Automatic testing system and method for industrial internet software

Abstract

The invention relates to the technical field of software testing, in particular to an automatic testing system and method for industrial internet software. Comprising a software demand analysis module, a test case module, a test case distribution module and a test result analysis and visualization module, and the software demand analysis module defines test parameters of to-be-tested software according to demand specification description of the to-be-tested software; the test case module designs test cases through a model inspection method modified by an improved genetic algorithm, perfects the test cases in combination with an AI technology for automatically aided design of the test cases, and generates a test case group. Through a transfer learning-based use case generation algorithm, an automatic use case generation method for intelligent test of customized software is constructed, a model inspection method based on multi-source domain transfer learning is adopted, automation and intelligentization of a test process are realized, a software test method is optimized, and concurrency performance of automatic test of customized software and physical models in production is improved.

Description

Technical Field

[0001] The present invention relates to the field of software testing technology, and in particular to an industrial Internet software automation testing system and method. Background Art

[0002] With the rapid development of science and technology and society, computer software engineering has become increasingly important. Mechanical automation, intelligent control, the Internet, the Internet of Things, and other industries in various fields have increasingly higher requirements for the efficiency, control, or running speed of software development. Software development is the process of building a software system or the software part of a system according to user requirements. Software development is a systematic project that includes requirements capture, requirements analysis, design, implementation, and testing.

[0003] Intelligent testing of software is not only about letting computers automatically complete the testing process, but also requires the testing system to be able to autonomously identify different controls in customized software based on machine learning. The functional customized software used by the industrial Internet platform often has very different control design styles. If models are established and verified for all customized functions, it will greatly increase the difficulty of developing intelligent testing tools and affect the efficiency of intelligent testing. In addition, the automatic generation technology of test cases for complete paths is not yet mature. At present, this technology mainly uses path coverage technologies such as symbolic execution to obtain constraints on the path, solve constraints, and generate test cases covering the path. Among them, the complexity of software systems is usually very high, with a large number of paths and branches. Symbolic execution is also limited by program scale, execution time, and resource consumption. Error points are prone to occur during the test case generation process, resulting in inaccurate case generation. Summary of the invention

[0004] The purpose of the present invention is to provide an industrial Internet software automation testing system and method to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides an industrial Internet software automation testing system, including a software requirement analysis module, a test case module, a test case allocation module and a test result analysis and visualization module, wherein: The software requirement analysis module specifies the test parameters of the software to be tested according to the requirement specifications of the software to be tested; The test case module designs test cases by improving the model checking method modified by the genetic algorithm, and improves the test cases by combining the AI ​​technology of automatically assisting in the design of test cases, thereby generating a test case group; Adopt the model verification method based on multi-source domain transfer learning to realize the automation and intelligence of the test process, and use the existing model verification results for model-based system engineering to complete the model equivalence verification of new software or new functions; The test case allocation module allocates the test tasks one by one to the execution clients of the idle execution machines in the associated test machine cluster, dynamically generates and allocates test cases from the model library according to the data flow feedback, until all the test cases in the test task have been completed; The test result analysis and visualization module generates a test report based on the test results, searches for various errors and optimal solutions based on a random mutation algorithm, and enables users to interact with the software in a visual manner through intelligent test result visualization technology.

[0006] As a further improvement of the present technical solution, the test parameters include functional requirements, performance requirements, security requirements and compatibility requirements of the software.

[0007] As a further improvement of the present technical solution, the test case module includes a test case design unit, and the test case design unit designs test cases by improving the model checking method corrected by the genetic algorithm, and the specific steps are as follows: S110, initialize the population; the design population is ,in, Indicates test cases, each test case consists of a set of parameters; S120, fitness evaluation; define a fitness function Used to evaluate test cases Execution results on industrial Internet software, and adaptability function Based on multiple indicators, such as test coverage, error detection rate, response time, etc., through the adaptive function To calculate the fitness value; S130, random mutation operation; S140, iterative optimization; by repeating the mutation steps until the stopping condition is met, when the stopping condition is met, the optimal test case set is output.

[0008] As a further improvement of the technical solution, the test case module includes a model equivalence verification unit. In the model equivalence verification method based on multi-source domain deep transfer learning in the model equivalence verification unit, the samples in the source domain are: ; ; in, It's data. It’s a label; Indicates domain; Indicates the number of samples in the domain; Indicates the first samples; the samples in the target domain are: ; .

[0009] As a further improvement of the technical solution, the dynamic generation and allocation of test cases in the test case allocation module is a dynamic case allocation technology based on data flow feedback, wherein the test case allocation result adopts the following formula: ; Probability Vector For the Subdomain After the first test, the reliability capability can be enhanced in this subdomain; During the test, Tests will be assigned to subdomains middle; Under importance sampling, the superposition of information is obtained, which is close to reducing the number of samples. This paper attempts an adaptive application of importance sampling and uses the following formula to obtain : ; is the difference between the estimated sample distribution and the true distribution, As the expectation for this approximation, It is in The number of subfields for which at least one test case was executed in iterations. Is the significance level The normal distribution of is The number of test cases to be executed in each iteration.

[0010] As a further improvement of the present technical solution, the algorithm for finding various errors and finding the optimal solution is based on the random mutation algorithm in the random mutation operation in S120. The following is the pseudo code of the random mutation algorithm: Initializing the population Afterwards, for each iteration , do the following: The first step is for each individual , generating a random mutation solution ; Step 2: Calculation and The fitness value of and ; Step 3: If , then accept As the next generation solution, otherwise keep As a solution for the next generation; Step 4: End the algorithm and return the optimal solution.

[0011] As a further improvement of the present technical solution, the formula of the random mutation algorithm is: Generate variant solutions: ; in p is a random vector indicating the direction and magnitude of the mutation; The following is the formula to calculate the probability of acceptance: ; Where T is a parameter that controls the acceptance probability, which usually decreases as the number of iterations increases. The following is the formula of the fitness function: ; In the formula, It is the result of the individual running on the functional model of the software being tested. It is expressed as the value to be found. Output value It represents the individual that is most fit in the population.

[0012] The second object of the invention is to provide a method for operating the above-mentioned industrial Internet software automated testing method, including the following method steps: S1. According to the requirements specification of the software to be tested, clarify the test parameters of the software to be tested; S2. Design test cases by improving the model checking method modified by genetic algorithm, and improve the test cases by combining AI technology for automatic auxiliary design of test cases to generate a test case group; Adopt the model verification method based on multi-source domain transfer learning to realize the automation and intelligence of the test process, and use the existing model verification results for model-based system engineering to complete the model equivalence verification of new software or new functions; S3, assigning the test tasks one by one to the execution clients of the idle execution machines in the associated test machine cluster, dynamically generating and assigning test cases from the model library according to the data flow feedback, until all test cases in the test task have been completed; S4. Generate a test report based on the test results, find various errors and optimal solutions based on the random mutation algorithm, and use intelligent test result visualization technology to enable users to interact with the software in a visual way.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. In the industrial Internet software automation testing system, an automatic case generation method for customized software intelligent testing is constructed through a case generation algorithm based on transfer learning. A model verification method based on multi-source domain transfer learning is adopted to realize the automation and intelligence of the testing process, optimize the software testing method, and improve the concurrent performance of automatic testing of customized software and physical models in production. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram of an industrial Internet software automation testing system of the present invention; Figure 2 A schematic diagram of a process flow of an automated testing method for industrial Internet software according to the present invention; Figure 3 It is a feature extraction statistical graph of the present invention; Figure 4 A schematic diagram of the execution of a use case based on the evenly distributed testing technology of the present invention; Figure 5 It is a schematic diagram of the execution of a use case set of the dynamic use case allocation technology based on data flow feedback of the present invention.

[0015] The meaning of each number in the figure is: 100. Software requirement analysis module; 200. Test case module; 210. Test case design unit; 220. Model equivalence verification unit; 300. Test case allocation module; 400. Test result analysis and visualization module. DETAILED DESCRIPTION

[0016] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, an industrial Internet software automation testing system is provided, including a software requirement analysis module 100, a test case module 200, a test case allocation module 300 and a test result analysis and visualization module 400, wherein: The software requirement analysis module 100 specifies the test parameters of the software to be tested according to the requirement specifications of the software to be tested; The test case module 200 designs test cases by improving the model checking method modified by the genetic algorithm, and improves the test cases by combining the AI ​​technology of automatically assisting in the design of test cases, thereby generating a test case group; Adopt the model verification method based on multi-source domain transfer learning to realize the automation and intelligence of the test process, and use the existing model verification results for model-based system engineering to complete the model equivalence verification of new software or new functions; The test case allocation module 300 allocates the test tasks one by one to the execution clients of the idle execution machines in the associated test machine cluster, and dynamically generates and allocates test cases from the model library according to the data flow feedback until all the test cases in the test task have been completed; The test result analysis and visualization module 400 generates a test report based on the test results, searches for various errors and optimal solutions based on a random mutation algorithm, and enables users to interact with the software in a visual manner through intelligent test result visualization technology.

[0018] Among them, for the test parameters of the software to be tested in the software requirement analysis module 100, the test parameters include the functional requirements, performance requirements, security requirements and compatibility requirements of the software. And according to the requirements of the software, the test team also needs to formulate a detailed test plan to clarify the test scope, test strategy, test progress and required test resources. The test strategy can include different test types such as unit testing, integration testing, and system testing.

[0019] Secondly, before testing the industrial Internet software, it is necessary to design test cases. To this end, the test case module 200 includes a test case design unit 210. The test case design unit 210 designs test cases by improving the model checking method modified by the genetic algorithm. The specific steps are as follows: S110, initialize the population; the design population is ,in, Indicates test cases, each test case consists of a set of parameters. For example: ; S120, fitness evaluation; define a fitness function Used to evaluate test cases Execution results on industrial Internet software, and adaptability function Based on multiple indicators, such as test coverage, error detection rate, response time, etc., through the adaptive function To calculate the fitness value; S130, random mutation operation; S140, iterative optimization; by repeating the mutation steps until the stopping condition is met (such as reaching the maximum number of iterations, the fitness value no longer increases, etc., then no more iterations are performed), when the stopping condition is met, the optimal test case set is output.

[0020] After outputting the optimal test case set, the AI ​​technology for automatically assisting in the design of test cases can reduce manual input while ensuring the perfection of test cases. On the basis of the optimal test case set, the introduction of AI technology for automatically assisting in the design of test cases can further improve the perfection and efficiency of test cases, including intelligently recommending new test cases, or supplementing and modifying existing test cases, and automatically generating test cases for specific functions or scenarios based on the software's code structure, data model and other information, and optimizing test cases, including adjusting the order of test cases, merging redundant test cases, and optimizing the input data of test cases.

[0021] Because intelligent testing of software is not only about letting the computer automatically complete the testing process, but also requires the testing system to be able to autonomously identify different controls in customized software based on machine learning. The functional customized software used by the Industrial Internet platform often has very different control design styles. If models are established and verified for all customized functions, it will greatly increase the difficulty of developing intelligent testing tools and also affect the efficiency of intelligent testing. Therefore, the model equivalence test method based on transfer learning is used to complete the model equivalence test of new software or new functions.

[0022] The test case module 200 includes a model equivalence verification unit 220. In the model equivalence verification method based on multi-source domain deep transfer learning in the model equivalence verification unit 220, the samples in the source domain are: ; ; in, It's data. It’s a label; Indicates domain; Indicates the number of samples in the domain; Indicates the first samples; the samples in the target domain are: ; ; Reflect the data features of inconsistent fields (such as two different data sets) into the same feature space, making their distances in this space as close as possible, and constructing Figure 3 Feature extraction statistics shown.

[0023] That is, suppose that the color information of a certain area is used as an image feature. As shown in the figure, the red line represents the information value distribution represented by the color of the source data set, and the blue line represents the color information value distribution represented by the target data set. For this feature, the data of these two domains are originally converted to each other. However, this transformation will greatly reduce the accuracy of our model analysis, resulting in the color information of the area not matching the selected feature. This is the principle of multi-source domain learning selection.

[0024] Through the above model equivalence verification of customized software intelligent testing, and through the application of transfer learning, especially by designing an adaptation layer and an additional domain confusion loss, the generalization ability of the model between different domains can be effectively promoted while maintaining the invariance of sentence meaning and domain; ; in, is the maximum mean difference. Apply the loss function: ; Here, represents the final loss of the network, represents the conventional classification loss on data that the network already has annotations for (mostly the source domain) (this coincides with ordinary deep networks), represents the adaptive loss of the network, The last part is the uniqueness of transfer learning. is the weight parameter that weighs the two parts.

[0025] Through the above operations, the versatility of model verification is improved. When using the original convolutional layer intermediate results, some parameters need to be adjusted according to the actual software functions and data characteristics. Unify the results of multi-source domain deep transfer learning and classic machine learning when identifying the same model object.

[0026] In addition, as the software usage environment becomes increasingly complex, it is difficult to evaluate its usage status. Once a problem occurs during use, it is difficult to reproduce, and maintenance during use becomes increasingly difficult. Although the content and relationship of requirements, functions, logic, and physical processes have been determined, there is no definition of how to test against the requirements and implement small closed loops in the process to avoid discovering problems at the end. In response to this situation, model-based testing and verification of data flow feedback information can discover problems introduced during the implementation process as early as possible, shortening the development cycle and reducing costs overall.

[0027] Therefore, the dynamic generation and allocation of test cases in the test case allocation module 300 is a dynamic case allocation technology based on data flow feedback, wherein the test case allocation result adopts the following formula: ; Probability Vector For the Subdomain After the first test, the reliability capability can be enhanced in this subdomain; During the test, Tests will be assigned to subdomains middle; Under importance sampling, the superposition of information is obtained, which is close to reducing the number of samples. This paper attempts an adaptive application of importance sampling and uses the following formula to obtain : ; is the difference between the estimated sample distribution and the true distribution, As the expectation for this approximation, It is in The number of subfields for which at least one test case was executed in iterations. Is the significance level The normal distribution of is The number of test cases to be executed in each iteration.

[0028] Compared with evenly distributed testing, this technology distributes the test work to each function in proportion to the probability of occurrence according to the running profile of the software to be tested, thus reducing the failure rate of use case execution. Figure 4 and Figure 5 As shown, the execution of use cases based on the uniform distribution test technology and the execution of use case sets based on the dynamic use case allocation technology based on data flow feedback are respectively shown.

[0029] Through the above research, we can customize the software distributed cluster parallel testing and realize the data flow monitoring of multiple database resources at the same time. We can dynamically adjust the test case generation strategy according to the data offset. We can solve the data conversion problem between data deviation analysis and process execution and break the information barrier between data monitoring and process execution.

[0030] Furthermore, in the process of software intelligent testing, if there is a lack of performance testing for industrial software, or if performance testing is difficult to reach the limit level, there is a high probability that industrial software will have systemic or functional errors under special working conditions, which will lead to closed-loop errors in production control and cause safety accidents. At the same time, in the existing process of industrial software testing, test reports are generally summarized by manual means, which lack flexibility and have potential errors and loopholes.

[0031] To this end, through intelligent analysis of the test results, the errors and error locations reflected in the test process are displayed, and the results are explained. The report describes the important events generated in the test, such as error points and checkpoints. In-depth acquisition of the key points of any error that is not covered in the test scope. Considering that the characteristics of genetic algorithms are search and optimization skills that can adaptively adjust functions, the design of this solution uses genetic algorithms to automatically generate test cases for intelligent analysis of software functions to achieve intelligent testing. Among them, the algorithm for finding various errors and finding the optimal solution is based on the random mutation algorithm in the S130 random mutation operation. The following is the pseudo code of the random mutation algorithm: Initializing the population Afterwards, for each iteration , do the following: The first step is for each individual , generating a random mutation solution ; Step 2: Calculation and The fitness value of and ; Step 3: If , then accept As the next generation solution, otherwise keep As a solution for the next generation; Step 4: End the algorithm and return the optimal solution.

[0032] During the execution of this algorithm, by randomly mutating each individual, sorting them into new individuals, calculating fitness and retaining the results, it is possible to find out the individuals that are not suitable for this algorithm, and also find out the errors. At the same time, with step-by-step iterations, it is also possible to find the optimal individual that adapts to this population, that is, the optimal solution.

[0033] The following is the formula for the random mutation algorithm: Generate variant solutions: ; in p is a random vector indicating the direction and magnitude of the mutation; The following is the formula to calculate the probability of acceptance: ; Where T is a parameter that controls the acceptance probability, which usually decreases as the number of iterations increases. The following is the formula of the fitness function: ; In the formula, It is the result of the individual running on the functional model of the software being tested. It is represented by the value that needs to be found.

[0034] Output value It represents the individual that is most fit in the population.

[0035] Intelligent test result visualization technology: Add snapshots of each step in the results to express the experience and results of the test run. And based on the test process and the various test process data generated and recorded. Considering that Matlab can be used to create visualizations and user interfaces of software functions, and Matlab has built-in functions and neural network toolboxes for creating graphical user interfaces (GUIs) that allow users to interact with the software in a visual way, the Matlab language is used in the intelligent test visualization.

[0036] Regarding software function testing, the Matlab training program of our work is as follows: Function Train() / / Training process; In=Mormal(p) / / Sample input data normalization; Out=Mormal(n) / / Normalization of sample output data; net=newff(minmax(x),[20,3],{'logsig','purelin'},'trainbr'); / / Set the number of network layers, hidden layer neural function, inter-layer transfer function, and training algorithm.

[0037] Before the network starts training, p obtains training sample n and then stores all input data in the training sample into matrix p and all output data into matrix n. p and n constitute the entire training sample. The training program first normalizes p and n. Secondly, the network structure is set, including the number of network layers, the number of hidden layer neurons, the inter-layer transfer function, and the training algorithm. The network of this program is a three-layer structure. The input layer neurons are determined according to the input amount of the sample. The hidden layer neurons are 20 and the output neurons are 3. The hidden layer and the output layer correspond to the S-type and linear transfer functions, and the training algorithm is Bayesian regularization.

[0038] The logarithmic sigmoid function formula is: ; exp represents the natural exponential function. The logarithmic sigmoid function is a commonly used activation function that maps any real number to a value between 0 and 1. The formula of the linear transfer function is: ; Through the above research, the test report is automatically generated, including test case information, execution status information, execution result information, execution result status, execution process record, and error information. Directly display the location and error in the test, and explain the results. Intuitively display the test process and execution interface pictures, display interface output value information, including data tables during execution, database call information, etc. Establish a logical mapping relationship between the test results and the report description language in the database. Solve the problem of visual display of different information in the test report.

[0039] like Figure 2 As shown, the second object of the present invention is to provide an industrial Internet software automation testing method, including the following method steps: S1. According to the requirements specification of the software to be tested, clarify the test parameters of the software to be tested; S2. Design test cases by improving the model checking method modified by genetic algorithm, and improve the test cases by combining AI technology for automatic auxiliary design of test cases to generate a test case group; Adopt the model verification method based on multi-source domain transfer learning to realize the automation and intelligence of the test process, and use the existing model verification results for model-based system engineering to complete the model equivalence verification of new software or new functions; S3, assigning the test tasks one by one to the execution clients of the idle execution machines in the associated test machine cluster, dynamically generating and assigning test cases from the model library according to the data flow feedback, until all test cases in the test task have been completed; S4. Generate a test report based on the test results, find various errors and optimal solutions based on the random mutation algorithm, and use intelligent test result visualization technology to enable users to interact with the software in a visual way.

[0040] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An industrial Internet software automated testing system, characterized by: The system comprises a software requirement analysis module (100), a test case module (200), a test case allocation module (300) and a test result analysis and visualization module (400), wherein: The software requirement analysis module (100) specifies the test parameters of the software to be tested according to the requirement specification of the software to be tested; The test case module (200) designs test cases by improving the model checking method modified by the genetic algorithm, and improves the test cases by combining the AI ​​technology of automatically assisting in the design of test cases, thereby generating a test case group; Adopt the model verification method based on multi-source domain transfer learning to realize the automation and intelligence of the test process, and use the existing model verification results for model-based system engineering to complete the model equivalence verification of new software or new functions; The test case allocation module (300) allocates the test tasks one by one to the execution clients of the idle execution machines in the associated test machine cluster, dynamically generates and allocates test cases from the model library according to the data flow feedback, until all the test cases in the test task have been completed; The test result analysis and visualization module (400) generates a test report according to the test results, searches for various errors and optimal solutions based on a random mutation algorithm, and enables users to interact with the software in a visualized manner through intelligent test result visualization technology.

2. The industrial Internet software automation testing system according to claim 1 is characterized in that: The test parameters include the software's functional requirements, performance requirements, security requirements, and compatibility requirements.

3. The industrial Internet software automation testing system according to claim 1 is characterized in that: The test case module (200) comprises a test case design unit (210), wherein the test case design unit (210) designs a test case by improving a model checking method modified by a genetic algorithm, and the specific steps are as follows: S110, initialize the population; the design population is ,in, Indicates test cases, each test case consists of a set of parameters; S120, fitness evaluation; define a fitness function Used to evaluate test cases Execution results on industrial Internet software, and adaptability function Based on multiple indicators, such as test coverage, error detection rate, response time, etc., through the adaptive function To calculate the fitness value; S130, random mutation operation; S140, iterative optimization; by repeating the mutation steps until the stopping condition is met, when the stopping condition is met, the optimal test case set is output.

4. The industrial Internet software automation testing system according to claim 1 is characterized in that: The test case module (200) comprises a model equivalence verification unit (220). In the model equivalence verification method based on multi-source domain deep transfer learning in the model equivalence verification unit (220), the samples in the source domain are: ； ； in, It's data. It’s a label; Indicates domain; Indicates the number of samples in the domain; Indicates the first samples; the samples in the target domain are: ； 。 5. The industrial Internet software automation testing system according to claim 1 is characterized in that: The dynamic generation and allocation of test cases in the test case allocation module (300) is a dynamic case allocation technology based on data flow feedback, wherein the test case allocation result adopts the following formula: ； Probability Vector For the Subdomain After the first test, the reliability capability can be enhanced in this subdomain; During the test, Tests will be assigned to subdomains middle; Under importance sampling, the superposition of information is obtained, which is close to reducing the number of samples. This paper attempts an adaptive application of importance sampling and uses the following formula to obtain : ； is the difference between the estimated sample distribution and the true distribution, As the expectation for this approximation, It is in The number of subfields for which at least one test case was executed in iterations. Is the significance level The normal distribution of is The number of test cases to be executed in each iteration.

6. The industrial Internet software automation testing system according to claim 3 is characterized in that: The algorithm for finding various errors and finding the optimal solution is based on the random mutation algorithm in the random mutation operation of S120. The following is the pseudo code of the random mutation algorithm: Initializing the population Afterwards, for each iteration , do the following: The first step is for each individual , generating a random mutation solution ; Step 2: Calculation and The fitness value of and ; Step 3: If , then accept As the next generation solution, otherwise keep As a solution for the next generation; Step 4: End the algorithm and return the optimal solution.

7. The industrial Internet software automation testing system according to claim 6 is characterized in that: The formula of the random mutation algorithm is: Generate variant solutions: ； in p is a random vector indicating the direction and magnitude of the mutation; The following is the formula to calculate the probability of acceptance: ； Where T is a parameter that controls the acceptance probability, which usually decreases as the number of iterations increases. The following is the formula of the fitness function: ； In the formula, It is the result of the individual running on the functional model of the software being tested. It is expressed as the value to be found. Output value It represents the individual that is most fit in the population.

8. A method for operating an automated testing method for industrial Internet software as described in claim 1, characterized in that: The method comprises the following steps: S1. According to the requirements specification of the software to be tested, clarify the test parameters of the software to be tested; S2. Design test cases by improving the model checking method modified by genetic algorithm, and improve the test cases by combining AI technology for automatic auxiliary design of test cases to generate a test case group; Adopt the model verification method based on multi-source domain transfer learning to realize the automation and intelligence of the test process, and use the existing model verification results for model-based system engineering to complete the model equivalence verification of new software or new functions; S3, assigning the test tasks one by one to the execution clients of the idle execution machines in the associated test machine cluster, dynamically generating and assigning test cases from the model library according to the data flow feedback, until all test cases in the test task have been completed; S4. Generate a test report based on the test results, find various errors and optimal solutions based on the random mutation algorithm, and use intelligent test result visualization technology to enable users to interact with the software in a visual way.

Images