A Rapid Technical Support Safety Evaluation Method for Complex Equipment

Abstract

The present invention discloses a method for evaluating the safety of rapid technical support for complex equipment, which decomposes it into three primary evaluation factors; and then gradually decomposes them into 15 secondary evaluation factors, 9 tertiary evaluation factors, and 7 quaternary evaluation factors, and uses the method of scoring by experts at the same level for comprehensive scoring. In order to adapt to the irregular quaternary structure, a special radial basis neural network with a five-level weight superposition type is established, which has a network structure that shrinks layer by layer and upward, so as to be able to adapt to irregular data patterns; at the same time, the posterior data of experts at the same level and higher-level experts in the past are used to train the weights of the neural network, and a weight adaptive adjustment law with a sine-cosine hybrid fractional-order error feedback is designed according to the network training error to achieve weight convergence. Finally, the powerful parallel computing ability of the neural network is used to simulate the complex relationship between safety and quaternary evaluation factors, and an evaluation conclusion is made on the safety of the rapid technical support of the equipment to be evaluated.

Description

Technical Field

[0001] The present invention relates to a method for rapidly evaluating the safety of technical support for complex equipment, and belongs to the fields of equipment reliability evaluation and safety evaluation prediction. Background Art

[0002] The formation of the rapid technical support ability of equipment is a long-term and complex process, which involves many links and very high technical content. The quality requirements for its support process are correspondingly very high. Each operation and each inspection have strict process requirements, and it is required that the operators must have high technical qualities and the quality of the completed operations meet the quality control requirements; there are many key processes, many key control points, and strict technical acceptance; especially when accelerating, it is very easy to bring potential safety hazards, so the safety requirements are also high. Among them, safety regulations, personnel safety, equipment safety, facility safety, information safety, etc. are affected by many factors, so its safety evaluation is a very complex matter; although the existing on-site expert scoring method is simple to operate, it cannot take into account the empirical value of historical data in the past, so it is increasingly difficult to meet higher safety evaluation requirements.

[0003] For the above reasons, the present invention decomposes its evaluation into 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors. In order to adapt to the irregular four-level structure, a five-level weight superposition type special radial basis neural network is established, which has a network structure that shrinks layer by layer and upward, so as to be able to adapt to the irregular historical data patterns given by previous expert scores; use the previous local expert score data and the posterior data of the superior experts' safety evaluation to train the weights of the neural network, and design a weight adaptive adjustment law with a mixed sine-cosine fractional-order error feedback according to the network training error to achieve weight convergence. Finally, use the powerful parallel computing ability of the neural network to simulate the complex relationship between safety and the four-level evaluation factors, and realize the safety evaluation of the equipment to be evaluated. This method not only has high innovation in theory, but also has high engineering practical value. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: a method for rapidly evaluating the safety of technical support for complex equipment, so as to solve the problem of difficult hierarchical modeling of the irregular structure of rapid technical support in the above-mentioned existing safety evaluation methods.

[0005] The technical solution adopted by the present invention is: a method for rapidly evaluating the safety of technical support for complex equipment, and the method includes the following steps:

[0006] Step S10: Select the historical data of n pieces of equipment. Divide the safety assessment of the rapid technical support process for each piece of equipment into three first-level evaluation factors: safety of support conditions, safety of support protection, and safety of support processes. Then, decompose the safety of support conditions into four second-level evaluation factors: temperature and humidity, static electricity, high-pressure gas, and electronic devices. Among them, the temperature and humidity are further decomposed into four third-level evaluation factors: environmental temperature and humidity, support equipment temperature and humidity, temperature and humidity of explosive products, and temperature and humidity of flammable products. Decompose the safety of support protection into four second-level evaluation factors: safety of protection devices, explosion protection measures, prevention of combustibles, and anti-theft safety. Among them, the safety of protection devices is further decomposed into explosion-proof equipment in ordinary test rooms, protection devices in explosive test rooms, and protection devices in flammable test rooms, which are three third-level evaluation factors. The explosion protection measures are divided into two third-level evaluation factors: internal measure safety and external energy excitation safety. Among them, the internal measure safety is decomposed into five fourth-level evaluation factors: grounding, short-circuit sockets, explosion-proof sockets, explosion-proof air conditioners, and ventilation in filling areas. The external energy excitation safety is decomposed into two fourth-level evaluation factors: isolation walls and isolation nets. The safety of support processes is decomposed into seven fourth-level evaluation factors: disassembly safety, transportation safety, sub-assembly test safety, installation safety, overall test safety, technical preparation safety, and departure safety.

[0007] Step S20: For the 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors in the rapid technical support process of each piece of equipment, conduct a comprehensive score using the method of expert scoring at this level. First, conduct a comprehensive score for the 7 fourth-level evaluation factors of grounding, short-circuit sockets, explosion-proof sockets, explosion-proof air conditioners, ventilation in filling areas, isolation walls, and isolation nets of the i-th piece of equipment, using a percentage value, denoted as a ijkwl , which represents the evaluation score of the l-th fourth-level evaluation factor of the w-th third-level evaluation factor of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th piece of equipment; then conduct a comprehensive score for the 7 third-level evaluation factors that are not decomposed into fourth-level evaluation factors, denoted as a ijkw , which represents the evaluation score of the w-th third-level evaluation factor of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th piece of equipment; then conduct a comprehensive score for the 12 second-level evaluation factors that are not decomposed into third-level evaluation factors, denoted as a ijk , which represents the evaluation score of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th piece of equipment; secondly, conduct a comprehensive score for the 12 first-level evaluation factors that are not decomposed into second-level evaluation factors, denoted as a ijk , which represents the evaluation score of the j-th first-level evaluation factor of the i-th piece of equipment; finally, record the comprehensive safety evaluation score of the i-th piece of equipment after the superior experts complete the task, denoted as b i .

[0008] Step S30: Based on the evaluation score data of 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the rapid technical support process of n pieces of equipment by experts at this level, a special radial basis neural network with an irregular five-layer progressive structure is established. First, according to the evaluation scores of the 7 fourth-level evaluation factors of the i-th piece of equipment, namely grounding, short-circuit socket, explosion-proof socket, explosion-proof air conditioner, ventilation in the filling area, isolation wall, and isolation net, 30 center values of the fourth-level nodes of the neural network are selected respectively. Then, the deviation data of the fourth-level center points of the network is obtained by comparison. And the sensitive intervals of the second-level nodes of the neural network are set, and after the absolute value exponential transformation, the absolute value square exponential transformation, and the absolute value square root exponential transformation of the deviation are carried out respectively and then linearly superimposed, the mixed transformation data of the deviation index of the fourth-level center points of the network is obtained. Then, it is multiplied by the corresponding weights of the fourth-level neural network and accumulated for 30 nodes to obtain the comprehensive output of the neural network for the l-th fourth-level evaluation factor, the w-th third-level evaluation factor, the k-th second-level evaluation factor, and the j-th first-level evaluation factor of the i-th piece of equipment.

[0009] Step S40: According to the structural composition characteristics of 3 first-level evaluation factors, 15 second-level evaluation factors, and 9 third-level evaluation factors of the rapid technical support process of the equipment, first, the comprehensive outputs of the neural network for 7 fourth-level evaluation factors are divided into two groups for signal synthesis to obtain the comprehensive scores of 2 third-level evaluation factors decomposed into fourth-level evaluation factors. Then, the comprehensive scores of the 2 third-level evaluation factors are combined with the comprehensive scores of the 7 third-level evaluation factors not decomposed into fourth-level evaluation factors to form the comprehensive scores of 9 third-level evaluation factors. Then, for the comprehensive scores of 9 third-level evaluation factors, 30 center values of the third-level nodes of the neural network are selected respectively. Then, the deviation data of the third-level center points of the network is obtained by comparison. And according to the sensitive intervals of the second-level nodes of the neural network, after the absolute value exponential transformation, the absolute value square exponential transformation, and the absolute value square root exponential transformation of the deviation are carried out respectively and then linearly superimposed, the mixed transformation data of the deviation index of the third-level center points of the network is obtained. Then, it is multiplied by the corresponding weights of the third-level neural network and accumulated for 30 nodes to obtain the comprehensive output of the neural network for the w-th third-level evaluation factor, the k-th second-level evaluation factor, and the j-th first-level evaluation factor of the i-th piece of equipment.

[0010] Step S50: According to the structural composition characteristics of the three first-level evaluation factors and fifteen second-level evaluation factors in the rapid technical support process of the equipment, first, based on the comprehensive output of the neural network for nine third-level evaluation factors, they are divided into three groups for signal synthesis to obtain the comprehensive scores of the three second-level evaluation factors decomposed into third-level evaluation factors; then, combined with the comprehensive scores of the twelve second-level evaluation factors that are not decomposed into third-level evaluation factors, the comprehensive scores of the fifteen second-level evaluation factors are formed; then, for the comprehensive scores of the fifteen second-level evaluation factors, the central values of 30 neural network nodes are respectively selected; then, the deviation data of the network secondary center points are obtained by comparison; and according to the sensitive intervals of the neural network secondary nodes, after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation and then performing linear superposition, the mixed transformation data of the network secondary center point deviation index are obtained; then, multiplied by the corresponding neural network secondary weights and accumulated for 30 nodes, the comprehensive output of the neural network for the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th equipment is obtained.

[0011] Step S60: According to the structural composition characteristics of the three first-level evaluation factors in the rapid technical support process of the equipment, first, based on the comprehensive output of the neural network for fifteen second-level evaluation factors, they are divided into three groups for signal synthesis to obtain the comprehensive scores of all three first-level evaluation factors; then, for the comprehensive scores of the three first-level evaluation factors, the central values of 30 neural network first-level nodes are respectively selected; then, the deviation data of the network first-level center points are obtained by comparison; and by setting the sensitive intervals of the neural network first-level nodes, after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation and then performing linear superposition, the mixed transformation data of the network first-level center point deviation index are obtained, and then multiplied by the corresponding neural network first-level weights and accumulated for 30 nodes, the comprehensive output of the neural network for the j-th first-level evaluation factor of the i-th equipment is obtained; then, multiplied by the corresponding neural network primary weights to obtain the total output of the safety assessment of the rapid technical support process of the i-th equipment.

[0012] Step S70: Compare the total output data of the safety assessment of the rapid technical support process of the i-th equipment obtained from the neural network with the comprehensive safety assessment score of the i-th equipment by the superior experts after the equipment completes the task to obtain the network training error data; then, respectively design the adaptive adjustment laws of the network primary weights, network first-level weights, network second-level weights, network third-level weights, and network fourth-level weights based on the positive and cosine hybrid fractional-order error feedback according to the network training error data, and finally, respectively perform adaptive updates on the network primary weights, network first-level weights, network second-level weights, network third-level weights, and network fourth-level weights through integral operations; when the network training error converges to the vicinity of 0, stop training and weight update.

[0013] Step S80: Substitute the evaluation score data of 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the rapid technical support process of the equipment to be evaluated by the local experts into the trained neural network to obtain the safety evaluation score of the rapid technical support process of the equipment to be evaluated by the neural network.

[0014] In an exemplary embodiment of the present invention, according to the evaluation scores of 7 fourth-level evaluation factors, namely the grounding, short-circuit socket, explosion-proof socket, explosion-proof air conditioner, ventilation in the filling area, isolation wall, and isolation net of the i-th equipment, 30 neural network fourth-level node central values are respectively selected; then the deviation data of the network fourth-level center point is obtained by comparison; and the sensitive intervals of the neural network second-level nodes are set, and after the absolute value exponential transformation, the absolute value square exponential transformation, and the square root exponential transformation of the deviation are respectively performed and linearly superimposed, the mixed transformation data of the deviation index of the network fourth-level center point is obtained; then it is multiplied by the corresponding neural network fourth-level weight and accumulated for 30 nodes to obtain the comprehensive output of the neural network for the l-th fourth-level evaluation factor, the w-th third-level evaluation factor, the k-th second-level evaluation factor, and the j-th first-level evaluation factor of the i-th equipment, including:

[0015] ε ijkwlp = a ijkwl - a 1jkwlp ;

[0016]

[0017] where a 1jkwlp is the central value of 30 neural network fourth-level nodes, ε ijkwlp is the deviation data of the network fourth-level center point; σ jk is the sensitive interval of the neural network second-level node, which is a constant parameter; c1 and c2 are constant parameters; γ ijkwlp is the mixed transformation data of the deviation index of the network fourth-level center point; u ijkwl is the comprehensive output of the neural network for the l-th fourth-level evaluation factor, the w-th third-level evaluation factor, the k-th second-level evaluation factor, and the j-th first-level evaluation factor of the i-th equipment, and k 4jkwlp is the neural network fourth-level weight.

[0018] In an exemplary embodiment of the present invention, the comprehensive output of seven four-level evaluation factors by a neural network is divided into two groups for signal integration to obtain the comprehensive scores of two three-level evaluation factors decomposed into four-level evaluation factors; then, the comprehensive scores of the seven three-level evaluation factors that are not decomposed into four-level evaluation factors are combined to form the comprehensive scores of nine three-level evaluation factors; then, for the comprehensive scores of the nine three-level evaluation factors, 30 central values of the three-level nodes of the neural network are respectively selected; then, the deviation data of the three-level center points of the network is obtained by comparison; and according to the sensitive intervals of the two-level nodes of the neural network, after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation and then performing linear superposition, the deviation index mixed transformation data of the three-level center points of the network is obtained; then, it is multiplied by the corresponding three-level weights of the neural network and accumulated for 30 nodes, and the comprehensive output of the neural network for the w-th three-level evaluation factor of the k-th secondary evaluation factor of the j-th primary evaluation factor of the i-th equipment includes:

[0019]

[0020] ε ijkwp =a ijkw -a 1jkwp ;

[0021]

[0022] where a 1jkwp is the central values of 30 three-level nodes of the neural network. When p = 1, 2,..., 30, j = 2, k = 8, w = 1, l = 1, 2,..., 5; when j = 2, k = 8, w = 2, l = 1, 2; ε ijkwp is the deviation data of the three-level center points of the network; γ ijkwp is the deviation index mixed transformation data of the three-level center points of the network; u ijkw is the comprehensive output of the neural network for the w-th three-level evaluation factor of the k-th secondary evaluation factor of the j-th primary evaluation factor of the i-th equipment, k 3jkwp is the three-level weights of the neural network.

[0023] In an exemplary embodiment of the present invention, the comprehensive output of nine third-level evaluation factors based on a neural network is divided into three groups for signal synthesis to obtain the comprehensive score of three second-level evaluation factors decomposed into third-level evaluation factors; then, the comprehensive score of 12 second-level evaluation factors not decomposed into third-level evaluation factors is combined to form the comprehensive score of 15 second-level evaluation factors; then, for the comprehensive scores of 15 second-level evaluation factors, 30 neural network node central values are respectively selected; then, the deviation data of the network second-level center points is obtained by comparison; and according to the sensitive intervals of the neural network second-level nodes, after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation, linear superposition is performed to obtain the network second-level center point deviation index mixed transformation data; then, it is multiplied by the corresponding neural network second-level weight and accumulated for 30 nodes, and the comprehensive output of the neural network for the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th equipment includes:

[0024]

[0025] ε ijkp =a ijk -a 1jkp ;

[0026]

[0027] where a 1jkp is the central value of 30 neural network second-level nodes, where p = 1, 2, …, 30. When j = 1 and k = 1, w = 1, 2, 3, 4; when j = 2 and k = 1, w = 1, 2, 3, 4; when j = 2 and k = 4, w = 1, 2, and δ ijkp is the deviation data of the network second-level center point; γ ijkp is the network second-level center point deviation index mixed transformation data; u ijk is the comprehensive output of the neural network for the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th equipment, and k 2jkp is the neural network second-level weight.

[0028] In an exemplary embodiment of the present invention, the comprehensive output of 15 secondary evaluation factors based on a neural network is divided into three groups for signal synthesis to obtain the comprehensive scores of all 3 primary evaluation factors; then, for the comprehensive scores of the 3 primary evaluation factors, 30 center values of the first-level nodes of the neural network are respectively selected; then, the deviation data of the first-level center points of the network is obtained by comparison; and the sensitive intervals of the first-level nodes of the neural network are set, and after the absolute value exponential transformation, the square exponential transformation of the absolute value of the deviation, and the square root exponential transformation of the absolute value of the deviation are respectively performed and linearly superimposed, the mixed transformation data of the deviation index of the first-level center points of the network is obtained, and then multiplied by the corresponding first-level weights of the neural network and accumulated for 30 nodes to obtain the comprehensive output of the neural network for the j-th primary evaluation factor of the i-th equipment; then multiplied by the corresponding primary weights of the neural network to obtain the total output of the safety evaluation of the rapid technical support process of the i-th equipment by the neural network, including:

[0029]

[0030] ε ijp =a ij -a 1jp ;

[0031]

[0032] where a 1jp is the center value of 30 first-level nodes of the neural network, where p = 1, 2,..., 30. When j = 1, k = 1, 2, 3, 4; when j = 2, k = 1, 2, 3; when j = 3, k = 1, 2,..., 7, ε ijp is the deviation data of the first-level center points of the network; σ j is the sensitive interval of the first-level nodes of the neural network, which is a constant parameter; γ ijp is the mixed transformation data of the deviation index of the first-level center points of the network; u ij is the comprehensive output of the neural network for the j-th primary evaluation factor of the i-th equipment, k 1jp is the secondary weight of the neural network; k 0j is the primary weight of the neural network; y i is the total output of the safety evaluation of the rapid technical support process of the i-th equipment by the neural network.

[0033] In an exemplary embodiment of the present invention, based on the network training error data, the adaptive adjustment laws of the primary weights of the network based on the mixed fractional-order error feedback of sine and cosine, the adaptive adjustment law of the first-level weights of the network, the adaptive adjustment law of the secondary weights of the network, the adaptive adjustment law of the third-level weights of the network, and the adaptive adjustment law of the fourth-level weights of the network are respectively designed. Finally, the primary weights of the network, the first-level weights of the network, the secondary weights of the network, the third-level weights of the network, and the fourth-level weights of the network are adaptively updated through integral operations, including:

[0034] e i = y i -b i ;

[0035]

[0036] k 0j (n + 1)= k 0j (n)+ k 0jd ;

[0037] k 1jp (n + 1)= k 1jp (n)+ k 1jpd ;

[0038]

[0039] k 2jkp (n + 1)= k 2jkp (n)+ k 2jkpd ;

[0040]

[0041] k 3jkwp (n + 1)= k 3jkwp (n)+ k 3jkwpd ;

[0042]

[0043] k 4jkwlp (n + 1)= k 4jkwlp (n)+ k 4jkwlpd ;

[0044] In an exemplary embodiment of the present invention, according to the evaluation score data of 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the rapid technical support process of the equipment to be evaluated by the experts at this level, substituting them into the trained neural network, the safety evaluation score of the rapid technical support process of the equipment to be evaluated by the neural network includes:

[0045]

[0046] where γ Djkwlp is the network fourth-level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; γ Djkwp is the network third-level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; γ Djkp is the network second-level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; γ DjpThe network-level central point deviation index mixed transformation data of the equipment to be evaluated by the neural network; the above data is generated by substituting the evaluation score data of 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the rapid technical support process of the equipment to be evaluated by experts at this level into the trained neural network; u Djkwl is the comprehensive output of the neural network for the l-th fourth-level evaluation factor, the w-th third-level evaluation factor, the k-th second-level evaluation factor, and the j-th first-level evaluation factor of the equipment to be evaluated, u Djkw is the comprehensive output of the neural network for the w-th third-level evaluation factor, the k-th second-level evaluation factor, and the j-th first-level evaluation factor of the equipment to be evaluated, u Djk is the comprehensive output of the neural network for the k-th second-level evaluation factor and the j-th first-level evaluation factor of the equipment to be evaluated, u Dj is the comprehensive output of the neural network for the j-th first-level evaluation factor of the equipment to be evaluated, y D is the safety evaluation score of the rapid technical support process of the equipment to be evaluated by the neural network.

[0047] Advantages of the present invention

[0048] Compared with the prior art, the method adopted by the present invention has the following three major innovation points. The first is to decompose the rapid technical support of complex equipment into three first-level evaluation factors: guarantee condition safety, guarantee protection safety, and guarantee process safety according to the background characteristics and actual situation of the rapid technical support of complex equipment; and then gradually decompose them into 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors, establishing an irregular four-level structure model. The second is to establish a five-level weight superposition type dedicated radial basis neural network, which has a network structure that fills in layers and shrinks upward, so as to be able to adapt to the irregular data patterns of the safety evaluation of the rapid technical support of complex equipment. The third is to propose a weight adaptive adjustment law using a sine-cosine hybrid fractional-order error feedback to achieve rapid convergence of the neural network weights, so as to accelerate the network training process, speed up the weight convergence speed, and reduce the probability of network weight divergence during the training process. Description of the drawings

[0049] The drawings here are incorporated into the specification and form a part of this specification, showing embodiments in line with the present invention, and are used together with the specification to explain the principles of the present invention. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0050] Figure 1 is a flowchart of a method for evaluating the safety of the rapid technical support of complex equipment;

[0051] Figure 2It is the structure diagram of 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors for the rapid technical support of the equipment provided by the embodiments of the present invention;

[0052] Figure 3 It is the diagram of the convergence of the neural network training error of the method provided by the embodiments of the present invention;

[0053] Figure 4 It is the convergence diagram of the neural network weight k 01 of the method provided by the embodiments of the present invention;

[0054] Figure 5 It is the convergence diagram of the neural network weight k 111 of the method provided by the embodiments of the present invention;

[0055] Figure 6 It is the convergence diagram of the neural network weight k 2111 of the method provided by the embodiments of the present invention. Specific embodiments

[0056] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the following further elaborates on the present invention in detail with reference to specific embodiments and the appended Figure 1 drawings.

[0057] Embodiment 1: Taking the rapid technical support process of 260 pieces of equipment stored in the warehouse, the comprehensive scoring data of 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the rapid technical support process of these 260 pieces of equipment by the experts at this level, and the comprehensive evaluation score of the safety of the equipment after completing the task by the experts at the higher level, to train the established neural network; finally, using the comprehensive scoring data of the safety of 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the equipment to be evaluated by the experts at this level to complete the safety comprehensive evaluation of the rapid technical support process of the equipment to be evaluated as a background case, a safety evaluation method for the rapid technical support of complex equipment is described. This method includes the following steps:

[0058] Step S10 can be decomposed into the following three sub-steps. First step, select the historical data of 260 pieces of equipment, and divide the safety evaluation of the rapid technical support process of each piece of equipment into three first-level evaluation factors: safety of guarantee conditions, safety of guarantee protection, and safety of guarantee process; as shown in the structure diagram in the appended Figure 2 drawings, then decompose the safety of guarantee conditions into four second-level evaluation factors: temperature and humidity, static electricity, high-pressure gas, and electronic devices.

[0059] In the second step, the temperature and humidity are decomposed into four third-level evaluation factors: ambient temperature and humidity, temperature and humidity of support equipment, temperature and humidity of explosive products, and temperature and humidity of flammable products; ensuring protection safety is decomposed into four second-level evaluation factors: safety of protection devices, explosion protection measures, prevention of combustibles, and anti-theft safety; among which, the safety of protection devices is further decomposed into explosion-proof equipment in ordinary test rooms, protection devices in explosive test rooms, and protection devices in flammable test rooms, three third-level evaluation factors; explosion protection measures are divided into two third-level evaluation factors: internal measure safety and external energy excitation safety.

[0060] In the third step, the internal measure safety is decomposed into five fourth-level evaluation factors: grounding, short-circuit sockets, explosion-proof sockets, explosion-proof air conditioners, and ventilation in filling areas; the external energy excitation safety is decomposed into two fourth-level evaluation factors: isolation walls and isolation nets; ensuring process safety is decomposed into seven fourth-level evaluation factors: disassembly safety, transportation safety, sub-assembly test safety, installation safety, overall test safety, technical preparation safety, and departure safety.

[0061] Step S20: For the 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors in the rapid technical support process of each piece of equipment, a comprehensive score is obtained by means of expert scoring at this level. Specifically, it can be decomposed into the following five sub-steps.

[0062] In the first step, a comprehensive score is obtained for the 7 fourth-level evaluation factors of grounding, short-circuit sockets, explosion-proof sockets, explosion-proof air conditioners, ventilation in filling areas, isolation walls, and isolation nets of the i-th (1 ≤ i ≤ 260) piece of equipment, using a percentage value, denoted as a ijkwl , which represents the evaluation score of the l-th fourth-level evaluation factor of the w-th third-level evaluation factor of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th piece of equipment.

[0063] In the second step, a comprehensive score is obtained for the 7 third-level evaluation factors that are not decomposed into fourth-level evaluation factors, denoted as a ijkw , which represents the evaluation score of the w-th third-level evaluation factor of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th piece of equipment.

[0064] In the third step, a comprehensive score is obtained for the 12 second-level evaluation factors that are not decomposed into third-level evaluation factors, denoted as a ijk , which represents the evaluation score of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th piece of equipment.

[0065] In the fourth step, a comprehensive score is obtained for the 12 first-level evaluation factors that are not decomposed into second-level evaluation factors, denoted as a ijk , which represents the evaluation score of the j-th first-level evaluation factor of the i-th piece of equipment.

[0066] The fifth step is to record the comprehensive safety evaluation score of the i-th equipment after the superior expert completes the task, which is recorded as b i .

[0067] In step S30, an irregular five-layer progressive specialized radial basis function neural network is established based on the evaluation scores of three first-level evaluation factors, 15 second-level evaluation factors, nine third-level evaluation factors, and seven fourth-level evaluation factors of the rapid technical support process for 260 pieces of equipment by experts at this level. This can be broken down into the following three steps.

[0068] In the first step, the evaluation scores of the seven four-level evaluation factors, namely, grounding, short-circuit sockets, explosion-proof sockets, explosion-proof air conditioners, ventilation of filling sites, isolation walls, and isolation nets of the i-th equipment are recorded as the seven four-level evaluation factors. 30 neural network four-level node center values are selected respectively; then, the four-level center point deviation data of the network is obtained by comparison as follows:

[0069] ε ijkwlp =a ijkwl -a 1jkwlp ;

[0070] where a 1jkwlp is the center value of the fourth-level nodes of 30 neural networks, where ε ijkwlp It is the deviation data of the fourth-level center point of the network.

[0071] In the second step, the sensitive interval of the secondary node of the neural network is set, and the absolute value exponential transformation, the absolute value square exponential transformation, and the absolute value square root exponential transformation of the deviation are respectively performed, and then linear superposition is performed to obtain the hybrid transformation data of the deviation exponential of the fourth-level center point of the network as follows:

[0072]

[0073] where σ jk is the sensitive interval of the secondary node of the neural network, which is a constant parameter; c1 and c2 are constant parameters; c1 = 0.8, c2 = 0.4, γ ijkwlp It is the hybrid transformation data of the deviation index of the fourth-level center point of the network.

[0074] In the third step, multiply the corresponding four-level weights of the neural network and perform 30 node accumulation to obtain the comprehensive output of the neural network for the jth first-level evaluation factor, the kth second-level evaluation factor, the wth third-level evaluation factor, and the lth fourth-level evaluation factor of the i-th equipment as follows:

[0075]

[0076] where u ijkwl is the comprehensive output of the neural network for the jth first-level evaluation factor, the kth second-level evaluation factor, the wth third-level evaluation factor, and the lth fourth-level evaluation factor of the i-th equipment, k4jkwlp It is the fourth-level weight of the neural network.

[0077] Step S40 can be specifically decomposed into the following four sub-steps. First, according to the structural composition characteristics of the 3 first-level evaluation factors, 15 second-level evaluation factors, and 9 third-level evaluation factors in the rapid technical support process of the equipment, based on the comprehensive output of the neural network for 7 fourth-level evaluation factors, they are divided into two groups for signal synthesis to obtain the comprehensive scores of 2 third-level evaluation factors decomposed into fourth-level evaluation factors; then, combined with the comprehensive scores of 7 third-level evaluation factors that are not decomposed into fourth-level evaluation factors, the comprehensive scores of 9 third-level evaluation factors are formed as follows:

[0078]

[0079] Second, for the comprehensive scores of 9 third-level evaluation factors, 30 neural network third-level node center values are respectively selected; then, the deviation data of the network third-level center points are obtained by comparing with them as follows:

[0080] ε ijkwp = a ijkw -a 1jkwp ;

[0081] where a 1jkwp is the 30 neural network third-level node center values. When p = 1, 2,..., 30, j = 2, k = 8, w = 1, l = 1, 2,..., 5; when j = 2, k = 8, w = 2, l = 1, 2; ε ijkwp is the deviation data of the network third-level center points.

[0082] Third, according to the sensitive intervals of the neural network second-level nodes, after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation and then performing linear superposition, the deviation index mixed transformation data of the network third-level center points are obtained as follows:

[0083]

[0084] where γ ijkwp is the deviation index mixed transformation data of the network third-level center points.

[0085] Fourth, multiply by the corresponding neural network third-level weights and perform accumulation for 30 nodes to obtain the comprehensive output of the neural network for the w-th third-level evaluation factor of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th equipment as follows:

[0086]

[0087] where u ijkwis the comprehensive output of the neural network for the w-th third-level evaluation factor of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th equipment, k 3jkwp is the third-level weight of the neural network.

[0088] Step S50 can be specifically decomposed into the following four sub-steps. First, according to the structural composition characteristics of the 3 first-level evaluation factors and 15 second-level evaluation factors in the rapid technical support process of the equipment, first, based on the comprehensive output of the neural network for 9 third-level evaluation factors, they are divided into three groups for signal synthesis to obtain the comprehensive scores of the 3 second-level evaluation factors decomposed into third-level evaluation factors; then, combined with the comprehensive scores of the 12 second-level evaluation factors not decomposed into third-level evaluation factors, the comprehensive scores of the 15 second-level evaluation factors are formed as follows:

[0089]

[0090] Second, for the comprehensive scores of the 15 second-level evaluation factors, 30 neural network node central values are respectively selected; then, the deviation data of the network second-level center points are obtained by comparing with them as follows:

[0091] ε ijkp = a ijk - a 1jkp ;

[0092] where a 1jkp is the central value of the 30 neural network second-level nodes, where p = 1, 2,..., 30. When j = 1, k = 1, w = 1, 2, 3, 4; when j = 2, k = 1, w = 1, 2, 3, 4, and when j = 2, k = 4, w = 1, 2.

[0093] Third, according to the sensitive intervals of the neural network second-level nodes, after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation and then performing linear superposition, the deviation index mixed transformation data of the network second-level center points are obtained as follows:

[0094]

[0095] where ε ijkp is the deviation data of the network second-level center points; γ ijkp is the deviation index mixed transformation data of the network second-level center points.

[0096] Fourth, multiply by the corresponding neural network second-level weights and perform accumulation of 30 nodes to obtain the comprehensive output of the neural network for the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th equipment as follows:

[0097]

[0098] where uijk is the comprehensive output of the neural network for the k-th secondary evaluation factor of the j-th primary evaluation factor of the i-th equipment, k 2jkp is the secondary weight of the neural network.

[0099] Step S60 can be specifically decomposed into the following four sub-steps. First, according to the structural composition characteristics of the three primary evaluation factors in the rapid technical support process of the equipment, first divide the comprehensive outputs of the neural network for the 15 secondary evaluation factors into three groups for signal synthesis, and obtain the comprehensive scores of all three primary evaluation factors as follows:

[0100]

[0101] Second, for the comprehensive scores of the three primary evaluation factors, respectively select the central values of 30 neural network primary nodes; then compare with them to obtain the deviation data of the network primary center points as follows:

[0102] ε ijp = a ij - a 1jp ;

[0103] where a 1jp is the central value of 30 neural network primary nodes, where p = 1, 2,..., 30. When j = 1, k = 1, 2, 3, 4; when j = 2, k = 1, 2, 3; when j = 3, k = 1, 2,..., 7, ε ijp is the deviation data of the network primary center point.

[0104] Third, set the sensitive interval of the neural network primary nodes, perform linear superposition after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation, and obtain the deviation exponential mixed transformation data of the network primary center points as follows:

[0105]

[0106] where σ j is the sensitive interval of the neural network primary nodes and is a constant parameter; γ ijp is the deviation exponential mixed transformation data of the network primary center point.

[0107] Fourth, multiply by the corresponding neural network primary weights and perform accumulation of 30 nodes to obtain the comprehensive output of the neural network for the j-th primary evaluation factor of the i-th equipment as follows:

[0108]

[0109] where u ij is the comprehensive output of the neural network for the j-th primary evaluation factor of the i-th equipment, k 1jpIt is the secondary weight of the neural network.

[0110] Step 5: Multiply by the corresponding primary weights of the neural network to obtain the total output of the safety assessment of the rapid technical support process for the i-th piece of equipment by the neural network as follows:

[0111]

[0112] where k 0j is the primary weight of the neural network; y i is the total output of the safety assessment of the rapid technical support process for the i-th piece of equipment by the neural network.

[0113] Step S70 can be specifically decomposed into the following three sub-steps. Step 1: Compare the total output data of the safety assessment of the rapid technical support process for the i-th piece of equipment by the neural network with the comprehensive safety assessment score of the i-th piece of equipment after the superior expert has completed the task to obtain the network training error data as follows:

[0114] e i = y i - b i ;

[0115] where e i is the network training error data, and its convergence situation is as Figure 3 shown.

[0116] Step 2: Design the adaptive adjustment laws of the primary weights of the network, the first-level weights of the network, the second-level weights of the network, the third-level weights of the network, and the fourth-level weights of the network based on the positive and cosine hybrid fractional-order error feedback respectively according to the network training error data as follows:

[0117]

[0118] where k 0jd 、k 1jpd 、k 2jkpd 、k 3jkwpd 、k 4jkwlpd are the adaptive adjustment laws of the primary weights of the network, the first-level weights of the network, the second-level weights of the network, the third-level weights of the network, and the fourth-level weights of the network respectively; l 0j 、l 1jp 、l 2jkp 、l 3jkwp 、l 4jkwlp are constant parameters used to adjust the convergence speed of the neural network weights.

[0119] Step 3. Finally, the primary weights, first-level weights, second-level weights, third-level weights, and fourth-level weights of the network are adaptively updated through integral operations as follows:

[0120] k 0j (n + 1)= k 0j (n)+ k 0jd ;

[0121] k 1jp (n + 1)= k 1jp (n)+ k 1jpd ;

[0122] k 2jkp (n + 1)= k 2jkp (n)+ k 2jkpd ;

[0123] k 3jkwp (n + 1)= k 3jkwp (n)+ k 3jkwpd ;

[0124] k 4jkwlp (n + 1)= k 4jkwlp (n)+ k 4jkwlpd ;

[0125] Step 4. When the network training error converges to the region near 0, stop the network training and weight update. Finally, the convergence conditions of the neural network weights k 01 、k 111 、k 2111 are as shown in Figure 4 、 5 、6.

[0126] Step S80. The evaluation score data of the 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the rapid technical support process of the equipment to be evaluated by the local expert are as follows:

[0127] [85, 82, 78, 92, 88, 94, 76, 81, 89, 90, 88, 83, 87, 75, 92, 80, 90, 93, 92, 86, 74, 79, 86, 88, 93, 89]. Substitute them into the trained neural network to obtain the safety evaluation score of the rapid technical support process of the equipment to be evaluated by the neural network as follows:

[0128]

[0129]

[0130] where γ Djkwlp is the mixed transformation data of the network fourth-level center point deviation index of the neural network for the equipment to be evaluated; γDjkwp is the network three - level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; γ Djkp is the network two - level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; γ Djp is the network one - level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; The above data is generated by substituting the evaluation score data of 3 first - level evaluation factors, 15 second - level evaluation factors, 9 third - level evaluation factors, and 7 fourth - level evaluation factors in the rapid technical support process of the equipment to be evaluated by experts at this level into the trained neural network; u Djkwl is the comprehensive output of the neural network for the l - th fourth - level evaluation factor, the w - th third - level evaluation factor, the k - th second - level evaluation factor, and the j - th first - level evaluation factor of the equipment to be evaluated, u Djkw is the comprehensive output of the neural network for the w - th third - level evaluation factor, the k - th second - level evaluation factor, and the j - th first - level evaluation factor of the equipment to be evaluated, u Djk is the comprehensive output of the neural network for the k - th second - level evaluation factor and the j - th first - level evaluation factor of the equipment to be evaluated, u Dj is the comprehensive output of the neural network for the j - th first - level evaluation factor of the equipment to be evaluated, y D is the safety evaluation score of the rapid technical support process of the neural network for the equipment to be evaluated. The final result obtained is 86.73, so the conclusion is that the safety is good.

[0131] In the above - described specific embodiments, the purpose, technical solution, and beneficial effects of the present invention have been further described in detail. It should be understood that the above - described are only specific embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A fuzzy comprehensive evaluation method for complex equipment technical support, characterized by the following steps: Step S10: The rapid technical support process safety assessment of each piece of equipment is divided into three primary evaluation factors: support condition safety, protection safety, and process safety. The security assurance conditions are then broken down into four secondary evaluation factors: temperature and humidity, static electricity, high-pressure gas, and electronic devices; Among them, temperature and humidity are decomposed into four three-level evaluation factors: ambient temperature and humidity, support equipment temperature and humidity, explosive material temperature and humidity, and flammable material temperature and humidity; ensuring protection safety is decomposed into four two-level evaluation factors: protective device safety, explosion-proof measures safety, flammable material safety, and anti-theft safety; and among them, protective device safety is decomposed into three three-level evaluation factors: explosion-proof equipment in ordinary test rooms, protective devices in explosive test rooms, and protective devices in flammable test rooms; explosion-proof measures safety is divided into two three-level evaluation factors: internal measures safety and external energy excitation safety. Among them, internal measures safety is decomposed into five four-level evaluation factors: grounding, short-circuit sockets, explosion-proof sockets, explosion-proof air conditioners, and ventilation of filling sites; external energy excitation safety is decomposed into two four-level evaluation factors: isolation walls and isolation nets; ensuring process safety is decomposed into seven four-level evaluation factors: disassembly safety, transportation safety, extension test safety, installation safety, overall test safety, technical preparation safety, and departure safety; Step S20: For the 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors in the rapid technical support process of each piece of equipment, a comprehensive score is obtained by having experts at this level score. First, a comprehensive score is calculated for the 7 fourth-level evaluation factors of the grounding, short-circuit socket, explosion-proof socket, explosion-proof air conditioner, ventilation in the filling area, isolation wall, and isolation net of the i-th piece of equipment. Using a percentage value, it is denoted as a ijkwl , which represents the evaluation score of the l-th fourth-level evaluation factor, the w-th third-level evaluation factor, the k-th second-level evaluation factor, and the j-th first-level evaluation factor of the i-th piece of equipment; then, a comprehensive score is calculated for the 7 third-level evaluation factors that are not decomposed into fourth-level evaluation factors, denoted as a ijkw , which represents the evaluation score of the w-th third-level evaluation factor, the k-th second-level evaluation factor, and the j-th first-level evaluation factor of the i-th piece of equipment; then, a comprehensive score is calculated for the 12 second-level evaluation factors that are not decomposed into third-level evaluation factors, denoted as a ijk , which represents the evaluation score of the k-th second-level evaluation factor and the j-th first-level evaluation factor of the i-th piece of equipment; secondly, a comprehensive score is calculated for the 12 first-level evaluation factors that are not decomposed into second-level evaluation factors, denoted as a ijk , which represents the evaluation score of the j-th first-level evaluation factor of the i-th piece of equipment; finally, record the comprehensive evaluation score of the safety of the i-th piece of equipment after the superior experts complete the task, denoted as b i ; Step S30, based on the evaluation score data of 3 first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the rapid technical support process of n equipment by experts at this level, establish an irregular five-layer progressive special radial basis neural network; first, based on the evaluation scores of 7 fourth-level evaluation factors of the grounding, short-circuit socket, explosion-proof socket, explosion-proof air conditioner, ventilation of the filling site, isolation wall, and isolation net of the i-th equipment, 30 fourth-level node center values of the neural network are selected respectively; then, the fourth-level center point deviation data of the network is obtained by comparison; and the second-level node sensitive interval of the neural network is set, and the deviation absolute value exponential transformation, the deviation absolute value square exponential transformation, and the deviation absolute value square root exponential transformation are respectively performed and linearly superimposed to obtain the network fourth-level center point deviation exponential mixed transformation data; then, multiplied by the corresponding fourth-level weight of the neural network and accumulated by 30 nodes, the comprehensive output of the neural network for the jth first-level evaluation factor, the kth second-level evaluation factor, the wth third-level evaluation factor, and the lth fourth-level evaluation factor of the i-th equipment is obtained as follows: ε ijkwlp = a ijkwl -a 1jkwlp ; where a 1jkwlp is the central value of 30 four - level nodes of the neural network, and ε ijkwlp is the deviation data of the four - level center point of the network; σ jk is the sensitive interval of the second - level node of the neural network, which is a constant parameter; c1 and c2 are constant parameters; γ ijkwlp is the deviation - index mixed transformation data of the four - level center point of the network; u ijkwl is the comprehensive output of the neural network for the l - th four - level evaluation factor of the w - th three - level evaluation factor of the k - th second - level evaluation factor of the j - th first - level evaluation factor of the i - th equipment, and k 4jkwlp is the weight of the four - level of the neural network; Step S40: According to the structural composition characteristics of the three first-level evaluation factors, fifteen second-level evaluation factors, and nine third-level evaluation factors in the rapid technical support process of the equipment, first, based on the neural network, the comprehensive outputs of seven fourth-level evaluation factors are divided into two groups for signal synthesis to obtain the comprehensive scores of two third-level evaluation factors decomposed into fourth-level evaluation factors; then, combined with the comprehensive scores of the seven third-level evaluation factors not decomposed into fourth-level evaluation factors, the comprehensive scores of nine third-level evaluation factors are formed; then, for the comprehensive scores of the nine third-level evaluation factors, 30 neural network third-level node central values are selected respectively; then, the deviation data of the network third-level center points are obtained by comparison; and according to the sensitive intervals of the neural network second-level nodes, after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation and then performing linear superposition, the network third-level center point deviation index mixed transformation data are obtained; then, multiplied by the corresponding neural network third-level weights and accumulated for 30 nodes, the comprehensive output of the neural network for the w-th third-level evaluation factor of the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th equipment is as follows: ε ijkwp = a ijkw -a 1jkwp ; where a 1jkwp is the central value of 30 three - level nodes of the neural network. When p = 1, 2, …, 30, j = 2, k = 8, w = 1, l = 1, 2, …, 5; when j = 2, k = 8, w = 2, l = 1, 2; ε ijkwp is the deviation data of the three - level center point of the network; γ ijkwp is the mixed transformation data of the deviation index of the three - level center point of the network; u ijkw is the comprehensive output of the neural network for the w - th three - level evaluation factor of the k - th secondary evaluation factor of the j - th primary evaluation factor of the i - th equipment, k 3jkwp is the three - level weight of the neural network; Step S50: According to the structural composition characteristics of the three first-level evaluation factors and fifteen second-level evaluation factors in the rapid technical support process of the equipment, first, based on the neural network, the comprehensive outputs of nine third-level evaluation factors are divided into three groups for signal synthesis to obtain the comprehensive scores of three second-level evaluation factors decomposed into third-level evaluation factors; then, combined with the comprehensive scores of the twelve second-level evaluation factors not decomposed into third-level evaluation factors, the comprehensive scores of fifteen second-level evaluation factors are formed; then, for the comprehensive scores of the fifteen second-level evaluation factors, 30 neural network node central values are selected respectively; then, the deviation data of the network second-level center points are obtained by comparison; and according to the sensitive intervals of the neural network second-level nodes, after respectively performing deviation absolute value exponential transformation, deviation absolute value square exponential transformation, and deviation absolute value square root exponential transformation and then performing linear superposition, the network second-level center point deviation index mixed transformation data are obtained; then, multiplied by the corresponding neural network second-level weights and accumulated for 30 nodes, the comprehensive output of the neural network for the k-th second-level evaluation factor of the j-th first-level evaluation factor of the i-th equipment is as follows: ε ijkp = a ijk -a 1jkp ; where a 1jkp is the central value of 30 neural network secondary nodes, where p = 1, 2, …, 30. When j = 1, k = 1, w = 1, 2, 3, 4; when j = 2, k = 1, w = 1, 2, 3; when j = 2, k = 4, w = 1, 2, ε ijkp is the deviation data of the secondary center point of the network; γ ijkp is the deviation index mixed transformation data of the secondary center point of the network; u ijk is the comprehensive output of the neural network for the k-th secondary evaluation factor of the j-th primary evaluation factor of the i-th equipment, k 2jkp is the secondary weight of the neural network; Step S60: According to the structural composition characteristics of the three first-level evaluation factors in the rapid technical support process of the equipment, first, based on the neural network, the comprehensive outputs of the 15 second-level evaluation factors are grouped into three groups for signal synthesis to obtain the comprehensive scores of all three first-level evaluation factors; then, for the comprehensive scores of the three first-level evaluation factors, 30 central values of the first-level nodes of the neural network are selected respectively; then, the deviation data of the first-level center points of the network are obtained by comparison; and the sensitive intervals of the first-level nodes of the neural network are set, and after the absolute value exponential transformation, the absolute value square exponential transformation, and the square root exponential transformation of the absolute value of the deviation are carried out respectively and then linearly superimposed, the mixed transformation data of the deviation index of the first-level center points of the network are obtained, and then multiplied by the corresponding first-level weights of the neural network and accumulated for 30 nodes to obtain the comprehensive output of the neural network for the j-th first-level evaluation factor of the i-th equipment; then multiplied by the corresponding primary weights of the neural network to obtain the total output of the safety assessment of the rapid technical support process of the i-th equipment by the neural network as follows: ε ijp = a ij -a 1jp ; where a 1jp is the central value of 30 first-level nodes of the neural network. Here, p = 1, 2, …, 30. When j = 1, k = 1, 2, 3, 4; when j = 2, k = 1, 2, 3, 4; when j = 3, k = 1, 2, …, 7, ε ijp is the deviation data of the first-level center point of the network; σ j is the sensitive interval of the first-level nodes of the neural network and is a constant parameter; γ ijp is the deviation index mixing transformation data of the first-level center point of the network; u ij is the comprehensive output of the neural network for the j-th first-level evaluation factor of the i-th equipment, k 1jp is the second-level weight of the neural network; k 0j is the primary weight of the neural network; y i is the total output of the safety assessment of the rapid technical support process of the i-th equipment by the neural network; Step S70: Compare the total output data of the safety assessment of the rapid technical support process of the i-th equipment by the neural network with the comprehensive safety assessment score of the i-th equipment after the superior experts complete the task to obtain the network training error data; Then, based on the network training error data, design the adaptive adjustment laws of the primary weights of the network, the first-level weights of the network, the second-level weights of the network, the third-level weights of the network, and the fourth-level weights of the network based on the positive and cosine hybrid fractional-order error feedback respectively. Finally, adaptively update the primary weights of the network, the first-level weights of the network, the second-level weights of the network, the third-level weights of the network, and the fourth-level weights of the network through integral operations respectively; when the network training error converges to the area near 0, stop training and weight update; e i = y i -b i ; k 0j (n + 1) = k 0j (n) + k 0jd ; k 1jp (n + 1)= k 1jp (n)+ k 1jpd ; k 2jkp (n + 1) = k 2jkp (n) + k 2jkpd ; k 3jkwp (n + 1)= k 3jkwp (n)+ k 3jkwpd ; k 4jkwlp (n + 1) = k 4jkwlp (n) + k 4jkwlpd ; where e i is the network training error data, and k 0jd , k 1jpd , k 2jkpd , k 3jkwpd , k 4jkwlpd are the primary weight adaptive adjustment law of the network, the first-level weight adaptive adjustment law of the network, the second-level weight adaptive adjustment law of the network, the third-level weight adaptive adjustment law of the network, and the fourth-level weight adaptive adjustment law of the network respectively; l 0j , l 1jp , l 2jkp , l 3jkwp , l 4jkwlp are constant parameters used to adjust the convergence speed of the neural network weights. Step S80: Substitute the evaluation score data of the three first-level evaluation factors, 15 second-level evaluation factors, 9 third-level evaluation factors, and 7 fourth-level evaluation factors of the rapid technical support process of the equipment to be evaluated by the local experts into the trained neural network to obtain the comprehensive evaluation score of the rapid technical support process of the equipment to be evaluated by the neural network as follows: where γ Djkwlp is the network four - level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; γ Djkwp is the network three - level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; γ Djkp is the network two - level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; γ Djp is the network one - level center point deviation index mixed transformation data of the neural network for the equipment to be evaluated; The above data is generated by substituting the evaluation score data of 3 first - level evaluation factors, 15 second - level evaluation factors, 9 third - level evaluation factors, and 7 fourth - level evaluation factors in the rapid technical support process of the equipment to be evaluated by experts at this level into the trained neural network; u Djkwl is the comprehensive output of the neural network for the l - th fourth - level evaluation factor, the w - th third - level evaluation factor, the k - th second - level evaluation factor, and the j - th first - level evaluation factor of the equipment to be evaluated, u Djkw is the comprehensive output of the neural network for the w - th third - level evaluation factor, the k - th second - level evaluation factor, and the j - th first - level evaluation factor of the equipment to be evaluated, u Djk is the comprehensive output of the neural network for the k - th second - level evaluation factor and the j - th first - level evaluation factor of the equipment to be evaluated, u Dj is the comprehensive output of the neural network for the j - th first - level evaluation factor of the equipment to be evaluated, y D is the safety evaluation score of the rapid technical support process of the neural network for the equipment to be evaluated.

Images