Train network control system software reliability evaluation method

Abstract

This invention relates to a method for evaluating the reliability of train network control system software, comprising the following steps: Step 1: Calculating a reliability factor based on collected failure data; Step 2: Determining data trends based on the significance level under a normal distribution and the calculated reliability factors of the failure data; Step 3: Evaluating the reliability of the software based on multiple software reliability models; Step 4: Evaluating the predicted quality of each software reliability model based on the relative accuracy of the predicted software reliability. This method for evaluating the reliability of train network control system software, after obtaining failure data and before selecting a specific model, performs trend analysis on the data, and the analysis results can serve as guidance for model selection and the use of effective data; after evaluating the reliability of the software, it evaluates the predicted quality of the model based on the predicted effectiveness, thereby obtaining the optimal reliability evaluation result based on the rationality and relative accuracy of the reliability evaluation.

Description

Technical Field

[0001] This invention relates to the field of train network control system technology, and in particular to a method for evaluating the reliability of train network control system software. Background Technology

[0002] Software reliability assessment, or software reliability evaluation, refers to "the process of determining the level of reliability achieved by an existing system or system components." Software reliability assessment is defined as "the application of statistical techniques to observable failure data collected during system testing and operation to evaluate software reliability." Software reliability assessment is a quantitative estimate and evaluation of the software reliability level after obtaining software failure data.

[0003] Software failure data can be obtained in two ways: first, in the later stages of the testing phase, through software reliability testing, failure data is collected during the testing process to estimate the software's reliability level and predict the potential future reliability level; second, after the software is put into use, failure data is collected during actual use to assess the software's reliability and predict the potential future reliability level. The software reliability assessment results at this stage are considered a true reflection of the software's ability to perform its specified functions. That is, the reliability assessment results at this stage not only provide the actual reliability level but also serve as a reference for determining the quantitative reliability requirements of next-generation software or similar software.

[0004] One of the main purposes of collecting failure data is to build software reliability models in order to assess software reliability. The previous approach was to directly use data to build reliability growth models, but in practice, it has been found that using reliability growth models in the early stages of program development and the later stages of program debugging is not ideal. Summary of the Invention

[0005] The present invention aims to solve the technical problems in the prior art by providing a method for evaluating the reliability of train network control system software.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] A method for evaluating the reliability of train network control system software includes the following steps:

[0008] Step 1: Calculate the reliability factor based on the collected failure data;

[0009] Step 2: Determine the data trend based on the significance level under the normal distribution and the calculated reliability factor of the failure data;

[0010] Step 3: Evaluate the reliability of the software based on multiple software reliability models;

[0011] Step 4: Evaluate the predicted quality of each of the multiple software reliability models based on the relative accuracy of the software reliability predictions.

[0012] In the above technical solution, step 1 specifically includes:

[0013] When the collected failure data is the failure interval time, let θ j Let j = 1, 2, ..., L be random variables T. j One implementation of this is such that, for each failure i, the reliability factor u(i) satisfies:

[0014]

[0015] When the collected failure data is failure intensity, assuming the time interval [0, t] is uniformly divided into k parts, and n(i) is the number of failures per unit time, i = 1…k, then the reliability factor u(k) satisfies:

[0016]

[0017] When the collected failure data is the cumulative number of failures, let N(k) be the cumulative number of failures before the k-th time unit, then the reliability factor u(k) satisfies:

[0018]

[0019] In the above technical solution, step 2 specifically means: when the significance level under the normal distribution is 5%, if |u(k)| or |u(i)| < 1.96, the trend is considered stable;

[0020] When |u(k)| or |u(i)| < 1.96 is not satisfied: if u(k) or u(i) < 1.645, the reliability is considered to have increased; if u(k) or u(i) > -1.645, the reliability is considered to have decreased.

[0021] In the above technical solutions, the software reliability models in step 3 include: Jelinski-Moranda model, Goel-Okumoto model, and Duane model.

[0022] In the above technical solution, step 4 specifically includes:

[0023] After j-1 failures occur, the time T for the j-th failure to occur is... jIt follows a true distribution, and the failure rate of this true distribution is f. j (t), Model A and Model B respectively apply f based on the previous j-1 failure times. j (t) Make predictions respectively, and the predicted results are as follows: and

[0024] During software operation, the time t at which the j-th failure occurs is observed. j , will t j Substitute into and Go to the middle, get and

[0025] if but Closer to f j If the value of (t) is large, then Model A will predict better quality;

[0026] if but Closer to f j If the value of (t) is large, the B model will predict better quality.

[0027] if Therefore, the predicted quality of Model A and Model B is the same.

[0028] The beneficial effects of this invention are:

[0029] The train network control system software reliability assessment method of this invention performs trend analysis on the failure data after obtaining it, but before selecting a specific model. The analysis results can serve as guidance for model selection and the use of effective data. Trend analysis of failure data can preliminarily screen model types, which can greatly improve modeling efficiency and avoid blind selection to some extent.

[0030] The train network control system software reliability assessment method of the present invention, after assessing the reliability of the software, performs a predicted quality evaluation on the model based on the predicted effectiveness, thereby obtaining the best reliability assessment result based on the rationality and relative accuracy of the reliability assessment. Attached Figure Description

[0031] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0032] Figure 1 This is a flowchart illustrating the steps of the train network control system software reliability assessment method of the present invention. Detailed Implementation

[0033] The present invention will now be described in detail with reference to the accompanying drawings.

[0034] like Figure 1 As shown, the train network control system software reliability assessment method of the present invention includes the following steps:

[0035] Step 1: Calculate the reliability factor based on the collected failure data.

[0036] By analyzing the patterns in failure data collected during software reliability testing, the reliability of the software can be assessed. The data being tested can be the failure interval, failure intensity, or cumulative number of failures.

[0037] When the collected failure data is the failure interval time, it is assumed that the failure data follows a non-homogeneous Poisson distribution (NHPP), and let θ j Let j = 1, 2, ..., L be random variables T. j One implementation of this is such that, for each failure i, the reliability factor u(i) satisfies:

[0038]

[0039] When the collected failure data is failure intensity, assuming the failure data follows a non-homogeneous Poisson distribution (NHPP), and the time interval [0,t] is uniformly divided into k parts, where n(i) is the number of failures per unit time, i = 1…k, then the reliability factor u(k) satisfies:

[0040]

[0041] As t increases, k also increases, but the unit time length remains unchanged.

[0042] When the collected failure data is the cumulative number of failures, assuming the failure data follows a non-homogeneous Poisson distribution (NHPP), and let N(k) be the cumulative number of failures up to the k-th time unit (inclusive), then the reliability factor u(k) satisfies:

[0043]

[0044] Step 2: Determine the data trend based on the significance level under the normal distribution and the reliability factor of the calculated failure data.

[0045] In practical applications, the significance level under a normal distribution is generally considered. When the significance level is 5%, all i or k satisfy:

[0046] When |u(k)| or |u(i)| < 1.96, the trend is considered stable;

[0047] When |u(k)| or |u(i)| < 1.96 is not satisfied: if u(k) or u(i) < 1.645, the reliability is considered to have increased; if u(k) or u(i) > -1.645, the reliability is considered to have decreased.

[0048] Step 3: Evaluate the reliability of the software based on multiple software reliability models.

[0049] Analyze the patterns in failure data collected during software reliability testing to assess software reliability. Make hypotheses about software failure behaviors, analyze the failure data using statistical methods, and provide software reliability assessment results.

[0050] Software reliability models include: the Jelinski-Moranda model (JM model), the Goel-Okumoto model (GO model), and the Duane model. Among them:

[0051] The Jelinski-Moranda model (JM model) is a reliability model developed by Jelinski and Moranda. It is one of the earliest established software reliability models and was used in the McDonnell Douglas Naval Engineering program. The failure interval t is considered to follow an exponential distribution, with the parameter of the exponential distribution proportional to the number of residual errors in the software. Specifically, the average failure interval at time t is 1 / φ(N-(i-1)), where t is any time between the (i-1)th failure and the ith failure. The quantity φ is a proportionality constant, and N is the total number of errors in the software since the start of observation.

[0052] The Goel-Okumoto model (GO model) was first proposed by Amrit Goel and Kazu Okumoto in 1979. It became the basis for a set of models that use the number of errors observed per unit time. Many other models also originated from the GO model, such as the S-type model.

[0053] The Duane model was originally applied to hardware reliability. First discovered by Duane while working at General Electric, it was later completed by Crow. It is now used to predict the reliability of software systems, and various software reliability assessment methods have been developed based on its results.

[0054] Step 4: Evaluate the predicted quality of each of the multiple software reliability models based on the relative accuracy of the software reliability predictions.

[0055] Two different reliability growth models made different predicted sequences for the same data source. The relative accuracy of the model predictions was assessed by calculating the sequence likelihood ratio.

[0056] For a specific piece of software, its next failure time during operation is random, following a real distribution, but this actual distribution is unknowable. Software reliability modeling involves making assumptions about this distribution and then making predictions. Different models make different assumptions about this distribution and thus arrive at different predictions. Comparing the relative quality of predictions from two models means comparing the reasonableness of their assumptions about this distribution and the relative accuracy of their predictions.

[0057] For a specific piece of software, after j-1 failures have occurred, the time T for the j-th failure to occur is... j It follows a true distribution, and we assume that the failure rate of this distribution is f. j (t), two different models A and B, based on their own assumptions and the preceding j-1 failure times, each apply f to f. j (t) made a prediction. Subsequently, during software operation, the time t at which the j-th failure occurred was observed. j , t j T is a random variable that follows an unknown true distribution. j From a probabilistic perspective, the implementation of t j It should be located at f j (t) larger areas.

[0058] t j Substitute into and Get it from the middle and if but Closer to f j The maximum value of (t), or in other words with f j (t j Model A is closer to reality than Model B, and its predictions are of higher quality.

[0059] Therefore, if model A predicts better quality than model B, then there should be It tends to be greater than 1. In other words, if but Closer to f j If the value of (t) is large, then model A will predict better quality; if but Closer to f j If the value of (t) is large, the B model will predict better quality; if Therefore, the predicted quality of Model A and Model B is the same.

[0060] The train network control system software reliability assessment method of this invention performs trend analysis on the failure data after obtaining it, but before selecting a specific model. The analysis results can serve as guidance for model selection and the use of effective data. Trend analysis of failure data can preliminarily screen model types, which can greatly improve modeling efficiency and avoid blind selection to some extent.

[0061] The train network control system software reliability assessment method of the present invention, after assessing the reliability of the software, also needs to evaluate the predicted quality of the model based on its expected effectiveness. This allows for the acquisition of the optimal reliability assessment result based on the rationality and relative accuracy of the reliability assessment.

[0062] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for evaluating software reliability of a train network control system, characterized by, Includes the following steps: Step 1: Calculate the reliability factor based on the collected failure data; Step 1 is as follows: When the collected failure data is failure interval time, set For one implementation of the random variable T j For each failure i, the reliability factor u(i) satisfies: When the collected failure data is failure intensity, assuming the time interval [0, t] is uniformly divided into k parts, and n(i) is the number of failures per unit time, i = 1…k, then the reliability factor u(k) satisfies: When the collected failure data is the cumulative number of failures, let N(k) be the cumulative number of failures before the k-th time unit, then the reliability factor u(k) satisfies: ； Step 2: Determine the data trend based on the significance level under the normal distribution and the calculated reliability factor of the failure data; Step 3: Evaluate the reliability of the software based on multiple software reliability models; Step 4: Evaluate the predicted quality of multiple software reliability models based on the relative accuracy of the software reliability predictions.

2. The train network control system software reliability assessment method according to claim 1, characterized in that, Step 2 specifically involves the following: When the significance level under a normal distribution is 5%, if |u(k)| or |u(i)| < 1.96, the trend is considered stable. When |u(k)| or |u(i)| < 1.96 is not satisfied: if u(k) or u(i) < 1.645, the reliability is considered to have increased; if u(k) or u(i) > -1.645, the reliability is considered to have decreased.

3. The train network control system software reliability assessment method according to claim 1, characterized in that, The software reliability model in step 3 is either the Jelinski-Moranda model, the Goel-Okumoto model, or the Duane model.

4. The method for evaluating the reliability of train network control system software according to claim 1, characterized in that, Step 4 is as follows: After j-1 failures occur, the time of the j-th failure is... It follows a true distribution, and the failure rate of this true distribution is... Model A and Model B respectively, based on the previous j-1 failure times, Make predictions separately, and the predicted results are as follows: and ; During software operation, the time of occurrence of the j-th failure was observed. ,Will Substitute into and Go to the middle, get and ; if > ,but Closer to If the value of is large, then Model A will predict better quality. if < ,but Closer to If the value of is large, then the B model will predict better quality. if = If the predicted quality of Model A and Model B is the same, then Model A and Model B are of the same quality.

Images