Multi-valued system diagnosis strategy optimization method considering test uncertainty

Abstract

The invention discloses a multi-valued system diagnosis strategy optimization method considering test uncertainty, and relates to the technical field of spacecraft testability design and fault diagnosis. The method solves the problems that in existing aerospace multi-valued system diagnosis, test uncertainty is not considered, multi-target optimization is missing, and the global search capability is insufficient, and the method comprises the steps that based on a fault test correlation matrix, a multi-target optimization model fusing average test cost, test reliability and a fault isolation rate as optimization indexes is established; a multi-valued information entropy heuristic function is introduced to improve the local search efficiency of the multi-target optimization model; on the basis of adopting a discrete particle swarm algorithm and designing a linear decline inertia weight, realizing global optimal search of a test sequence; three indexes including the average test cost, the reliability and the fault isolation rate are optimized cooperatively through the fitness function, and a diagnosis strategy meeting engineering requirements is output. And the method is also suitable for a multi-valued test scene with test uncertainty in an aerospace complex system.

Description

Technical Field

[0001] This invention relates to the field of spacecraft testability design and fault diagnosis technology, and specifically to an optimization method for multi-valued system diagnostic strategies that takes into account test uncertainties. Background Technology

[0002] Against the backdrop of rapid development in aerospace technology, the functions and structures of complex equipment such as spacecraft are becoming increasingly complex, placing stringent demands on fault diagnosis capabilities for their safe operation. Design for testability, as a core means to improve system testing and diagnosis efficiency and reduce life-cycle costs, has become a crucial link in equipment quality characterization. Diagnostic strategy optimization, as a core element of design for testability, requires selecting optimal test points and constructing optimal test sequences to determine the system fault state with minimal time and cost; this problem has been proven to be NP-complete.

[0003] In the field of multi-valued system diagnostic strategy optimization, existing technologies, in order to simplify research complexity, often default the system to a binary output mode of "pass / fail," losing a large amount of critical fault-test correlation information in multi-valued testing. Moreover, they generally ignore uncertainties such as missed detections and false alarms that exist in actual testing. The optimization objective often focuses only on the average test cost, without incorporating test reliability and fault isolation rate into the core optimization scope. For example, algorithms such as Rollout and single information entropy heuristic algorithms attempt to adapt to multi-valued scenarios, but they have defects such as either failing to handle test uncertainties or failing to achieve multi-objective collaborative optimization. When traditional particle swarm optimization algorithms are directly applied, they are also prone to getting stuck in local optima and are difficult to adapt to the selection requirements of discrete test sequences.

[0004] In the field of multi-valued system diagnostic strategy optimization, existing technologies have significant limitations: most algorithms, in order to reduce research complexity, default to a simplified binary system model of "pass / fail," resulting in the loss of a large amount of fault-test correlation information in multi-valued testing, which cannot adapt to the actual testing scenarios of complex aerospace equipment. At the same time, they generally ignore the uncertainty problems such as missed detections and false alarms in the testing process. Even some studies on multi-valued testing, such as information entropy optimization methods, Rollout algorithms, and ANT-TS algorithms, only focus on fault detection accuracy without incorporating false alarm rates, and do not take into account testing uncertainties. Moreover, the optimization objectives are mostly limited to average test costs, failing to achieve multi-objective collaborative optimization of test costs, reliability, and fault isolation rates. Although the multi-strategy fusion whale optimization algorithm proposed by Zhao et al. improves optimization performance, its adaptability to multi-valued systems and global search capabilities are still inferior to the swarm intelligence characteristics of particle swarm optimization. When traditional particle swarm optimization is directly applied, it is also prone to getting trapped in local optima, has difficulty handling discrete test sequence selection, and does not consider multi-valued information gain. Summary of the Invention

[0005] To overcome the problems in existing technologies regarding the diagnosis of multi-valued aerospace systems, such as the lack of consideration for testing uncertainties, the absence of multi-objective optimization, and insufficient global search capabilities, this invention proposes a new particle swarm optimization algorithm, MUPSOIG, which integrates the traditional particle swarm optimization algorithm and the information entropy algorithm, achieving a trade-off between expected testing cost and testing reliability.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: Option 1: An optimization method for diagnostic strategies of multi-valued systems considering testing uncertainty, the method being implemented based on the particle swarm optimization information entropy algorithm, and comprising the following steps: Step 1: Establish the fused average test cost based on the fault test correlation matrix. J Test reliability R, fault isolation rate A multi-objective optimization model for optimizing indicators; Step 2: Introduce a multi-valued information entropy heuristic function to improve the local search efficiency of the multi-objective optimization model established in Step 1; Step 3: Based on Step 2, the discrete particle swarm optimization algorithm is used and a linearly decreasing inertia weight is designed to achieve the global optimal search of the test sequence; Step 4: Co-optimize three key metrics using a fitness function. These three metrics include average test cost J, test reliability R, and fault isolation rate. It outputs diagnostic strategies that meet engineering requirements.

[0007] Furthermore, a preferred embodiment is provided, wherein the particle swarm information entropy algorithm includes the following steps: Step 1.1: Initialize the parameters, including learning factors c1 and c2, weight coefficients a, b, and c, inertia weight coefficient ω, the number of tests randomly selected in each iteration l, the population size N, the population size d, and the maximum number of iterations iter. Step 1.2: Initialize the population. Use binary encoding for the test set and randomly generate a population of size d, that is, the elements in the position vector are randomly initialized to 0 or 1, and the velocities of each particle are randomly generated. Step 1.3: The information entropy algorithm is used to calculate the fitness function. The test set corresponding to each particle position vector is used to construct a fault tree using the information entropy algorithm. The fitness values ​​are calculated and compared. The best position of the particle and the corresponding fitness value are stored in a local best position set pbest. The best particle position corresponding to the smallest fitness value among all particles is saved to the global best position gbest. Step 1.4: Population iterative update. The population is updated using BPSO with a linearly decreasing inertia weight factor to update the position and velocity of each particle. Step 1.5, repeat steps 1.3 and 1.4 until the maximum number of iterations is reached and iter or gbest remains unchanged across multiple iterations.

[0008] Furthermore, a preferred embodiment is provided, wherein the multi-valued system includes describing the multi-valued test diagnostic strategy problem of complex systems using a quintuple test set:

[0009] in Represents the finite set of system states of the target system. This represents the system state when only the i-th failure occurs. Represents the prior probability distribution set of various diagnostic conclusions for the target system; express Failure probability, This represents the test set available to the system. This indicates the j-th available test, with a result that has states other than pass and fail; This represents the testing resources consumed in each test, including factors such as test time, cost, and test complexity. The correlation matrix represents the relationship between the fault set and the test set; This represents the logical relationship between the i-th test and the m fault modes. In multi-valued tests, the elements of the correlation matrix have values ​​other than 0 and 1.

[0010] Furthermore, a preferred embodiment is provided, wherein the average testing cost described in step 1 is... J Used to isolate faults The testing costs and time generated by the test sequence; the calculation method is as follows:

[0011] in, The representative, in the current diagnostic strategy, isolating to The test sequence experienced by the fault. Representative sequence The Middle k The testing cost of each test.

[0012] Furthermore, a preferred embodiment is provided, in which the reliability R in step 1 is calculated through two levels: single-test reliability and sequence reliability. The calculation method is as follows: The formula for calculating the reliability of a single test is:

[0013] in, For testing For the fault The accuracy of the test For testing The false alarm rate, when hour, , ;when hour, , ; The reliability calculation formula for the test sequences obtained by the diagnostic strategy is as follows:

[0014] in, Representative sequence The reliability of the k-th test.

[0015] Furthermore, a preferred embodiment is provided, wherein the fault isolation rate in step 1 is... Used to determine fault sets Testable sequences The conditions for isolation are:

[0016] in, Representative fault and Corresponding test sequence The row vectors of the relevant matrices are not the same; test sequence The formula for calculating the fault isolation rate after fault isolation is as follows:

[0017] Furthermore, a preferred embodiment is provided, wherein the specific steps of the multi-valued information entropy heuristic function in step 2 are as follows: assuming the set of fault states of the system under test is S and the set of test sequences is T, calculate the unit information gain for each test point according to the information entropy function, and compare and select the test point corresponding to the maximum unit information gain value. Select this test point Test and isolate the fuzzy set of system faults.

[0018] Furthermore, a preferred embodiment is provided, wherein the fitness function in step 4 is a fitness function designed by combining the minimization of test cost, the maximization of reliability, and the maximization of fault isolation rate, i.e.,

[0019] in, , and These represent the expected test cost, reliability, and fault isolation rate of the fault diagnosis tree obtained by using the information entropy heuristic algorithm IG from the test subsets corresponding to the individual particles in the particle swarm.

[0020] Option 2: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in Option 1.

[0021] Option 3: A computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the method described in Option 1.

[0022] The advantages of this invention are: This invention presents a diagnostic strategy optimization method for multi-valued systems that considers testing uncertainty. Based on traditional particle swarm optimization (PSO) and information entropy algorithms, a novel PSO algorithm, MUPSOIG, is proposed. This algorithm integrates PSO and information entropy to achieve a trade-off between expected testing cost and testing reliability. This method better considers uncertain testing costs than the multi-valued Rollout algorithm and offers comprehensive optimization advantages in multi-valued testing uncertainty scenarios, making it significant for practical engineering applications.

[0023] The multi-valued system diagnostic strategy optimization method described in this invention, which considers test uncertainty, integrates and optimizes three core indicators: "average test cost, test reliability, and fault isolation rate," thereby overcoming the single-objective limitations of traditional algorithms and meeting the comprehensive requirements of complex systems such as aerospace for "low cost, high reliability, and high isolation."

[0024] The present invention provides a multi-valued system diagnostic strategy optimization method that considers test uncertainty. By correcting the test reliability calculation with detection accuracy Pd and false alarm rate Ps, it effectively reduces the deviation of missed detections or false alarms from the diagnostic results and improves the practicality of strategy engineering. The present invention provides a diagnostic strategy optimization method for multi-valued systems that considers test uncertainty. By integrating the information entropy heuristic function with the linearly decreasing inertia weight BPSO, it avoids the particle swarm algorithm from getting stuck in local optima and accelerates convergence through heuristic information. It has a lower test cost considering reliability than the traditional Rollout algorithm and IG algorithm. The present invention provides an optimization method for diagnostic strategies of multi-valued systems that considers test uncertainties. Based on multi-valued fault-test correlation matrices and multi-valued information entropy, it fully preserves the fault correlation information of multi-valued tests. It has achieved a 100% fault isolation rate in the case of aircraft engine systems, and its versatility has been verified by 20-100 dimensional random correlation matrices. It has better engineering adaptability for test design applications of complex systems such as spacecraft and aircraft.

[0025] This invention is also applicable to multi-value testing scenarios in complex aerospace systems where testing uncertainties exist (missed detections, false alarms). Attached Figure Description

[0026] Figure 1 This is a flowchart of the particle swarm information entropy algorithm that takes into account test uncertainties, as described in Implementation Method 1.

[0027] Figure 2 This is a schematic diagram of the fault tree obtained after the test optimization results in an embodiment of the present invention.

[0028] Figure 3 This is a schematic diagram comparing the average cost of three algorithms in a random simulation experiment according to the embodiments of the present invention, considering reliability.

[0029] Figure 4 This is a schematic diagram comparing the running times of three algorithms in a random simulation experiment in this embodiment of the invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0031] Implementation Method 1: This implementation method proposes an optimization method for diagnostic strategies of multi-valued systems that considers testing uncertainties. The method specifically includes the following steps: This invention proposes an information-inspired improved particle swarm optimization algorithm (MUPSOIG): Based on the fault-test correlation matrix, a multi-objective optimization model integrating "average test cost, test reliability, and fault isolation rate" is established; a multi-valued information gain heuristic function (based on information entropy) is introduced to improve local search efficiency; a discrete particle swarm optimization (BPSO) algorithm is adopted and a linearly decreasing inertia weight is designed to achieve global optimal search of the test sequence; finally, the three major indicators are optimized through fitness function to output a diagnostic strategy that meets engineering requirements.

[0032] This invention involves the mathematical foundations of multi-valued test diagnosis problems. Addressing the complex systems and the polymorphic mapping relationships between fault symptom characteristics and measured parameters, the multi-valued test diagnosis strategy problem for complex systems is described using a quintuple test set.

[0033] in Represents the finite set of system states of the target system. This represents the system state when only the i-th failure occurs. Represents the prior probability distribution set of various diagnostic conclusions for the target system; express Failure probability, This represents the test set available to the system. This indicates the j-th available test, with a result that has states other than pass and fail; This represents the testing resources consumed in each test, including factors such as test time, cost, and test complexity. The correlation matrix represents the relationship between the fault set and the test set; This represents the logical relationship between the i-th test and the m fault modes. In multi-valued tests, the elements of the correlation matrix have values ​​other than 0 and 1.

[0034] This invention introduces a custom performance evaluation algorithm that considers the cost of reliability testing. The cost of reliability testing is shown in the following formula:

[0035] In the formula This represents the average test cost of the test sequence. To test the reliability of the sequence, the higher the reliability, the lower the testing cost, and the lower the corresponding testing cost considering unreliability, which means the better the optimization effect.

[0036] The specific implementation process includes: I. (1) Optimize the indicator model Starting with the costs of testing cost, testing reliability, and fault isolation rate, a mathematical model is established for the optimization problem of diagnostic strategies under testing uncertainty. Three optimization indices are designed as follows: Average test cost J Average test cost is used to isolate faults in test sequence optimization. The test sequence optimization aims to minimize the average test cost and time incurred by the test sequence. The main objective of test sequence optimization is to find a subset of tests from the original test set (which satisfies the testability parameters) that minimizes the average test cost while still meeting the testability parameters.

[0037]

[0038] The representative, in the current diagnostic strategy, isolating to The test sequence experienced by the fault. Representative sequence The test cost of the k-th test.

[0039] (2) Reliability R In complex real-world systems, test results are often uncertain due to unreliable factors. This can lead to missed faults and misdiagnosed components, resulting in a compounding effect of missed and misdiagnosed errors, ultimately causing a bias in the final diagnosis. Therefore, it is crucial to maximize the R-value to improve test reliability and reduce the unreliability of diagnostic reasoning. Considering both missed and false alarms, a two-level calculation of single-test reliability and sequence reliability is employed.

[0040] The formula for calculating the reliability of a single test is:

[0041] in For testing For the fault The accuracy of the test For testing The false alarm rate, when hour, , ;when hour, , .

[0042] The reliability calculation formula for the test sequences obtained by the diagnostic strategy is as follows:

[0043] in Representative sequence The reliability of the k-th test.

[0044] (3) Fault isolation rate

[0045] Fault isolation rate is a metric that measures the ability of a device or system to accurately and quickly isolate the faulty portion from the normal portion when a fault occurs. Fault set Testable sequences The conditions for isolation can be expressed using the following mathematical expression:

[0046] in Representative fault and Corresponding test sequence The row vectors of the correlation matrix are not the same.

[0047] test sequence The formula for calculating the fault isolation rate after fault isolation is as follows:

[0048] In summary, by combining the three optimization metrics mentioned above—reducing testing costs, increasing test sequence reliability, and improving fault isolation rate—a mathematical model for a diagnostic strategy based on multi-valued testing and testing uncertainty can be obtained as follows:

[0049] II. MUPSOIG Algorithm Design Particle Swarm Optimization Framework The position vector of the particle This is related to the selection of the test sequence T during the optimization process. Since the test optimization problem is a discrete optimization problem, the Discrete Particle Swarm Optimization (BPSO) algorithm is used. Given that the components of the current position can only take the values ​​{0, 1}, the velocity and position update formulas become as follows (assuming that at this point...). )

[0050] Where ω is the inertia weight (linearly decreasing: ), For learning factors.

[0051] In BPSO, the definition is as follows: When selected ,when When not selected Before each particle update transition, it is necessary to calculate the test cost, test reliability, and fault isolation rate of the test sequence corresponding to the particle, record and compare them, and then complete the subsequent update process.

[0052] Information entropy heuristic function For the test point selection problem, the fault isolation capability of each test can serve as heuristic information for the PSO algorithm. Based on information entropy theory, information gain is used to evaluate the isolation capability of test points. Considering test uncertainties, the set of test points for the system under test is as follows: Assuming multi-valued testing There are m results. Represent these results using different sets. From the theory of information entropy, the information entropy function can be obtained as:

[0053] In the formula For testing The corresponding set of the kth diagnostic conclusions, where s is the fuzzy set to be diagnosed.

[0054] Considering the uncertainty of testing, that is, assuming that each test is not completely reliable, then the test... The probability of the divided subsets for:

[0055] in for For the fault The detection accuracy.

[0056] By combining test costs and test information, i.e., selecting the test-separated fault fuzzy set that maximizes unit information gain, a new evaluation function can be constructed:

[0057] The specific steps of the information entropy-based heuristic algorithm are as follows: Assuming the set of fault states of the system under test is S and the set of test sequences is T, calculate the unit information gain for each test point according to the information entropy function, and select the test point corresponding to the maximum unit information gain value. Select this test point Test and isolate the fuzzy set of system faults.

[0058] Algorithm fitness function design and definition The fitness function of the particle swarm optimization algorithm is designed based on the optimization index model in the previous section. That is, the fitness function is designed by combining the principles of minimizing test cost, maximizing reliability, and maximizing fault isolation rate. Therefore, the fitness function of an individual particle can be constructed as follows:

[0059] in, , and Let a, b, and c represent the expected test cost, reliability, and fault isolation rate of the fault diagnosis tree obtained by the Information Entropy Heuristic Algorithm (IG) from the test subsets corresponding to the individual particles in the particle swarm. The process of constructing the fault diagnosis tree using the IG algorithm is shown in steps 2). a, b, and c are parameters for the expected test cost, reliability, and fault isolation rate, which can control the weights of the three in the fitness function.

[0060] The core steps of the particle swarm optimization (PSO) information entropy algorithm for solving multi-valued tests and considering test uncertainty are as follows: Figure 1 As shown.

[0061] Step 1: Initialize the parameters, including learning factors c1 and c2, weight coefficients a, b, and c, inertia weight coefficient ω, the number of tests randomly selected in each iteration l, population size N, population size d, and maximum number of iterations iter.

[0062] Step 2: Initialize the population. Use binary encoding for the test set and randomly generate a population of size d, that is, the elements in the position vector are randomly initialized to 0 or 1, and the velocities of each particle are randomly generated.

[0063] Step 3: The information entropy algorithm is used to calculate the fitness function. A fault tree is constructed on the test set corresponding to each particle position vector using the information entropy algorithm. The fitness values ​​are calculated and compared. The best position of the particle and the corresponding fitness value are stored in a local best position set pbest. The best particle position corresponding to the smallest fitness value among all particles is saved to the global best position gbest.

[0064] Step 4: Population Iterative Update. The population is updated using BPSO with a linearly decreasing inertia weighting factor to update the position and velocity of each particle. Step 5: Repeat steps 3 and 4 until the maximum number of iterations is reached and iter or gbest remains unchanged across multiple iterations.

[0065] III. Examples and Diagnostic Methods for Aircraft Engine Systems Taking an aircraft engine as an example, the test sequence was optimized using the method described in this invention. Through analysis of the system's working principle and composition, a set of 13 possible fault states for the system was obtained. Furthermore, the system has eight associated test sets. 0, 1, and 2 are set as the three test conditions for the detection signal, meaning the elements of the correlation matrix take values ​​from 0 to 2. Based on statistical data, the probability of each fault state is obtained as follows: The correlation matrix is ​​shown in Table 1 below, with the last column representing the probability of failure. Fault detection accuracy is shown in Table 2, and testing cost and false alarm rate are shown in Table 3.

[0066] Table 1. Correlation Matrix of Fault Testing for a Certain Aircraft Engine System

[0067] Table 2 Accuracy Matrix of Fault Detection for a Certain Aircraft Engine System

[0068] Table 3. Testing Costs and False Alarm Rates of an Aircraft Engine System

[0069] For the optimal testing of the aforementioned engine equipment system, the following parameters are set: population size d=30, population number N=10, and binary particle swarm optimization learning factor. , Inertia weighting coefficient Change to a linearly decreasing relational weight factor, i.e. ,in , Maximum number of iterations Weight coefficients in the population fitness function , , .

[0070] The optimization result obtained after the iteration is That is, the selected test is At this point, the fault isolation rate is 100%. After the optimized tests are sorted globally, the fault tree can be obtained as shown in the figure below. The expected test cost can be obtained by the average test cost formula, which is 11.8765. The test reliability of this test sequence is 0.012.

[0071] (2) Validation by randomized trial To verify the algorithm's applicability and superiority, N was randomly generated. The multi-valued correlation matrices of N (20, 40, 60, 80, and 100, including values ​​0, 1, and 2) are compared with the proposed algorithm using an information-heuristic algorithm considering test uncertainty, a multi-valued Rollout algorithm, and the algorithm itself. To reduce the weighting influence of individual test costs, all test costs are set to 1, the sum of all fault occurrence probabilities is 1 and does not exceed 0.1, elements in the fault detection accuracy matrix are randomly selected between 0.79 and 0.97, the false alarm rate is randomly selected between 0.01 and 0.04, and the number of tests is randomly selected in each iteration. .

[0072] from Figure 3 It is clear that the reliability testing costs of the three algorithms are taken into account. Both increase with the increase of the dimension of the correlation matrix (system size), but show differences at the same dimension. The MUPSOIG algorithm... The fitness function consistently ranks lowest, followed by Rollout, with IG ranking highest. On one hand, it integrates the global search capability of particle swarm optimization, escaping local optima through population iteration, making it easier to find a globally optimal solution that balances "low average test cost (J)" and "high reliability." On the other hand, its fitness function directly incorporates... J , R The MUPSOIG algorithm achieves a balance among multiple objectives—fault isolation rate, reliability, and reliability calculation—through weight adjustment, particularly for reliability calculations related to test uncertainty, significantly reducing the hidden costs caused by missed detections or false alarms. Therefore, across all dimensions, the MUPSOIG algorithm... All values ​​remain at the lowest level, fully demonstrating its comprehensive optimization advantages in multi-value testing uncertainty scenarios.

[0073] from Figure 4It can be observed that the runtime of the particle swarm optimization algorithm proposed in this paper, which considers test uncertainty and combines information heuristics, increases with the dimension of the correlation matrix. Furthermore, under the same matrix dimension, the MUPSOIG algorithm has a slightly longer runtime than the Rollout algorithm, while the IG algorithm has the shortest runtime. The MUPSOIG algorithm in this paper embeds the information entropy heuristic as a module for calculating the fitness function within the particle swarm optimization framework. Therefore, the algorithm's time complexity is determined not only by the basic single-step heuristic search complexity but also by the population size d, test dimension N, and maximum number of iterations (iter) in the particle swarm algorithm. Thus, the complexity dimension of the algorithm in this paper is similar to that of the Rollout algorithm, but the actual runtime is higher due to the population maintenance, high-frequency information heuristic calls, and parameter settings within the particle swarm framework.

[0074] Those skilled in the art will understand that the above description is merely a preferred embodiment of the present invention, and the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. This is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0075] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for multi-valued system diagnosis strategy optimization considering test uncertainty, characterized in that, The method is realized based on a particle swarm information entropy algorithm, and comprises the following steps: Step 1, based on the fault test correlation matrix, establish the fusion average test cost J , test reliability R, fault isolation rate Multi-objective optimization model for optimization index; Step 2, introducing a multi-value information entropy heuristic function to improve the local search efficiency of the multi-objective optimization model established in step 1; Step 3, based on step 2, adopting a discrete particle swarm algorithm and designing a linearly decreasing inertia weight to realize global optimal search of the test sequence; Step 4, three indexes including average test cost J, test reliability R, and fault isolation rate are cooperatively optimized by fitness function , and a diagnosis strategy meeting engineering requirements is output.

2. The method of claim 1, wherein, The particle swarm information entropy algorithm comprises the following steps: Step 1.1, initializing parameters, initializing learning factors c1 and c2, weight coefficients a, b and c, inertia weight coefficient ω, test times l selected randomly each time, population quantity N, population size d, and maximum iteration times iter; Step 1.2, initializing the population, adopting binary coding for the test set, and randomly generating a population with a size of d, that is, the elements in the position vector are randomly initialized as 0 or 1, and the speed of each particle is randomly generated; Step 1.3, the information entropy algorithm calculates the fitness function, constructs a fault tree for the test set corresponding to each particle position vector by using the information entropy algorithm, compares the size of the fitness value, stores the best position of the particle and the corresponding fitness value in a local best position set pbest, and saves the best particle position corresponding to the minimum fitness value in all particles to a global best position gbest; Step 1.4, population iteration update, updating the population by using BPSO with a linearly decreasing inertia weight factor, updating the position and speed of each particle; Step 1.5, repeating steps 1.3 and 1.4 until the maximum iteration times iter are reached or gbest remains unchanged in multiple iterations.

3. The method of claim 1, wherein the method further comprises: The multi-value system comprises describing the multi-value test diagnosis strategy problem of a complex system by using a five-tuple test set: wherein S = {s1, s2,..., sn} represents the finite set of system states of the target system, Si = {si1, si2,..., sin} represents the set of system states only when the ith fault occurs, P = {p1, p2,..., pm} represents the set of prior probability distributions of various diagnostic conclusions of the target system; F = {f1, f2,..., fn} represents the fault probabilities, T = {t1, t2,..., tm} represents the set of tests available to the system; tj = {tj1, tj2,..., tjn} represents the jth available test, the test result has states other than pass and fail; R = {r1, r2,..., rm} represents the test resources consumed by each test, including test time, cost, test complexity and other factors; C = {c1, c2,..., cm} represents the correlation matrix between the fault set and the test set; Ci = {ci1, ci2,..., cin} represents the logical relationship between the ith test and the m fault modes, in the multi-value test, the correlation matrix elements have values other than 0 and 1.

4. The method of claim 1, wherein the method further comprises: The average test cost described in step 1 J For isolating faults The test cost and time generated by the test sequence; it is calculated as: wherein, represents the test sequence that the fault has undergone in the current diagnostic strategy, represents the sequence of tests in the k th test.​ 5. The method of claim 1, wherein the method further comprises: The reliability R in step 1 is obtained by two-level calculation of single test reliability and sequence reliability, and the calculation method is as follows: The reliability calculation formula of a single test is as follows: wherein To test the accuracy of the test for faults, the accuracy of the test for faults, To test the false alarm rate of the test, when , , ; when , , ; The reliability calculation formula of the test sequence obtained by the diagnosis strategy is as follows: wherein Representative sequences In the middle of the k Reliability of the test.

6. The method of claim 1, wherein the method further comprises: Fault isolation rate in step 1 For determining the fault set Can be tested sequences Isolated conditions, namely wherein representing a fault and corresponding test sequence the correlation matrix row vectors are not identical; Test sequence The fault isolation rate after fault isolation is calculated according to the following formula: 。 7. The method of claim 1, wherein the method further comprises: The specific steps of the multi-value information entropy heuristic function in step 2 are as follows: assuming that the fault state set of the measured system is S, and the test sequence set is T, the unit information gain of each test point is calculated according to the information entropy function, and the test point corresponding to the maximum unit information gain value is selected by comparison The test point is selected The system fault fuzzy set is tested and isolated.

8. The method of claim 1, wherein the method further comprises: The fitness function in step 4 is a fitness function designed by combining the minimum test cost, the maximum reliability and the maximum fault isolation rate, that is, wherein, , and respectively represent the expected test cost, reliability and fault isolation rate of the fault diagnosis tree obtained by the information entropy heuristic algorithm IG for the test subset corresponding to the particle swarm individual.

9. A computer storage medium having stored thereon a computer program, characterized in that The computer program is executed by a processor to realize the method in any one of claims 1-8.

10. A computer device, comprising: Comprise: a memory, a processor and a computer program stored on the memory and executable on the processor, and the processor executes the program to realize the method in any one of claims 1-8.

Images