Engine test scheme optimization method based on reliability

Abstract

The invention discloses a reliability-based engine test scheme optimization method, and belongs to the technical field of spaceflight quality and reliability. The test cost of the whole engine and components is considered, a large-scale intelligent differential optimization algorithm is adopted, a global optimal solution can be quickly obtained, the situation that optimization is terminated in advance when a local optimal solution is obtained or the global optimal solution cannot be found for a long time and is caught in a rigid bureau is avoided, and better calculation optimization efficiency is achieved; by limiting the reasonable test time range, the number of test samples and the like of the engine and the components, the basic rule of the reliability test is followed, a test scheme can be customized conveniently, rapid convergence of a large-scale intelligent optimization algorithm is accelerated, and calculation optimization is accelerated to obtain an optimal test scheme; and aiming at the reliability index requirement of the engine, the reliability index requirement of the engine can be verified at the minimum test cost.

Description

Technical Field

[0001] The present invention belongs to the technical field of aerospace quality and reliability, and particularly relates to an optimization method for an engine test plan based on reliability. Background Art

[0002] Reliability test verification of liquid rocket engines requires a large number of whole-machine and component tests, resulting in high costs. To control and reduce the development costs of engines, it is necessary to optimize the engine and component tests and plan the optimal test plan to minimize the reliability test costs.

[0003] Under the determined system reliability requirements, the key problem to be solved is how to obtain the optimal reliability test plan for the engine, control the minimum test time and the number of test stands for the engine and each component, so as to minimize the engine test costs while ensuring that the system reliability index requirements can be met. Summary of the Invention

[0004] In view of this, in accordance with the characteristics of the development of liquid rocket engines and combined with engineering practice, the present invention proposes an optimization method for a test plan based on reliability, filling the gap in the cost control method for liquid rocket engines based on reliability, providing technical support for engineering decision-making in engine reliability test projects, and having better computational optimization efficiency, better accuracy and flexibility.

[0005] An optimization method for an engine test plan based on reliability includes:

[0006] Step 1: Establish an optimization model for an engine test plan based on reliability, specifically including:

[0007] (1) Identify and determine the engine reliability test items

[0008] Determine the test items for the engine component and whole-machine reliability tests, and then identify the reliability key characteristic parameters for establishing a reliability evaluation model;

[0009] (2) Construct an engine reliability evaluation model;

[0010] (3) Determine the parameters and their ranges of the reliability evaluation model;

[0011] According to the determined engine reliability test items above, take the factors to be determined in the test items as parameters, and further determine the value ranges of the parameters; take the above factors as the decision variables of the reliability test plan optimization model, and take their value ranges as one of the constraint conditions of the optimization model;

[0012] (4) Test cost analysis and test cost calculation:

[0013] The test costs for time-related components include: the basic cost of the test and the cost of test implementation;

[0014] The test costs for time-unrelated components include: the cost of test implementation;

[0015] If the component test is a destructive test, the test cost also includes the cost of the test samples;

[0016] (5) The optimization model for the engine test plan based on reliability includes:

[0017] Taking the minimization of test cost as the objective function; taking the reliability of the engine system being greater than the set reliability threshold and the decision variables within their respective value ranges as constraints, as the optimization model;

[0018] Step 2: Solve the optimization model for the engine test plan based on reliability using the differential evolution method, specifically including:

[0019] Step 1): Set the parameters of the differential evolution algorithm: including population size, maximum number of evaluations, mutation factor, and crossover probability;

[0020] Step 2): Randomly generate an initial population containing the set population size; where, the individuals in the population are vectors composed of all decision variables, and the initial values of the decision variables are randomly generated;

[0021] Step 3): Based on the current population, calculate the objective function value, that is, calculate the reliability test cost of each individual in the population;

[0022] Step 4): Calculate the constraints of reliability, that is, calculate the reliability of each individual in the population;

[0023] Step 5): Evaluate the fitness of the population individuals, that is, evaluate the situation where each individual in the population meets the constraint conditions, and obtain the feasible solutions in the population;

[0024] Step 6): Select and record the optimal individual among the feasible solutions of the population;

[0025] Step 7): Based on the feasible solutions, cross and mutate to evolve a new generation of population;

[0026] Step 8): If the maximum number of evaluations has not been reached, return to Step 3) and continue to execute; if the maximum number of evaluations has been reached, judge whether the optimal solution has been obtained. If so, output the optimal solution to obtain the optimal reliability test plan; if there is no optimal solution, return to Step 1) to check the parameter settings or the optimization model.

[0027] Preferably, when the failure modes of the engine component include random failure, wear-out failure, and strength failure, the corresponding reliability assessments adopt the exponential distribution, Weibull distribution, and strengthening coefficient evaluation model; the reliability assessment of the whole machine adopts the Weibull distribution assessment model.

[0028] The decision variables of the exponential distribution and Weibull distribution assessment models include the test time and the number of test samples.

[0029] The decision variables of the strengthening coefficient evaluation model are the test time and the K value.

[0030] Preferably, the engine system reliability evaluation function is the reliability based on Bayesian and using the Metropolis-Hasting algorithm to fuse the component and engine whole machine information:

[0031]

[0032] In the formula, δ is the whole machine test run data, π(.) is the reliability posterior distribution after supplementing the whole machine test run data, L(.) is the likelihood function obtained through the whole machine test run data, and π 0 (R i ) is the reliability prior distribution of the engine; R i represents the reliability of the i-th component.

[0033] Preferably, the calculation method of the engine reliability prior distribution is: assuming that the failure probability distribution function of the i-th component is G i , the engine reliability prior probability obtained from the engine component test is 1 - ∏ i G i .

[0034] Preferably, for the reliability calculation of the traditional Weibull distribution assessment method after the engine whole machine test run data, the shape parameter m, the characteristic life The reliability additional information calculation method is as follows:

[0035]

[0036] In the formula, is the reliability point estimate, R L is the lower limit of the reliability, t is the engine mission time, r is the number of failures, n represents the number of test products, t i is the test time of the i-th sample, is the chi-square distribution with a confidence lower limit of γ.

[0037] The present invention has the following beneficial effects:

[0038] (1) Considering the test costs of the entire engine and its components, the present invention adopts a large-scale intelligent differential optimization algorithm, which can quickly obtain the global optimal solution, avoiding the situation of premature termination of the optimization due to obtaining a local optimal solution or being stuck in a deadlock for a long time without finding the global optimal solution, and has better computational optimization efficiency.

[0039] (2) In the present invention, the test plan optimization under the multi-model joint modeling reliability evaluation method is innovatively used, rather than the optimization under the simple traditional reliability model evaluation method, and has better accuracy and flexibility.

[0040] (3) By defining the reasonable test time range and the number of test samples of the engine and its components, etc., it not only follows the basic laws of reliability tests but also facilitates the customization of test plans, accelerates the rapid convergence of the large-scale intelligent optimization algorithm, and accelerates the computational optimization to obtain the optimal test plan.

[0041] (4) The present invention comprehensively considers the failure modes of reliability tests and can solve the optimization problem of large-scale test plans containing hundreds or thousands of test parameters.

[0042] (5) According to the reliability index requirements of the engine, the present invention can verify the reliability index requirements of the engine at the minimum test cost.

[0043] (6) The algorithm of the present invention has strong robustness, high search efficiency, is not easily trapped in local optima, can quickly and efficiently obtain stable optimization calculation results, has been applied in liquid rocket engines, and can be extended to the reliability tests of other complex mechanical equipment. Description of the Drawings

[0044] Figure 1 is the implementation process of the reliability-based engine test plan optimization method of the present invention;

[0045] Figure 2 is a schematic diagram of obtaining the engine system reliability by the engine multi-model joint modeling evaluation technology;

[0046] Figure 3 is the flowchart of the optimization method;

[0047] Figure 4 is a schematic diagram of the reliability multi-model joint model;

[0048] Figure 5 is the case optimization algorithm process;

[0049] Figure 6 is an example of the optimization result diagram. Detailed Embodiments

[0050] The present invention will be described in detail below with reference to the drawings and by way of examples.

[0051] In view of the product characteristics of liquid rocket engines and combining the engine test and failure mode situations, the present invention proposes a reasonable and feasible optimization method for the engine test plan based on reliability, plans the engine test plan, and verifies the reliability index requirements of the engine at the lowest test cost. It provides technical support for the engineering decision-making of liquid rocket engine flight missions. The implementation process of the optimization method for the engine test plan based on reliability is as Figure 1 shown, and the specific implementation process is as follows:

[0052] (1) Establish an optimization model for the engine test plan based on reliability

[0053] (1) Identify and determine the engine reliability test items

[0054] During the test plan planning stage, it is necessary to determine the test items for component and whole-machine reliability tests, and then identify the key reliability characteristic parameters for establishing a reliability evaluation model.

[0055] (2) Construct an engine reliability evaluation model

[0056] The engine reliability evaluation model based on multi-level product data is the most important constraint condition in the optimization of the engine test plan and the key criterion for measuring the feasibility of the engine test plan. The schematic diagram of the point estimate of the engine system reliability obtained by using the engine multi-model joint modeling evaluation technology is as Figure 2 shown.

[0057] The engine multi-model joint modeling evaluation constructs a reliability model by jointly modeling the event chain model and the fault tree model.

[0058] (3) Determine the parameters and their ranges of the reliability evaluation model

[0059] According to the above-determined engine reliability test items, the factors to be determined in the test items are used as parameters, and the value ranges of the parameters are further determined. For example, the test items with failure modes related to time include the number of test samples, test time, etc.; the test items with failure modes independent of time include the number of test samples, etc. It is necessary to specifically analyze and determine the reliability test items according to the specific reliability evaluation method.

[0060] The above factors are used as decision variables of the reliability test plan optimization model, and their value ranges are used as one of the constraint conditions of the optimization model.

[0061] (4) Test cost analysis and calculation method

[0062] The test costs of time-related components mainly include the basic test costs, the test costs per unit time, the test time, and the number of test samples; the test costs of time-independent components mainly include the costs of test implementation and the number of test samples. In addition, if the component test is a destructive test, the costs of the products participating in the test also need to be calculated. The test costs of the entire engine are calculated according to the time-related destructive test. For example, for a certain engine test plan, the calculation method of the test costs of engine components is shown in Table 1:

[0063] Table 1 Calculation Method of Test Costs of Engine Components

[0064]

[0065] Note: T is the test time, n is the number of samples, and C is the cost

[0066] (5) Optimization Model of Engine Test Plan Based on Reliability

[0067] The optimization of the test plan is an optimization problem of the cost objective under reliability constraints. To solve this problem, an optimization model of the engine test plan based on reliability needs to be established.

[0068] It is assumed that the known reliability index requirement of the engine system is not less than 0.985; the system consists of i randomly failing, j wear-out failing, and k strength failing components. Let T represent the test time of a single sample, n represent the number of test samples, K represent the K value of the enhanced coefficient method; the cost is represented by C X and the reliability model calculation function is represented by g(). Using the multi-model joint modeling and evaluation method introduced above to calculate the reliability value and the above engine test cost calculation method, the optimization model of the engine test plan can be expressed as follows:

[0069]

[0070] (2) Solving the Optimization Model of Engine Test Plan Based on Reliability

[0071] (1) Algorithm Analysis of the Optimization Model

[0072] The optimization model and the algorithm complement each other. After establishing the mathematical model of engine optimization, it is necessary to search for the optimal solution among the feasible solutions under the constraint conditions, and this process is realized through the optimization algorithm. The optimization algorithm is a search process or rule based on a certain idea and mechanism, and obtains the solution of the problem that meets the requirements through a certain way or rule.

[0073] The optimization model of the engine test scheme based on reliability presents complex characteristics such as strong non-linearity, strong constraints, multiple variables, and non-differentiable objective functions. The number of decision variables can reach hundreds or even thousands. The optimization of the engine test scheme based on reliability is a large-scale constrained optimization problem. This problem has complex constraint conditions that complicate the search process, and it is difficult to obtain its global optimal feasible solution.

[0074] There are many algorithms for solving constrained optimization problems. Generally, they can be divided into two categories: deterministic algorithms and stochastic algorithms according to their properties. Deterministic algorithms are usually gradient-based search methods. The main problems of this type of method are that the solution requires setting good initial points and the gradient information of the function, and mostly the obtained solutions are local optimal solutions. Compared with deterministic optimization methods, evolutionary algorithms are a type of stochastic intelligent optimization method, which is more suitable for solving constrained optimization problems. Evolutionary algorithms are a type of global optimization method that simulates natural processes. Compared with traditional optimization algorithms, evolutionary algorithms are a population-based search technique, with characteristics such as strong robustness, high search efficiency, and not easily getting stuck in local optima. Combining evolutionary algorithms with a certain constraint handling mechanism forms a constrained optimization evolutionary algorithm, which can better solve complex constrained optimization problems.

[0075] Differential evolution algorithm is a stochastic heuristic search algorithm, which is simple and easy to use, and has strong robustness and global optimization ability. Differential evolution algorithm is an intelligent optimization search algorithm generated through the cooperation and competition among individuals in the population. It is based on the global search strategy of the population, uses real number coding, simple mutation operations based on differences, and a "one-to-one" competitive survival strategy, reducing the complexity of evolutionary computing operations. At the same time, the unique memory ability of the differential evolution algorithm enables it to dynamically track the current search situation to adjust its search strategy, with strong global convergence ability and robustness, and does not need to rely on the characteristic information of the problem, and is suitable for solving complex optimization problems that are difficult or even impossible to solve using conventional mathematical programming methods.

[0076] The key idea of the differential evolution algorithm is different from traditional evolutionary methods: traditional methods use a pre-determined probability distribution function to determine vector perturbations; while the self-organization program of the differential evolution algorithm uses two randomly selected different vectors in the population to interfere with an existing vector, and every vector in the population has to interfere. The differential evolution algorithm uses a population of vectors, and the random perturbations of the population vectors can be carried out independently, so it is parallel.

[0077] The population reproduction scheme of the differential evolution algorithm is different from other evolutionary algorithms: it realizes "mutation" through the operation of adding the weighted difference vector between two members in the population to the third member to generate a new parameter vector; then it realizes "crossover" through the operation of mixing the parameters of the mutant vector and the parameters of another pre-determined target vector according to certain rules to generate a trial vector; finally, it realizes "selection" by selecting the lower one between the cost function of the trial vector and the cost function of the target vector to replace the target vector. All members in the population must be used as the target vector for such an operation once so that the same number of competitors appear in the next generation. In the process of evolution, the best parameter vector of each generation is evaluated to record the minimization process. In this way, by using the method of generating new individuals with random deviation perturbation, a very good convergence result can be obtained, guiding the search process to approach the global optimal solution.

[0078] Therefore, for the reliability-based engine test plan optimization problem with numerous decision variables, large space, non-linearity, and global optimization, the calculation process of the reliability assessment model is relatively complex. We use the differential evolution algorithm, an efficient parallel search algorithm, to solve and calculate this complex optimization problem of the reliability-based engine test plan optimization, and obtain good effects such as fast calculation speed, high efficiency, and stable optimization calculation results.

[0079] As Figure 5 shown, the differential evolution method specifically includes:

[0080] Step 1), setting the parameters of the differential evolution algorithm: including the population size (i.e., the number of individuals included in the population), the maximum number of evaluations, the mutation factor, and the crossover probability.

[0081] Step 2), randomly generating an initial population containing the set population size; among them, the individuals in the population are vectors composed of all decision variables, and the initial values of the decision variables are randomly generated;

[0082] Step 3), based on the current population, calculating the objective function value, that is, calculating the reliability test cost of each individual in the population;

[0083] Step 4), calculating the constraints of reliability, that is, calculating the reliability of each individual in the population;

[0084] Step 5), evaluating the fitness of the population individuals, that is, evaluating the situation where each individual in the population meets the constraint conditions to obtain the feasible solutions in the population;

[0085] Step 6), selecting and recording multiple optimal individuals in the population;

[0086] Step 7), based on multiple optimal individuals, cross-mutating and evolving to generate a new generation of population;

[0087] Step 8): If the maximum number of evaluations is not reached, return to Step 3) and continue; if the maximum number of evaluations is reached, determine whether the optimal solution is obtained. If so, output the optimal individual to obtain the optimal reliability test plan; if there is no optimal solution, return to Step 1) to check the parameter settings or optimize the model.

[0088] (2) Software implementation of the optimization model for engine test plans based on reliability

[0089] For large-scale complex optimization problems such as the optimization of engine test plans based on reliability, it is necessary to solve them by programming with the help of the fast computing power of a computer. The control parameters of the optimization algorithm play a crucial role in software design and debugging, and the control parameters have a great impact on a global optimization algorithm. In the selection of the algorithm and the main control parameters, after multiple test experiments and analysis of the optimization results, the main control parameters of the differential evolution algorithm adopted in the software design are shown in Table 2:

[0090] Table 2 Main control parameters of the differential evolution algorithm

[0091]

[0092] The optimization software process is as Figure 3 shown.

[0093] Implementation case:

[0094] The present invention provides an optimization method for engine test plans based on reliability, mainly by establishing an optimization model for engine test plans based on reliability and solving the model. On the premise of meeting the requirements of reliability indicators, an optimal test plan is obtained to minimize the cost of reliability tests.

[0095] Below, it is assumed that the entire engine consists of 8 components, including: 2 components with random failures, 3 components with wear-out failures, and 3 components with strength failures. Except for the K value (decision variable) of the strength failure strengthening coefficient method, the reliability evaluation parameters are known. According to the above method, an optimization model case for the 8-component engine test plan is constructed. For the 0.95 reliability index requirement of the engine, the main failure modes of the engine are analyzed, and the test plan for the engine is planned. The specific implementation steps are as follows:

[0096] (1) Establish an optimization model for engine test plans based on reliability

[0097] (1) Identify and determine the engine reliability test items

[0098] Analyze the failures of the 8 components Com1 to Com8. Assume that the reliability evaluations of the 8 components adopt the exponential distribution, Weibull distribution, and strengthening coefficient evaluation. The test items include:

[0099] 1) Tests of engine components: Specifically include test stress / test time; the number of samples put into the test.

[0100] 2) Engine full-load test run: Specifically include the single-engine test run time; the number of test samples.

[0101] Components Com1 - Com2 follow an exponential distribution, and components Com3 - Com5 follow a Weibull distribution. Both distributions are time-related. The decision variables for the 5 components include test times T1 - T5 and the number of test samples N1 - N5; the engine as a whole Fdj follows a Weibull distribution, and the decision variables are the test time Tfdj and the number of test samples Nfdj; components Com6 - Com8 are strengthening coefficients and are not related to time. The decision variables are test times N6 - N8. Additionally, there are three k values: K1 - K3.

[0102] There are a total of 18 decision variables for the model. See Table 3 for details.

[0103] Table 3 Decision Variables Table for Optimization Cases

[0104]

[0105] (2) Construct an engine reliability evaluation model

[0106] The multi-model joint model for calculating system reliability is as Figure 4 shown. The data required for the reliability of the multi-model joint model of reliability can be found in Table 4.

[0107] Table 4 System Reliability Data for Optimization Cases

[0108]

[0109]

[0110] (3) Determine the parameters and their ranges of the reliability evaluation model

[0111] The decision variables include the test time variable T, the number of test samples N, and the k values. Assuming the engine mission time is represented by t, the data range constraints for the decision variables are shown in Table 5.

[0112] Table 5 Data Range of Decision Variables for Optimization Cases

[0113]

[0114] (4) Test cost analysis and calculation method

[0115] The test cost data used in the case is shown in Table 6.

[0116] Table 6 Test Cost Data for Optimization Cases

[0117]

[0118] The objective function is to minimize the test cost, and the calculation principle of the test cost is as follows:

[0119] Time-related components: In the case of exponential and Weibull distributions, when the single-sample test time is greater than the mission time multiplied by the delivery threshold, it is a destructive test, and the cost of test samples needs to be increased, calculated using ∑ i ((C i基础费用 +C i单位试验时间 ×T i单件试验 )×n i +C i试验样件 ), otherwise calculated using ∑ j (C j基础费用 +C j单位试验时间 ×T j单件试验 )×n j .

[0120] Time-independent components: For the reinforcement coefficient evaluation method, when the reinforcement coefficient is greater than the K-value threshold, it is a destructive test, and the cost of test samples needs to be increased, calculated using ∑ k (C k单件试验 ×n k +C k试验样件 ), otherwise it is a non-destructive test, calculated using

[0121] ∑ l C l单件试验 ×n l .

[0122] (5) Optimization model of engine test plan based on reliability

[0123] The test plan optimization model is an optimization problem with the cost objective under reliability constraints. For the above 8-component case, the mathematical expression of the engine test plan optimization model is as follows:

[0124]

[0125] s.t.0.95 ≤ g(T1,T2,…T6,N1,N2,…N9,k1,k2,k3) ≤ 1

[0126]

[0127] In the above formula, n i is the number of test samples of the i-th component, and t i is the test time of the i-th component. C i1 is the basic cost of a single test of the i-th component, C i2 is the cost per unit test time of the i-th component, and C i3 is the cost of the test samples of the i-th component. Tfi is the threshold for judging whether the i-th component is subjected to a destructive test, which is related to the mission time of the engine. When conducting a destructive test on the component, the cost of test samples needs to be increased. T min is the minimum test time of the engine, T max is the set maximum test time of the engine.

[0128] The above reliability evaluation function g(T1, T2, … T6, N1, N2, … N9, k1, k2, k3) is the reliability based on Bayesian fusion of component and engine whole-machine information using the Metropolis-Hasting algorithm. The algorithm is as follows:

[0129]

[0130] In the formula, δ is the whole-machine test run data, π(.) is the reliability posterior distribution after supplementing the whole-machine test run data, L(.) is the likelihood function obtained through the whole-machine test run data, π 0 (R i ) is the reliability prior distribution of the engine, and R i represents the reliability of the i-th component.

[0131] 1) Calculation method of the engine reliability prior distribution

[0132] Assume that the component failure probability distribution function is G i (.), and the calculation method of the engine reliability prior probability obtained from the engine component test is as follows:

[0133]

[0134] The formula for the reliability evaluation model of the exponential distribution is as follows:

[0135] R(t) = e -λt , where λ is the failure rate;

[0136] The formula for the reliability evaluation model of the Weibull distribution is as follows:

[0137] m is the shape parameter and η is the characteristic life;

[0138] The formula for the stress-strength evaluation model is as follows:

[0139] μ x , μ y , are the mean and variance of strength and stress respectively.

[0140] 2) Calculation method of the reliability after the engine whole-machine test run

[0141] The traditional Weibull distribution evaluation method is adopted for the overall engine test. The shape parameter m, the characteristic life η, and the calculation method of reliability additional information are as follows:

[0142]

[0143] In the formula, is the reliability point estimate, and R L is the lower limit of reliability, t is the engine mission time, r is the number of failures, n represents the number of test products, and t i is the test time of the i-th sample. is the chi-square distribution with a confidence lower limit of γ.

[0144] (2) Solve the optimization model of the engine test plan based on reliability

[0145] The case sets the system reliability to be greater than 0.95, the significance level to be 0.1, the delivery threshold to be 1.5, the population size to be 90, the maximum number of evaluations to be 500, the mutation factor to be 0.5, the crossover probability to be 0.7, and the random seed to be 123. The optimization algorithm process solved by the differential evolution algorithm that uses the feasibility rule to handle the constraint conditions is as Figure 5 . Figure 5 Each individual in it contains 18 decision variables. The fitness of the individuals in the evaluation population is used to screen the feasible solutions that meet the constraints. The closer the reliability value is to 0.95, the better the fitness of the individual.

[0146] The list of case optimization results is shown in Table 7. The table lists the example of the case optimization results. The results show that when the system reliability reaches 0.95, the time and the number of samples required for components 1 to 5 to conduct tests, the number of test samples and the K value of components 6 to 8, and the test time and the number of samples required for the overall engine to conduct tests. The minimum cost in this plan is 186,085.54.

[0147] Table 7 Example of Case Optimization Results

[0148]

[0149]

[0150] Due to the randomness of the algorithm, the optimization results will fluctuate randomly. When the random seed is fixed for optimization, a unique solution can be obtained. Figure 6 is an example of the optimal value of the cost of the optimization plan obtained by multiple optimization calculations. The cost fluctuation does not exceed 0.1409%. In practical applications, multiple calculations can be performed and the optimal feasible test plan can be selected in combination with the actual situation.

[0151] In summary, the above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A reliability-based optimization method for engine test schemes, characterized in that, Including: Step 1: Establish an optimization model for the engine test plan based on reliability, specifically including: (1) Identify and determine the engine reliability test items Determine the test items for the engine component and whole-machine reliability tests, and then identify the key reliability characteristic parameters for establishing a reliability evaluation model; (2) Construct an engine reliability evaluation model; (3) Determine the parameters and their ranges of the reliability evaluation model; According to the determined engine reliability test items above, take the factors to be decided in the test items as parameters, and further determine the value ranges of the parameters; take the above factors as the decision variables of the reliability test plan optimization model, and take their value ranges as one of the constraint conditions of the optimization model; (4) Test cost analysis and test cost calculation: The test costs for components related to time include: the basic cost of the test and the cost of test implementation; The test costs for components not related to time include: the cost of test implementation; If the component test is a destructive test, the test cost also includes the cost of the test samples; (5) The optimization model for the engine test plan based on reliability, including: Taking the minimization of the test cost as the objective function; taking the engine system reliability being greater than the set reliability threshold and the decision variables within their respective value ranges as constraints, as the optimization model; Step 2: Solve the optimization model for the engine test plan based on reliability using the differential evolution method, specifically including: Step 1): Set the parameters of the differential evolution algorithm: including population size, maximum evaluation times, mutation factor, and crossover probability; Step 2): Randomly generate an initial population containing the set population size; where, the individuals in the population are vectors composed of all decision variables, and the initial values of the decision variables are randomly generated; Step 3): Based on the current population, calculate the objective function value, that is, calculate the reliability test cost of each individual in the population; Step 4): Calculate the constraints of reliability, that is, calculate the reliability of each individual in the population; Step 5): Evaluate the fitness of the population individuals, that is, evaluate the situation where each individual in the population meets the constraint conditions to obtain the feasible solutions in the population; Step 6): Select and record the optimal individual among the feasible solutions of the population; Step 7): Based on the feasible solutions, cross and mutate to evolve a new generation of population; Step 8): If the maximum evaluation number is not reached, return to Step 3) and continue to execute; if the maximum evaluation number is reached, judge whether the optimal solution is obtained. If so, output the optimal solution to obtain the optimal reliability test plan; if there is no optimal solution, return to Step 1) to check the parameter settings or the optimization model.

2. The optimization method of an engine test scheme based on reliability according to claim 1, wherein When the failure modes of the engine components include random failure, wear-out failure, and strength failure, the corresponding reliability evaluations adopt the exponential distribution, Weibull distribution, and strengthening coefficient evaluation models; the reliability evaluation of the whole machine adopts the Weibull distribution evaluation model; The decision variables of the exponential distribution and Weibull distribution evaluation models include test time and the number of test samples; The decision variables of the strengthening coefficient evaluation model are test time and the K value.

3. The optimization method for an engine test plan based on reliability as described in claim 2, wherein ,, characterized in that the reliability evaluation function of the engine system is the reliability after fusing component and engine information based on Bayesian using the Metropolis-Hasting algorithm: Where, δ is the data of the overall engine test run, π(.) is the reliability posterior distribution after supplementing the data of the overall engine test run, L(.) is the likelihood function obtained through the data of the overall engine test run, and π 0 (R i ) is the prior distribution of the reliability of the engine; R i represents the reliability of the i-th component.

4. The optimization method for an engine test plan based on reliability according to claim 3, wherein: The calculation method of the prior distribution of engine reliability is as follows: Assume that the failure probability distribution function of the i-th component is G i , and the prior probability of engine reliability obtained from engine component tests is 1 - ∏ ii G ii .

5. The optimization method for an engine test plan based on reliability as described in claim 3, characterized in that: Traditional Weibull distribution evaluation method for reliability calculation after engine full-load test run data, shape parameter m, characteristic life The calculation method of reliability additional information is as follows: In the formula, is the point estimate of reliability, R LL is the lower limit of reliability, t is the engine mission time, r is the number of failures, n represents the number of test products, t ii is the test time of the i-th sample, is the chi-square distribution with a confidence lower limit of γ.

Images