A Comprehensive Evaluation Method for Rapid Technical Support of Complex Equipment under Emergency Conditions

Abstract

The present invention discloses a comprehensive evaluation method for rapid technical support of complex equipment under emergency conditions. The emergency rapid support process of each piece of equipment is decomposed into 18 classifications in 9 sub-aspects of 2 aspects, namely equipment support quality and equipment support safety. The comprehensive score is obtained by means of scoring by experts at the same level, and the posterior evaluation score data of the equipment after completing the task by experts at the higher level are recorded. Then, a dedicated neural network based on the emphasis on equipment support time is established to simulate the complex non-linear relationship between the comprehensive evaluation of rapid technical support of complex equipment and the three aspects of equipment support quality, equipment support time, and equipment support safety. By utilizing the powerful parallel computing ability of the neural network, the existing historical data are fully utilized and the network error is trained to converge to the vicinity of 0, so as to substitute the data of the equipment to be evaluated into the trained neural network, and finally the comprehensive evaluation result of rapid technical support of complex equipment under emergency conditions is obtained.

Description

Technical Field

[0001] The present invention relates to a comprehensive evaluation method for rapid technical support of complex equipment under emergency conditions, and belongs to the fields of equipment reliability assessment and accident risk assessment and prediction. Background Art

[0002] Due to the influence of sudden accidents such as fires, earthquakes, floods, local wars, major sudden operation accidents, etc., under emergency conditions, complex equipment needs to be rapidly technically supported to prepare for its normal functions and use. The guarantee principle is to simplify the technical support process as much as possible, make the condition preparation as simple as possible, the necessary technical links cannot be less, the confirmatory technical links can be reduced according to the situation, and the auxiliary links should be reduced as much as possible. This technical support evaluation method is also very different from the focus of conventional technical support. Conventional technical support focuses on the quality and safety of overall support; while the technical support evaluation under emergency conditions, in addition to taking into account the support quality and safety, focuses more on the support time, hoping to accelerate the support speed so that the equipment can be quickly put into use. For the above reasons, the present invention proposes a dedicated neural network based on the emphasis on equipment support time, which is used to simulate the complex non-linear relationship among the comprehensive evaluation of rapid technical support of complex equipment, equipment support quality, equipment support time, and equipment support safety. By using the powerful parallel computing ability of the neural network, the existing historical data is fully utilized and the network error is trained to converge to the vicinity of 0, so as to substitute the data of the equipment to be evaluated into the trained neural network, and finally obtain the comprehensive evaluation result of the rapid technical support of complex equipment under emergency conditions. This method not only has high innovation in theory, but also has high engineering practical value. Summary of the Invention

[0003] The technical problem to be solved by the present invention is: a comprehensive evaluation method for rapid technical support of complex equipment under emergency conditions, so as to solve the problem of difficult modeling of the comprehensive evaluation method for the support time factor under emergency conditions in the above-mentioned prior art.

[0004] The technical solution adopted by the present invention is: a comprehensive evaluation method for rapid technical support of complex equipment under emergency conditions, the method comprising the following steps:

[0005] Step S10: Select the historical data of multiple pieces of equipment. Evaluate the emergency rapid support process of each piece of equipment in three aspects: equipment support quality, equipment support time, and equipment support safety. At the same time, decompose the equipment support quality aspect into five sub-aspects: equipment loading and unloading quality, equipment installation quality, equipment testing quality, equipment filling quality, and equipment startup preparation quality. Among them, the equipment loading and unloading quality sub-aspect is further divided into two categories: loading and unloading operation process quality and loading and unloading information collection process quality; the equipment installation quality sub-aspect is divided into two categories: installation operation process quality and installation information collection process quality; the equipment testing quality sub-aspect is divided into two categories: testing operation process quality and testing information collection process quality; the equipment filling quality sub-aspect is divided into two categories: filling operation process quality and filling information collection process quality; the equipment startup preparation quality sub-aspect is divided into two categories: preparation operation process quality and preparation information collection process quality. The equipment support time aspect is decomposed into two sub-aspects: equipment loading time and equipment technical preparation time; the equipment support safety is decomposed into four sub-aspects: personnel safety, equipment safety, facility safety, and environmental safety. Among them, the personnel safety sub-aspect is decomposed into two categories: personnel safety measures and personnel safety implementation; the equipment safety sub-aspect is decomposed into two categories: equipment safety measures and equipment safety implementation; the facility safety sub-aspect is decomposed into two categories: facility safety measures and facility safety implementation; the environmental safety sub-aspect is decomposed into two categories: environmental safety measures and environmental safety implementation.

[0006] Step S20: For the 18 categories in 9 sub-aspects of the equipment support quality and equipment support safety in the emergency rapid support process of each piece of equipment, use the method of expert scoring at this level for comprehensive scoring, denoted as a ijkw , which represents the evaluation score of the w-th category of the k-th sub-aspect of the j-th aspect of the i-th piece of equipment; then, for the 2 sub-aspects of the equipment support time aspect, use the method of expert scoring at this level for comprehensive scoring, denoted as b i3k , which represents the evaluation score of the k-th sub-aspect of the 3rd aspect of the i-th piece of equipment; at the same time, record the posterior evaluation score data of the i-th piece of equipment by the superior expert after the equipment completes the task, denoted as c i .

[0007] Step S30: Based on the evaluation score data of the \(w\)-th classification of the \(k\)-th sub-aspect of the \(j\)-th aspect of the \(i\)-th equipment in the emergency rapid support process of the local experts, a dedicated radial basis neural network based on emergency rapid support is established. First, according to the 18 classification data of the \(k\)-th sub-aspect of the \(j\)-th aspect of the \(i\)-th equipment, 100 neural network node center values are respectively selected. Then, the deviation data of the network center point is obtained by comparison. And the sensitive interval of the neural network node is set, and after absolute value linear and cubic mixed transformation and exponential transformation, the exponential transformation data of the network center point deviation is obtained. Finally, the neural network exponential weights, linear weights and cubic weights of each node are respectively superimposed. And the accumulation of 100 nodes is carried out to obtain the comprehensive output of the neural network for the \(w\)-th classification of the \(k\)-th sub-aspect of the \(j\)-th aspect of the \(i\)-th equipment.

[0008] Step S40: Classify and summarize the comprehensive output of the neural network for the \(w\)-th classification of the \(k\)-th sub-aspect of the \(j\)-th aspect of the \(i\)-th equipment to obtain the comprehensive output data of the neural network for the \(k\)-th sub-aspect of the \(j\)-th aspect of the \(i\)-th equipment, and according to the characteristics of rapid technical support of complex equipment under emergency conditions, summarize to obtain the comprehensive output data of the neural network for the \(j\)-th aspect of the \(i\)-th equipment. Finally, according to the comprehensive output of the neural network for the 3 aspects of the \(i\)-th equipment, the total output data of the \(i\)-th equipment of the neural network is obtained by superimposing the weights of the three network aspects, and compared with the posterior evaluation score data of the \(i\)-th equipment after the superior experts complete the task to obtain the network error data. Then, according to the network error data, a corresponding error adaptive weight adjustment rule is designed to perform integral adaptive iteration training on the linear and cubic mixed ratio weight coefficient, neural network exponential weight, neural network linear weight, and neural network cubic weight of the neural network. Then, according to the comprehensive output of the neural network for the 3 aspects of the \(i\)-th equipment, an error adaptive weight adjustment rule based on the emphasis on equipment support time is designed to perform integral adaptive iteration on the weights of the three network aspects until the network error data converges to the interval near 0, and the network training is stopped.

[0009] Step S50: According to the 18 classifications of 2 aspects, namely equipment support quality and equipment support safety, in the emergency rapid support process of the equipment to be evaluated by the local experts, a comprehensive score is made by the method of scoring by the local experts, denoted as \(a\) Gjkw , which represents the evaluation score of the \(w\)-th classification of the \(k\)-th sub-aspect of the \(j\)-th aspect of the equipment to be evaluated. Then, for the 2 sub-aspects of the equipment support time of the equipment to be evaluated, a comprehensive score is made by the method of scoring by the local experts, denoted as \(b\) G3k , solve the total score data of the \(k\)-th level of the \(j\)-th stage of the equipment to be evaluated, and then substitute it into the trained neural network to obtain the comprehensive evaluation score of the neural network for the emergency rapid support process of the equipment to be evaluated.

[0010] In an exemplary embodiment of the present invention, based on the evaluation score data of the i-th equipment, the j-th aspect, the k-th sub-aspect, and the w-th classification in the emergency rapid guarantee process by the local experts, establishing a dedicated radial basis neural network based on emergency rapid guarantee includes:

[0011] ε ijkwp = a ijkw - a 1jkwp ;

[0012]

[0013] where a 1jkwp is the central value of 100 neural network nodes, and ε ijkwp is the deviation data of the network center point; σ jk is the sensitive interval of the neural network node; γ ijkwp is the exponential transformation data of the network center point deviation; c 1jkwp is the linear and cubic mixed ratio weight coefficient, which is used to adjust the ratio value of the absolute value linear and cubic mixture in the exponential transformation data of the center point deviation; u ijkw is the comprehensive output of the neural network for the i-th equipment, the j-th aspect, the k-th sub-aspect, and the w-th classification, k 1jkwp is the exponential weight of the neural network, c 2jkwp is the linear weight of the neural network, c 3jkwp is the cubic weight of the neural network.

[0014] In an exemplary embodiment of the present invention, classifying and summarizing the comprehensive output of the neural network for the i-th equipment, the j-th aspect, the k-th sub-aspect, and the w-th classification to obtain the comprehensive output data of the neural network for the i-th equipment, the j-th aspect, and the k-th sub-aspect, and according to the characteristics of the rapid technical guarantee of complex equipment under emergency conditions, summarizing to obtain the comprehensive output data of the neural network for the i-th equipment and the j-th aspect; finally, adding the comprehensive output of the neural network for the three aspects of the i-th equipment and superimposing the weight values of the three network aspects to obtain the total output data of the neural network for the i-th equipment, and comparing it with the posterior evaluation score data of the i-th equipment after the superior experts complete the task to obtain the network error data including:

[0015]

[0016]

[0017] e i = f i - c i ;

[0018] where y ijk is the comprehensive output data of the neural network for the i-th equipment, the j-th aspect, and the k-th sub-aspect; oi1 , o i2 , o i3 are the comprehensive output data of the neural network for the first, second, and third aspects of the i-th equipment respectively; f i is the total output data of the neural network for the i-th equipment; e i is the network error data; k 1jkwpd is the error adaptive weight adjustment law of the exponential weight of the neural network; c 1jkwpd is the error adaptive weight adjustment law of the error of the linear and cubic mixed ratio weight coefficient of the neural network; c 2jkwpd is the error adaptive weight adjustment law of the linear weight of the neural network; c 3jkwpd is the error adaptive weight adjustment law of the cubic weight of the neural network.

[0019] In an exemplary embodiment of the present invention, according to the network error data, the corresponding error adaptive weight adjustment law is designed to perform integral adaptive iterative training on the linear and cubic mixed ratio weight coefficient, the exponential weight, the linear weight, and the cubic weight of the neural network; then, according to the comprehensive output of the neural network for 3 aspects of the i-th equipment, the error adaptive weight adjustment law based on the emphasis on equipment support time is designed, and the integral adaptive iteration of the weights of the three network aspects includes:

[0020]

[0021]

[0022] k 1jkwp (n + 1)=k 1jkwp (n)+k 1jkwpd ;

[0023] k 21 (n + 1)=k 21 (n)+k 21d ;

[0024] k 22 (n + 1)=k 22 (n)+k 22d ;

[0025] k 23 (n + 1)=k 23 (n)+k 23d ;

[0026] c 1jkwp (n + 1)=c 1jkwp (n)+c 1jkwpd ;

[0027] c 2jkwp (n + 1)=c 2jkwp(n) + c 2jkwpd ;

[0028] c 3jkwp (n + 1) = c 3jkwp (n) + c 3jkwpd ;

[0029] where k 21d is the error adaptive weight adjustment rule for the first network aspect weight; k 22d is the error adaptive weight adjustment rule for the second network aspect weight; k 23d is the error adaptive weight adjustment rule for the third network aspect weight; l1, l2, l3, l4, l5, l6, l7, l8, l9, l 10 、l 11 、l 12 、l 13 、l 14 are constant parameters used to adjust the speed of neural network weight convergence; k 21 、k 22 、k 23 are the weights of the first, second, and third network aspects respectively.

[0030] In an exemplary embodiment of the present invention, according to the 18 classifications of 9 sub - aspects in 2 aspects of equipment support quality and equipment support safety in the emergency rapid support process of the equipment to be evaluated by the local - level experts, the comprehensive scoring is carried out by the method of local - level expert scoring; then, for the 2 sub - aspects in the equipment support time aspect of the equipment to be evaluated, the comprehensive scoring is carried out by the method of local - level expert scoring; and then substituting into the trained neural network, the comprehensive evaluation score of the neural network for the emergency rapid support process of the equipment to be evaluated includes:

[0031] ε Gjkwp = a Gjkw - a 1jkwp ;

[0032]

[0033]

[0034] where a Gjkw represents the evaluation score of the w - th classification of the k - th sub - aspect of the j - th aspect of the equipment to be evaluated; b G3k is the total score data of the k - th level of the j - th stage of the equipment to be evaluated; ε Gjkwp is the network center point deviation data of the equipment to be evaluated; γ Gjkwp is the network center point deviation exponential transformation data of the equipment to be evaluated; u Gjkw is the comprehensive output of the neural network for the w - th classification of the k - th sub - aspect of the j - th aspect of the equipment to be evaluated, y Gjkis the comprehensive output data of the neural network for the k-th sub-aspect of the j-th aspect of the equipment to be evaluated; o G1 and o G2 and o G3 are the comprehensive output data of the neural network for the first, second, and third aspects of the equipment to be evaluated, respectively; f G is the comprehensive evaluation score of the neural network for the emergency rapid support process of the equipment to be evaluated.

[0035] Advantages of the present invention

[0036] 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 equipment under emergency conditions into 18 classifications in 9 sub-aspects of 2 aspects, namely equipment support quality and equipment support safety, and establish a support evaluation subdivision model suitable for the background of emergency conditions. The second is to establish a dedicated neural network matching the rapid technical support of equipment under emergency conditions, enabling it to better fit the background and actual situation of the rapid technical support of equipment, thereby accelerating the convergence of the network and improving its simulation accuracy. The third is to propose a weight training rule based on the emphasis on equipment support time in the neural network architecture, especially in the way of weight convergence, so as to realize the characteristic that the comprehensive support evaluation prefers shorter support time and faster support speed under emergency conditions. Description of the drawings

[0037] The drawings here are incorporated into the specification and form a part of this specification, showing the 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.

[0038] Figure 1 is a flowchart of a comprehensive evaluation method for rapid technical support of complex equipment under emergency conditions;

[0039] Figure 2 is a structure diagram of 18 classifications in three aspects and eleven sub-aspects of the method provided by the embodiment of the present invention;

[0040] Figure 3 is a diagram showing the convergence of the neural network training error of the method provided by the embodiment of the present invention;

[0041] Figure 4 is a diagram showing the weight convergence of the first network aspect of the method provided by the embodiment of the present invention;

[0042] Figure 5 is a diagram showing the weight convergence of the second network aspect of the method provided by the embodiment of the present invention;

[0043] Figure 6 It is the third network aspect weight convergence situation diagram of the method provided by the embodiments of the present invention. Specific implementation manners

[0044] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the following will further elaborate on the present invention in detail in conjunction with specific embodiments and with reference to the appended Figure 1 drawings.

[0045] Embodiment 1: Taking the rapid technical support process of 80 pieces of equipment stored in a warehouse under emergency conditions as an example, the comprehensive evaluation data of 18 classifications in 9 sub-aspects in 2 aspects of equipment support quality and equipment support safety, and 2 sub-aspects in terms of equipment support time during the emergency rapid support process of 80 pieces of equipment by local experts, as well as the posterior evaluation score data of the equipment after completing the task by superior experts are used to train the established neural network; finally, taking the comprehensive evaluation data of 18 classifications of the equipment to be evaluated by local experts and 2 sub-aspects in terms of equipment support time as a background case, a comprehensive evaluation method for the rapid technical support of complex equipment under emergency conditions is described. The method includes the following steps:

[0046] Step S10, select the historical data of 80 pieces of equipment. As shown in the appended Figure 2 drawings, the evaluation of the emergency rapid support process of each piece of equipment is divided into three aspects: equipment support quality, equipment support time, and equipment support safety; at the same time, the equipment support quality aspect is decomposed into five sub-aspects: equipment loading and unloading quality, equipment installation quality, equipment testing quality, equipment filling quality, and equipment start-up preparation quality; among them, the equipment loading and unloading quality sub-aspect is further divided into two classifications: loading and unloading operation process quality and loading and unloading information collection process quality; the equipment installation quality sub-aspect is divided into two classifications: installation operation process quality and installation information collection process quality; the equipment testing quality sub-aspect is divided into two classifications: testing operation process quality and testing information collection process quality; the equipment filling quality sub-aspect is divided into two classifications: filling operation process quality and filling information collection process quality; the equipment start-up preparation quality sub-aspect is divided into two classifications: preparation operation process quality and preparation information collection process quality; the equipment support time aspect is decomposed into two sub-aspects: equipment loading time and equipment technical preparation time; the equipment support safety is decomposed into four sub-aspects: personnel safety, equipment safety, facility safety, and environmental safety; among them, the personnel safety sub-aspect is decomposed into two classifications: personnel safety measures and personnel safety implementation; the equipment safety sub-aspect is decomposed into two classifications: equipment safety measures and equipment safety implementation; the facility safety sub-aspect is decomposed into two classifications: facility safety measures and facility safety implementation; the environmental safety sub-aspect is decomposed into two classifications: environmental safety measures and environmental safety implementation.

[0047] Step S20: For the 18 classifications in 9 sub - aspects of the equipment support quality and equipment support safety in the emergency rapid support process of each piece of equipment, a comprehensive score is obtained by means of expert scoring at this level, denoted as a ijkw , which represents the evaluation score of the w - th classification of the k - th sub - aspect of the j - th aspect of the i - th piece of equipment; then, for the 2 sub - aspects of the equipment support time, a comprehensive score is obtained by means of expert scoring at this level, denoted as b i3k , which represents the evaluation score of the k - th sub - aspect of the 3 - rd aspect of the i - th piece of equipment; meanwhile, record the posterior evaluation score data of the i - th piece of equipment after the superior experts complete the task, denoted as c i , where i = 1, 2, …, 80.

[0048] Step S30: Based on the evaluation score data of the w - th classification of the k - th sub - aspect of the j - th aspect of the i - th piece of equipment in the emergency rapid support process by experts at this level, a dedicated radial basis function neural network based on emergency rapid support is established. Specifically, it can be decomposed into the following three small steps. First step: First, according to the 18 classification data of the k - th sub - aspect of the j - th aspect of the i - th piece of equipment, 100 neural network node center values are respectively selected; then, the deviation data of the network center point is obtained by comparing with them as follows:

[0049] ε ijkwp = a ijkw - a 1jkwp ;

[0050] where a 1jkwp is the 100 neural network node center values, which are constant parameters, and ε ijkwp is the deviation data of the network center point.

[0051] Second step: Set the sensitive interval of the neural network nodes, and after absolute value linear and cubic mixed transformation and exponential transformation, the exponential transformation data of the network center point deviation is obtained as follows:

[0052]

[0053] where σ jk is the sensitive interval of the neural network nodes, which are constant parameters; γ ijkwp is the exponential transformation data of the network center point deviation; c 1jkwp is the linear and cubic mixed ratio weight coefficient, which is used to adjust the ratio of absolute value linear and cubic mixing in the exponential transformation data of the center point deviation.

[0054] Step 3: Multiply the data transformed by the network center point deviation index, the network center point deviation data, and the cube of the network center point deviation data by the neural network exponential weight, linear weight, and cubic weight of each node respectively, and then perform superposition; and accumulate for 100 nodes to obtain the comprehensive output of the neural network for the w-th classification of the k-th sub-aspect of the j-th aspect of the i-th equipment as follows:

[0055]

[0056] where u ijkw is the comprehensive output of the neural network for the w-th classification of the k-th sub-aspect of the j-th aspect of the i-th equipment, k 1jkwp is the neural network exponential weight, c 2jkwp is the neural network linear weight, c 3jkwp is the neural network cubic weight.

[0057] Step S40 can be specifically decomposed into the following nine sub-steps. Step 1: Classify and summarize according to the comprehensive output of the neural network for the w-th classification of the k-th sub-aspect of the j-th aspect of the i-th equipment to obtain the comprehensive output data of the neural network for the k-th sub-aspect of the j-th aspect of the i-th equipment as follows:

[0058]

[0059] where y ijk is the comprehensive output data of the neural network for the k-th sub-aspect of the j-th aspect of the i-th equipment.

[0060] Step 2: According to the characteristics of rapid technical support for complex equipment under emergency conditions, summarize the comprehensive output data of k sub-aspects to obtain the comprehensive output data of the neural network for the j-th aspect of the i-th equipment as follows:

[0061]

[0062] where o i1 、o i2 、o i3 are the comprehensive output data of the neural network for the first, second, and third aspects of the i-th equipment respectively.

[0063] Step 3: Finally, according to the comprehensive output of the neural network for 3 aspects of the i-th equipment, superimpose the weights of the three network aspects to obtain the total output data of the neural network for the i-th equipment as follows:

[0064]

[0065] where f i is the total output data of the neural network for the i-th equipment.

[0066] Step 4: Compare the total output data of the i-th equipment with the posterior evaluation score data of the i-th equipment after the superior expert completes the task to obtain the network error data as follows:

[0067] e i =f i -c i ;

[0068] where e i is the network error data; its convergence situation is as shown in the appendix Figure 3 .

[0069] Step 5: Design the corresponding error adaptive weight adjustment rule according to the network error data as follows:

[0070]

[0071]

[0072] where k 1jkwpd is the error adaptive weight adjustment rule of the exponential weight of the neural network; c 1jkwpd is the error adaptive weight adjustment rule of the error of the linear and cubic mixed ratio weight coefficient of the neural network; c 2jkwpd is the error adaptive weight adjustment rule of the linear weight of the neural network; c 3jkwpd is the error adaptive weight adjustment rule of the cubic weight of the neural network. Among them, l1, l2, l3, l4, l5, l6, l7, l8 are constant parameters used to adjust the speed of the neural network weight convergence, and their parameters are all selected as 0.002 in this case.

[0073] Step 6: Perform integral adaptive iterative training on the linear and cubic mixed ratio weight coefficient, exponential weight, linear weight, and cubic weight of the neural network as follows:

[0074] k 1jkwp (n + 1) = k 1jkwp (n) + k 1jkwpd ;

[0075] c 1jkwp (n + 1) = c 1jkwp (n) + c 1jkwpd ;

[0076] c 2jkwp (n + 1) = c 2jkwp (n) + c 2jkwpd ;

[0077] c 3jkwp (n + 1) = c 3jkwp (n) + c 3jkwpd ;

[0078] where k 1jkwp is the exponential weight of the neural network, c 2jkwp is the linear weight of the neural network, c 3jkwp is the cubic weight of the neural network, c 1jkwp is the weight coefficient for the linear and cubic mixed ratio.

[0079] Step 7: Design an error adaptive weight adjustment law based on the equipment support time preference according to the comprehensive output of the neural network for the three aspects of the i-th equipment as follows:

[0080]

[0081] where k 21d is the error adaptive weight adjustment law for the weight of the first network aspect; k 22d is the error adaptive weight adjustment law for the weight of the second network aspect; k 23d is the error adaptive weight adjustment law for the weight of the third network aspect. Among them, l9, l 10 , l 11 , l 12 , l 13 , l 14 are constant parameters used to adjust the speed of the neural network weight convergence. Their parameters are all selected as 0.005 in this case.

[0082] Step 8: Perform integral adaptive iteration on the weights of the three network aspects as follows:

[0083] k 21 (n + 1) = k 21 (n) + k 21d ;

[0084] k 22 (n + 1) = k 22 (n) + k 22d ;

[0085] k 23 (n + 1) = k 23 (n) + k 23d ;

[0086] where k 21 , k 22 , k 23 are the weights of the first, second, and third network aspects respectively. The convergence of their weights is shown in Appendix Figure 4 , Appendix Figure 5 , Appendix Figure 6 respectively.

[0087] Step 9: Stop the network training until the network error data converges to the interval near 0. ByFigures 3 - 6 It can be seen that although the error in the training process of the entire neural network fluctuates, the overall trend is stable convergence, and the weight training process is also unidirectionally stable convergence. Therefore, the network exhibits excellent convergence effect and reliability, without the unstable phenomenon of network divergence and collapse.

[0088] Step S50 can be specifically decomposed into the following three sub-steps. First, according to the 18 classifications in 9 sub-aspects of 2 aspects, namely equipment support quality and equipment support safety, of the emergency rapid support process of the equipment to be evaluated by the experts at this level, comprehensive scoring is carried out by means of scoring by the experts at this level, denoted as a Gjkw , which represents the evaluation score of the w-th classification of the k-th sub-aspect of the j-th aspect of the equipment to be evaluated. Its values are as follows:

[0089] [(85, 87; 92, 78; 84, 89; 82, 77; 81, 92), (82, 88; 79, 73; 82, 91; 93, 84)]; then, for the 2 sub-aspects of the equipment support time of the equipment to be evaluated, comprehensive scoring is carried out by means of scoring by the experts at this level, denoted as b G3k , and its values are as follows: [82, 89].

[0090] Second, according to the above scores, substitute them into the trained neural network to solve the total score data of the k-th layer of the j-th stage of the equipment to be evaluated by the neural network as follows:

[0091] ε Gjkwp = a Gjkw - a 1jkwp ;

[0092]

[0093] where ε Gjkwp is the network center point deviation data of the equipment to be evaluated; γ Gjkwp is the network center point deviation exponential transformation data of the equipment to be evaluated; u Gjkw is the comprehensive output of the neural network for the w-th classification of the k-th sub-aspect of the j-th aspect of the equipment to be evaluated.

[0094] Third, according to the weights of the trained neural network, further calculate to obtain the comprehensive evaluation score of the emergency rapid support process of the equipment to be evaluated by the neural network as follows:

[0095]

[0096] where y Gjk is the comprehensive output data of the neural network for the k-th sub-aspect of the j-th aspect of the equipment to be evaluated; o G1 , o G2 , o G3They are the comprehensive output data of the neural network for the first, second, and third aspects of the equipment to be evaluated; f G is the comprehensive evaluation score of the neural network for the emergency rapid support process of the equipment to be evaluated. The finally obtained f G = 87.324, so the conclusion is that the comprehensive evaluation of the emergency rapid support process of the equipment to be evaluated is good.

[0097] The specific embodiments described above have further elaborated on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is only for the specific embodiments of the present invention and is not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A comprehensive evaluation method for rapid technical support of complex equipment under emergency conditions, characterized by the following steps: Step S10, select historical data of multiple pieces of equipment, and divide the evaluation of the emergency rapid support process of each piece of equipment into three aspects: equipment support quality, equipment support time, and equipment support safety; at the same time, divide the equipment support quality aspect into five sub-aspects: equipment loading and unloading quality, equipment installation quality, equipment testing quality, equipment filling quality, and equipment start-up preparation quality; and among them, the equipment loading and unloading quality sub-aspect is further divided into two categories: loading and unloading operation process quality and loading and unloading information collection process quality; the equipment installation quality sub-aspect is divided into two categories: installation operation process quality and installation information collection process quality; the equipment testing quality sub-aspect is divided into two categories: testing operation process quality and testing information collection process quality; the equipment filling quality sub-aspect is divided into two categories: filling operation process quality and filling information collection process quality; the equipment start-up preparation quality sub-aspect is divided into two categories: preparation operation process quality and preparation information collection process quality; the equipment support time aspect is divided into two sub-aspects: equipment loading time and equipment technical preparation time; the equipment support safety is decomposed into four sub-aspects: personnel safety, equipment safety, facility safety, and environmental safety; among them, the personnel safety sub-aspect is decomposed into two categories: personnel safety measures and personnel safety implementation; the equipment safety sub-aspect is decomposed into two categories: equipment safety measures and equipment safety implementation; the facility safety sub-aspect is decomposed into two categories: facility safety measures and facility safety implementation; the environmental safety sub-aspect is decomposed into two categories: environmental safety measures and environmental safety implementation; Step S20, for each of the 18 categories in 9 sub - aspects of the equipment support quality and equipment support safety in the emergency rapid support process of each piece of equipment, a comprehensive score is obtained by means of expert scoring at this level, denoted as a ijkw , which represents the evaluation score of the w - th category of the k - th sub - aspect of the j - th aspect of the i - th piece of equipment; then, for the 2 sub - aspects of the equipment support time, a comprehensive score is obtained by means of expert scoring at this level, denoted as b i3k , which represents the evaluation score of the k - th sub - aspect of the 3 - rd aspect of the i - th piece of equipment; at the same time, record the posterior evaluation score data of the i - th piece of equipment after the superior experts complete the task of the equipment, denoted as c i ; Step S30, based on the evaluation score data of the i-th piece of equipment, the j-th aspect, the k-th sub-aspect, and the w-th classification in the emergency rapid support process by experts at this level, establish a dedicated radial basis neural network based on emergency rapid support; first, according to the 18 classification data of the i-th piece of equipment, the j-th aspect, and the k-th sub-aspect, respectively select 100 neural network node center values; then compare with them to obtain the network center point deviation data; and set the neural network node sensitive interval, and perform absolute value linear and cubic mixed transformation and exponential transformation to obtain the network center point deviation exponential transformation data; finally, respectively superimpose the neural network exponential weight, linear weight, and cubic weight of each node; and accumulate the 100 nodes to obtain the comprehensive output of the neural network for the i-th piece of equipment, the j-th aspect, the k-th sub-aspect, and the w-th classification as follows: ε ijkwp = a ijkw -a 1jkwp ; where a 1jkwp is the central value of 100 neural network nodes, and ε ijkwp is the deviation data of the network center point; σ jk is the sensitive interval of the neural network nodes; γ ijkwp is the deviation exponential transformation data of the network center point; c 1jkwp is the linear and cubic mixing ratio weight coefficient, which is used to adjust the ratio of the absolute value linear and cubic mixing in the deviation exponential transformation data of the center point; u ijkw is the comprehensive output of the neural network for the w-th classification of the k-th sub-aspect of the j-th aspect of the i-th equipment, k 1jkwp is the exponential weight of the neural network, c 2jkwp is the linear weight of the neural network, c 3jkwp is the cubic weight of the neural network; Step S40: Classify and summarize the comprehensive output of the w-th classification of the k-th sub-aspect of the j-th aspect of the i-th equipment according to the neural network, obtain the comprehensive output data of the k-th sub-aspect of the j-th aspect of the i-th equipment by the neural network, and summarize according to the characteristics of rapid technical support for complex equipment under emergency conditions to obtain the comprehensive output data of the j-th aspect of the i-th equipment by the neural network; finally, superimpose the weight values of the three network aspects on the comprehensive output of the three aspects of the i-th equipment by the neural network to obtain the total output data of the i-th equipment by the neural network, and compare it with the posterior evaluation score data of the i-th equipment after the superior experts complete the task to obtain the network error data; then design the corresponding error adaptive weight adjustment rule according to the network error data, and perform integral adaptive iteration training on the linear and cubic mixed ratio weight coefficients, neural network exponential weights, neural network linear weights, and neural network cubic weights of the neural network; then design the error adaptive weight adjustment rule based on the emphasis on equipment support time according to the comprehensive output of the three aspects of the i-th equipment by the neural network, and perform integral adaptive iteration on the weight values of the three network aspects until the network error data converges to the interval near 0, and stop the network training as follows: e i = f i - c i ; k 1jkwp (n + 1)= k 1jkwp (n)+ k 1jkwpd ; k 21 (n + 1)= k 21 (n)+ k 21d ; k 22 (n + 1)= k 22 (n)+ k 22d ; k 23 (n + 1)= k 23 (n)+ k 23d ; c 1jkwp (n + 1)= c 1jkwp (n)+ c 1jkwpd ; c 2jkwp (n + 1) = c 2jkwp (n) + c 2jkwpd ; c 3jkwp (n + 1) = c 3jkwp (n) + c 3jkwpd ; where y ijk is the comprehensive output data of the neural network for the k-th sub-aspect of the j-th aspect of the i-th equipment; o i1 、o i2 、o i3 are respectively the comprehensive output data of the neural network for the first, second, and third aspects of the i-th equipment; f i is the total output data of the neural network for the i-th equipment; e i is the network error data; k 1jkwpd is the error adaptive weight adjustment law of the exponential weight of the neural network; c 1jkwpd is the error adaptive weight adjustment law of the error of the linear and cubic mixed ratio weight coefficient of the neural network; c 2jkwpd is the error adaptive weight adjustment law of the linear weight of the neural network; c 3jkwpd is the error adaptive weight adjustment law of the cubic weight of the neural network; k 21d is the error adaptive weight adjustment law of the weight of the first network aspect; k 22d is the error adaptive weight adjustment law of the weight of the second network aspect; k 23d is the error adaptive weight adjustment law of the weight of the third network aspect; l1, l2, l3, l4, l5, l6, l7, l8, l9, l 10 、l 11 、l 12 、l 13 、l 14 are constant parameters used to adjust the speed of weight convergence of the neural network; k 21 、k 22 、k 23 are respectively the weights of the first, second, and third network aspects; Step S50: According to the 18 classifications in 9 sub - aspects of 2 aspects, namely equipment support quality and equipment support safety, of the emergency rapid support process of the equipment to be evaluated by experts at this level, a comprehensive score is obtained by means of scoring by experts at this level, denoted as a Gjkw , which represents the evaluation score of the w - th classification of the k - th sub - aspect of the j - th aspect of the equipment to be evaluated; then, for the 2 sub - aspects of the equipment support time of the equipment to be evaluated, a comprehensive score is obtained by means of scoring by experts at this level, denoted as b G3k , solve the total score data of the k - th layer of the j - th stage of the equipment to be evaluated by the neural network, and then substitute it into the trained neural network to obtain the comprehensive evaluation score of the emergency rapid support process of the equipment to be evaluated by the neural network as follows: ε Gjkwp = a Gjkw -a 1jkwp ; where ε Gjkwp is the network center point deviation data of the equipment to be evaluated; γ Gjkwp is the network center point deviation index transformation data of the equipment to be evaluated; u Gjkw is the comprehensive output of the neural network for the w-th classification of the k-th sub-aspect of the j-th aspect of the equipment to be evaluated, y Gjk is the comprehensive output data of the neural network for the k-th sub-aspect of the j-th aspect of the equipment to be evaluated; o G1 、o G2 、o G3 are the comprehensive output data of the neural network for the first, second, and third aspects of the equipment to be evaluated respectively; f G is the comprehensive evaluation score of the neural network for the emergency rapid support process of the equipment to be evaluated.

Images