" and "strong". In order to improve the transport equipment capability, its components are further analyzed: operational capability, mobility, survivability and general capability. According to the above evaluation results, P1>P2>P3>P4, that is, the operational capability is the strongest, the mobility is second, the survivability is third, and the general capability is the weakest; and only the operational capability is between "strong" and "very strong", the mobility is between "medium" and "strong", and the survivability and general capability are between "weak" and "medium".

Images