Comparative analysis of the data analysis capabilities of Weibull analysis tools[Copy link]
This post was last edited by lynnyong on 2018-7-5 09:53
Comparative analysis of data analysis capabilities of Weibull analysis tools
Abstract:This paper selects different types of data (including complete failure data, right-censored data) and different distribution types to compare and analyze the data analysis capabilities of mainstream Weibull analysis tools such as the Weibull analysis module of the PosVim platform, the Weibull++ module of the Reliasoft platform, and the reliability data analysis module of Minitab. According to the analysis results, the calculation capabilities of the above three software tools are not much different. The results calculated using the maximum likelihood method, the least squares method and other methods are basically consistent, with an error of <0.01%, which can meet the engineering requirements. However, from the perspective of reliability engineering application convenience and comprehensive analysis, the Weibull analysis module of the PosVim platform and the Weibull++ module of the Reliasoft platform are more suitable.
1. Analysis case 1: Complete failure data + Weibull distribution
The failure data of a product reliability test are as follows: 16, 34, 53, 75, 93, 120, in hours.
Weibull distribution is now selected as the distribution type for fitting, and the calculation methods are respectively selected as the maximum likelihood method and the least squares method, with a confidence level of 0.9 (bilateral). The Weibull analysis module of the PosVim platform, the Weibull++ module of the Reliasoft platform, and the reliability data analysis module of Minitab are used to calculate the above data. The calculation results are shown in the following table.
Result analysis: When the maximum likelihood method is selected, the shape parameter β and scale parameter η calculated by the three tools are basically consistent, with an error of <0.01%. When the least squares method is selected, the calculation results of the Weibull analysis module and the Weibull++ module of the PosVim platform are consistent, and the error with the calculation results of Minitab is also <0.01%, which can meet the engineering requirements. Table 1 Comparison of calculation results of Case 1 (maximum likelihood method) Input data 16, 34, 53, 75, 93, 120 Whether censored or not Censored or not Uncensored or not Censor ... [tr][td=1,4,112]
Figure 4 PosVim's case 1 calculation results (least squares method)
[align= Figure 5 Minitab’s calculation results for Case 1 (least squares method) 2. Analysis Case 2: Missing Data + Wei Boolean
The reliability test data of a fan system is shown in the following table. There are 70 test data records in total, of which the + is censored data (that is, the product has not failed at the end of the test).
Now select Weibull distribution, 0.95 confidence level (two-sided), and maximum likelihood method, and use the Weibull analysis module of the PosVim platform, the Weibull++ module of the Reliasoft platform, and the reliability data of Minitab respectively. The analysis module performs calculations. The calculation results are shown in the following table.
Result analysis: When the maximum likelihood method is selected, the shape parameters β and scale calculated by the three tools are The parameter η is basically consistent, with an error of <0.01%. Table 3 Comparison of calculation results of Case 2 (maximum likelihood method)
[/align Figure 6 Calculation results of Case 2 in PosVim (maximum likelihood method) Figure 7 Minitab's calculation results for Case 2 (maximum likelihood method) 3. Analysis case 3: Complete failure data + exponential distribution
After a certain product was tested for reliability, the failure data obtained were as follows: 7, 12, 19, 29, 41, 67.
Now select exponential distribution, 0.90 confidence level (two-sided), use maximum likelihood method, least squares method, Weibull analysis module of PosVim platform, Weibull++ module of Reliasoft platform, The reliability data analysis module of Minitab is used for calculation. The calculation results are shown in the following table.
Result analysis: When the maximum likelihood method is selected, the three tools calculate The failure rate and mean are basically consistent, with an error of <0.01%. The calculation results of the Weibull analysis module of the PosVim platform are consistent with those of the Weibull++ module. Table 4 Comparison of calculation results of Case 3 (maximum likelihood method) Input data
[/td][/tr] [tr][td=3,1,336]
7, 12, 19, 29, 41, 67
[/td][td=109]
Whether to censor
[/td][td=109]
No censoring
[/td][/tr] [ tr][td=5,1,553]
[/td][/tr] [tr][td=1,4,112]
Calculation Settings
[/td][td=112][align=right ]
Computational software
[/td][td=112]
PosVim's Weibull analysis module
[ /td][td=109]
Minitab
[/td][td=109]
Weibull++[/align ]
[/td][/tr] Calculation method However, maximum likelihood
Figure 8 PosVim's case 3 calculation results (maximum likelihood) 362440 Figure 9 Minitab Calculation results of case 3 (maximum likelihood) 362441 Figure 10 PosVim case 3 Calculation results (least squares method) 362442 Figure 11 Minitab’s calculation results for Case 3 (Least Squares Method)
4. Analysis Case 4: Analysis of a Company's Product Test Data
The author of this article was commissioned by a company to analyze the performance of a product (YY- CC) test data are analyzed and its B10 life is calculated. The 15 samples of this product were tested for 300 cycles. During the test, the resistance of the product was measured to see if it exceeded the standard. If the resistance exceeded the standard, the product was considered to be invalid. The test results of this product are as follows: 300+, 239, 300+, 288 , 137, 300+, 258, 175, 300+, 207, 102, 153, 276, 279, 300+. The data marked 300+ means that the sample has not failed after 300 cycles of testing. Now select Weibull distribution, 0.90 confidence level (two-sided), use maximum likelihood method, least squares method, Weibull analysis module of PosVim platform, and reliability data analysis module of Minitab for calculation. Calculation results As shown in the following table. Result analysis: When the maximum likelihood method and the least squares method are selected, the shape calculated by the Weibull analysis module of the PosVim platform and the reliability data analysis module of Minitab is The parameters and scale parameters are basically consistent, with an error of <0.01%. Table 6 Comparison of calculation results of Case 4 (maximum likelihood method) Input data 300+、239、300+、288、137、300+ , 258, 175, 300+, 207, 102, 153, 276, 279, 300+
[/td][td=109]
Whether missing
[/td][td=109]
The data marked with + are right censored data
[/td][/tr ] [tr][td=5,1,553]
[/td][/tr] [tr][td=1,4,112]
Calculation settings
[/td][td=112]
Calculation software
[/td][td= 112]
PosVim's Weibull analysis module
[/td][td=109]
Minitab
Weibull++ Calculation method However, maximum likelihood
[/td][/tr] [/table] Figure 12 Calculation results of Case 4 in PosVim (maximum likelihood-distribution fitting) Figure 13 Calculation results of Case 4 of PosVim (maximum likelihood-reliability parameter) Figure 14 Minitab's calculation results for Case 4 (maximum likelihood-distribution fitting) Figure 15 Calculation results of Case 4 of PosVim (least squares method - distribution fitting) Figure 16 Calculation results of Case 4 of PosVim (least square method - reliability parameter) Figure 17 Minitab’s calculation results for Case 4 (least squares method) 5 Conclusion =left] (1) Through the comparative analysis of the data calculation results of the above four cases, the calculation results of mainstream Weibull analysis related tools such as the Weibull analysis module of the PosVim platform, the Weibull++ module of the Reliasoft platform, and the reliability data analysis module of Minitab The results are basically consistent, and the calculation error is less than 0.01%, which meets the engineering accuracy requirements. (2) The Weibull analysis module of the PosVim platform and the Weibull++ module of the Reliasoft platform are better integrated with reliability engineering. It is the setting of calculation method and calculation parameters that better meets the requirements of reliability engineering. The Weibull analysis module of the PosVim platform and the Weibull++ module of the Reliasoft platform can calculate parameters such as BX% life, reliable life, conditional life (calculate its reliability when the product has been running for T hours), and can also calculate accelerated degradation test data. Supports commonly used exponential, Weibull, normal, base normal, logistic, gamma distribution, etc.
(3) In terms of data preprocessing, the Weibull analysis module of the PosVim platform also includes early Failure analysis, outlier detection, etc., are more in line with engineering requirements.
This content was originally created by EEWORLD forum user lynnyong, if you need to reprint or use Commercial use requires the author's consent and citation.
提升卡
变色卡
千斤顶
京公网安备 11010802033920号
