Numerical control machine tool reliability evaluation method based on multi-source discrete data fusion

Abstract

The invention discloses a numerical control machine tool reliability evaluation method based on multi-source discrete data fusion, and belongs to the field of machine tool reliability evaluation. And determining the quantile of the average fault interval time of the target machine tool by combining the average fault interval time of the similar numerical control machine tools, and inversely solving the parameter distribution interval of the target machine tool on the basis. Secondly, environment factors are introduced to convert reliability test stage data and discrete fault data of similar machine tools into external field stage data; and finally, solving and evaluating the distribution parameters of the discrete fault data in the external field stage by using a least square method, thereby completing fusion of multi-stage source data of the numerical control machine tool, and solving the problems of small number of samples and inaccurate evaluation of the numerical control machine tool.

Description

Technical Field

[0001] The invention relates to a reliability modeling and evaluation method for a numerically controlled machine tool, and belongs to the technical field of reliability evaluation of numerically controlled machine tools. Background Art

[0002] CNC machine tools are a typical electromechanical and hydraulic coupling equipment with high reliability, long life and maintainability. As the foundation and core manufacturing equipment of advanced manufacturing technology, they are increasingly widely used in mechanical production and restrict the development of my country's manufacturing field and high-tech. Although my country's domestic CNC machine tools have shortened the gap with the international advanced level in terms of reliability, precision and multi-axis linkage, the reliability index is still two times the international top level, which seriously affects the product reputation of domestic CNC machine tools and the competitiveness of domestic and foreign markets, resulting in domestic users not daring to use, not wanting to use, and not being able to use them. CNC machine tools are expensive. They are usually key equipment owned by enterprises in key industrial fields of the national economy. Their processing objects are often the core components of user products. Due to the low reliability of machine tools, they often cause failures and downtime, causing huge economic losses to users. Fully understanding the reliability of CNC machine tools is the premise for optimizing design, especially for new models. How to fully combine the reliability data of CNC machine tools at various stages to obtain real reliability evaluation indicators with high confidence is an urgent need of the industry and has great strategic significance.

[0003] The reliability of CNC machine tools is an important indicator to measure the quality of machine tools. The reliability of machine tools directly affects the processing quality, productivity and efficiency, and further affects the market competitiveness and user confidence. Therefore, CNC machine tools with improved reliability are urgently needed by the manufacturing industry. The reliability assessment of CNC machine tools is a key step in the reliability design of machine tools. Using the reliability assessment technology of machine tools, we can explore the weak links of CNC machine tools and carry out targeted optimization design. The existing multi-source information fusion method does not fully consider and utilize the reliability test information, resulting in incomplete information mastered in the expert evaluation process, and the mean time between failures (MTBF) and the percentile value given are quite different. In addition, the reliability test stage also belongs to the field test, but it still has differences with the actual processing process of the machine tool. Specifically, the reliability test stage belongs to the small batch processing process, while the field processing stage belongs to the large batch production process. The processing environment faced by the two stages is completely different. In the existing reliability evaluation process, the data of the reliability test stage is often ignored, resulting in the failure to truly reflect the actual reliability performance of the machine tool. Another point that needs to be noted is that in the actual machining process, the machine tool has often entered the "accidental failure stage" in the middle of the bathtub curve, and it is very likely that very few samples (1 to 4 data) or no fault samples will appear during the tracking time.

[0004] Based on the above background, combined with engineering research, it is found that there is an urgent need to propose a reliability assessment method that is easy to calculate, comprehensive in information, simple in process and easy to program for the current CNC machine tool assessment work. Therefore, this patent proposes a new CNC machine tool reliability assessment method based on multi-source discrete data fusion to solve the small sample problem in the assessment process, which specifically includes two important steps:

[0005] First, the multi-stage discrete reliability data collection stage;

[0006] The authenticity and availability of discrete fault data is an important prerequisite for ensuring the accuracy of reliability assessment results. The multiple stages include: reliability design stage, reliability test stage, and field test stage. The reliability design stage mainly collects what improvements have been made to the CNC machine tool on the original basis during the improvement and optimization process; the reliability test stage mainly collects the performance data and discrete fault data of the CNC machine tool during the simulated field operation stage, such as continuous idling test, loading test, static stiffness test, enhanced idling test, safety reliability test, and various precision indicators; the second is to collect discrete fault data that occurred during the small batch production stage; the field test stage mainly collects discrete fault data that occurred during the historical operation of similar function machine tools and discrete fault data of the target CNC machine tool to the current time.

[0007] Second, environmental factors are introduced to convert discrete data under different environments in multiple stages into the fault sample set of the target machine tool for reliability modeling and evaluation;

[0008] The MTBF percentile values ​​provided by the comprehensive experts are used to solve the parameter distribution range of the target machine tool, and the F statistic is used to test whether the failure mechanism of discrete fault data at different stages is consistent. Environmental factors are introduced to convert the discrete fault data at different stages into the fault data sample set of the target machine tool, and the reliability evaluation of the target machine tool is completed through the least squares method. Summary of the invention

[0009] The purpose of the present invention is to provide a new reliability evaluation method for CNC machine tools based on multi-source discrete data fusion. First, a number of experts obtain the MTBF value of the target machine tool at 1 / 4 and 3 / 4 based on the test data of the reliability test stage and the discrete fault data of similar machine tools, and then inversely solve the parameter distribution interval of the target machine tool. After that, the F statistic is used to test whether the discrete fault data of the reliability test stage and the discrete fault data of similar machine tools are the same as the failure mechanism of the target machine tool. On this basis, the environmental factor is introduced to convert the discrete fault data of multiple stages into the discrete fault data sample set of the field stage. Finally, the distribution parameters of the target machine tool are solved based on the least squares method, and its reliability index is calculated.

[0010] The technical solution adopted by the present invention is a new CNC machine tool reliability assessment method based on multi-source discrete data fusion. This method determines the percentile value of the average failure interval time of the target machine tool based on the test data of the reliability test phase and the average failure interval time of similar CNC machine tools by multiple experts, and then inversely solves the parameter distribution interval of the target machine tool on this basis. Afterwards, the environmental factor is introduced to convert the data of the reliability test phase and the discrete failure data of similar machine tools into the discrete failure data of the field stage. Finally, the least squares method is used to solve and evaluate the distribution parameters of the discrete failure data of the field stage, so as to complete the fusion of the multi-stage source data of the CNC machine tool and solve the problem of small number of samples and inaccurate evaluation of CNC machine tools.

[0011] The specific steps include:

[0012] Step 1: Collect discrete fault data from the historical operation of similar machine tools and solve the distribution parameters of similar machine tools by the least squares method;

[0013] Through literature research, it is found that most of the current CNC machine tools obey the Weibull distribution. This patent also assumes that the machine tools obey the Weibull distribution, and its specific expression is shown in formula (1).

[0014]

[0015] Where t represents the time variable, t≥0; η represents the scale parameter, η≥0; α represents the shape parameter, α≥0;

[0016] The least squares method is used to estimate and solve the parameters of the assumed distribution model. The specific solution steps are as follows: After transposing formula (1), we can get Taking logarithms twice on both sides of the equation gives make Further get the univariate linear regression equation Y l =αX l -αlnη, we can get

[0017] At the significance level γ = 0.05, the test formula (2) is used to test the goodness of fit of the assumed Weibull distribution model. If D n <D n,γ , then the test passes and the null hypothesis is accepted, otherwise the null hypothesis is rejected;

[0018] D=max|F s (t)-F(t)| (2)

[0019] Where F(t) represents the cumulative failure distribution function of the assumed distribution; F s (t) represents the empirical distribution function, which is specifically established by formula (3).

[0020]

[0021] Wherein, i=1,2,…,N; N represents the number of fault intervals.

[0022] Finally, the MTBF value of similar machine tools is solved by formula (4).

[0023]

[0024] Step 2: Based on the test results of the reliability test phase and the MTBF of similar machines obtained in step 1, multiple experts evaluate the MTBF quantiles of the target machine tool at 0.25 and 0.75.

[0025] GB / T 23567.1-2009 "Reliability Assessment of CNC Machine Tools: Part 1 General Principles" lists the various tests and tolerance values ​​that need to be conducted during the reliability test phase, and collects the test reports of the above phases as the basis for expert judgment. According to literature research, the most frequently asked questions for experts are the percentiles of the MTBF of CNC machine tools at 1 / 4 and 3 / 4. Therefore, this patent ultimately chooses to invite experts to judge the percentiles of the MTBF of the target machine tool at 0.25 and 0.75. The specific meaning of the percentiles is that for the reference machine tool, the probability that its failure interval is less than 0.25 times the MTBF is 25%, and the probability that its failure interval is less than 0.75 times the MTBF is 75%.

[0026] The calculation process of the two quantile values ​​is to derive the inverse function of time t based on formula (1), as shown in formula (5).

[0027]

[0028] Wherein, t is the MTBF of the target machine tool at the current percentile value; P is the input percentile value, which is 0.25 and 0.75 in this patent.

[0029] Due to the subjective factors in expert evaluation, it is necessary to fully consider multiple factors such as the expert's professional level, working time, academic level, scientific research title, and professional field status to set reasonable weight coefficients, and finally obtain the MTBF estimation interval of the target machine tool at the quantile value of 0.25 and the MTBF estimation interval at the quantile value of 0.75.

[0030] Step 3: Estimation interval combination and inverse parameter estimation interval;

[0031] Combine the upper and lower limits of the estimated interval of the quantile value 0.25 with the upper and lower limits of the estimated interval of the quantile value 0.75, respectively, and substitute the combined upper and lower limits into formula (5) to solve the shape parameter and scale parameter of each combined distribution function, and obtain the estimated intervals of Weibull distribution parameters α and η.

[0032] Step 4: Solve the posterior parameter distribution in the reliability test phase based on the prior information;

[0033] The estimated intervals of the Weibull distribution parameters α and η obtained in step 3 are used as the prior information of the Bayesian method. Combined with the discrete failure data in the reliability test phase, the Markov Chain Monte Carlo (MCMC) method based on Gibbs sampling is used to solve the posterior distribution parameters of the target machine tool. Since this method involves deep statistical theory, in order to facilitate the actual use of engineers, the free and open Win BUGS software can be used to complete step 3. The technicians can solve the parameters by simply learning through the search engine.

[0034] Step 5: Use F statistics to test whether the discrete fault data and discrete similar fault data in the reliability test phase obey the same fault failure distribution;

[0035] Steps 1 and 4 obtain the shape parameter α of the discrete fault data of similar machine tools and the discrete fault data of the reliability test phase S and α L , using statistics Check whether the discrete failure data of similar machine tools and the discrete failure data of the reliability test phase obey the same failure distribution. Assume that H 0 The failure distribution of the two stages is the same, assuming that H 1 The failure distributions of the two stages are different, that is,

[0036] H 0 :α S =α L ,H 1 :α S ≠α L .

[0037] According to the number of similar machine tool discrete fault data r S and the number of discrete failure data in the reliability test phase r L , check F through the "Reliability Test Table (Revised Edition)" 0.95 (r1,r2) and F 0.05 (r1,r2)=1 / F 0.95 (r1, r2), we get the interval acceptance domain W of the statistic T ={F0.05 (r1,r2)<F<F 0.95 (r1,r2)} and the rejection region W F ={F≤F 0.05 (r1,r2) or F ≥ F 0.95 (r1, r2)}. If the statistic does not fall into the rejection region WF, the null hypothesis H is accepted. 0 , if it falls into the rejection region W F In the middle, the original hypothesis H is not accepted 0 , accept hypothesis H 1 .

[0038] Step 6: Introduce environmental factors to complete the conversion of discrete fault data at different stages

[0039] For machine tools whose life follows Weibull distribution, the environmental factor expression of the discrete failure data in the similarity stage and the discrete failure data in the reliability test stage converted to the discrete failure data in the field stage is shown in formula (7).

[0040]

[0041] Where η is the scale parameter of the distribution function of the similarity stage, reliability test stage, and field stage.

[0042] The conversion process mainly includes two steps: 1) Calculate the environmental factor K of discrete failure data in the reliability test phase and discrete failure data of similar machine tools 1 , the failure data of the reliability test phase is transmitted through t 一次扩充 =K 1 ×Data 试验 Convert to the field data to complete the first sample expansion; 2) The first expanded fault data sample set is recalculated through the method of step 4 to recalculate the distribution parameter α after the first expansion E and η E ; Then use the test in step 5 to check whether the distribution parameters of the first expansion distribution are consistent with the failure distribution of similar failure data; after the F statistic test, calculate the environmental factor K by formula (7) 2 , completing the second conversion and expansion of the discrete fault data sample set.

[0043] Step 7: Use the least square method to perform reliability assessment on the discrete fault data after quadratic conversion;

[0044] Combined with the second expanded discrete fault data, use formula (1) to formula (4) described in step 1 to re-solve the distribution parameter α of the target machine tool F , η F and MTBF indicators.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The reliability assessment method for CNC machine tools provided by the present invention fully considers discrete fault data including the design stage, reliability test stage, and operation process of similar equipment of CNC machine tools, and proposes a relatively simple and easy-to-program theoretical fusion method to convert fault data in other stages or other environments into the fault sample set of the target machine tool, thereby solving the problem of inaccurate assessment caused by the scarcity of samples in the assessment process, and providing a scientific assessment method for realizing rapid assessment of machine tool reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart of the implementation of the method of the present invention. DETAILED DESCRIPTION

[0048] The present invention takes a certain improved CNC milling machine as an example to verify the above CNC machine tool reliability assessment method based on multi-source discrete data fusion. Specifically, it includes the following steps:

[0049] Step 1: Collect discrete fault data from the historical operation of similar CNC milling machines 相似 ={83.5,102.5,124,138,158.5,184.5,202,255,333,394,435,479,582,603}, and the MTBF of the similar CNC milling machine design stage is known to be 300h, the MTBF of the target CNC milling machine design stage is 650h, and the historical discrete fault data of the target CNC milling machine is collected. 外场 ={46,83.5,255,333}, and the distribution parameters of similar machine tools are solved by the least square method;

[0050] Through literature research, it is found that most of the current CNC machine tools obey the Weibull distribution. This patent also assumes that the machine tools obey the Weibull distribution, and its specific expression is shown in formula (1).

[0051]

[0052] Where t represents the time variable, t≥0; η represents the scale parameter, η≥0; α represents the shape parameter, α≥0;

[0053] The least squares method is used to estimate and solve the parameters of the assumed distribution model. The specific solution steps are as follows: After transposing formula (1), we can get Taking logarithms twice on both sides of the equation gives make Further get the univariate linear regression equation Y l =αX l -αlnη, which can be simplified to Data 相似 After the above steps, we can get α S =1.4506,η S =137.3013.

[0054] At the significance level γ = 0.05, the test formula (2) is used to test the goodness of fit of the assumed Weibull distribution model. If D n <D n,γ , then the test passes and the null hypothesis is accepted; otherwise, the null hypothesis is rejected.

[0055] D=max|F s (t)-F(t)| (2)

[0056] Where F(t) represents the cumulative failure distribution function of the assumed distribution; F s (t) represents the empirical distribution function, which is specifically established by formula (3).

[0057]

[0058] Wherein, i=1,2,…,N; N represents the number of fault intervals.

[0059] Knowable Data 相似 The data sample size is 14, D 14,0.05 =0.4356, maximum value D 14 =0.1452, D n <D n,γ , through testing, the null hypothesis is accepted, that is, the Weibull distribution is accepted as an appropriate distribution model for the target CNC machine tool.

[0060] Finally, the MTBF value of similar machine tools is solved by formula (4). S =1.4506,η S =137.3013 is substituted into formula (4), MTBF≈295h, which is close to the designed MTBF=300h. The relative error is 1.67%. Because the machine tool will be affected by environmental and human factors during operation, this error is within the normal acceptable range.

[0061]

[0062] Step 2: Based on the test results of the reliability test phase and the MTBF of similar machines obtained in step 1, multiple experts evaluate the MTBF quantiles of the target machine tool at 0.25 and 0.75.

[0063] GB / T 23567.1-2009 "Reliability Assessment of CNC Machine Tools: Part 1 General Principles" lists the various tests and tolerances that need to be carried out during the reliability test phase. The test reports of the above phases are collected. The reliability test of the target machine tool adopts a simulated field test plan, and its test results mainly include the following 7 items:

[0064] (1) The test results of geometric accuracy index and performance index are shown in Table 1.

[0065] Table 1 Geometric accuracy index and performance index test results

[0066]

[0067]

[0068] (2) Spindle axis loading test

[0069] The spindle axis loading test includes X / Y / Z three-axis loading. During the loading test, the target machine tool had no obvious faults and worked normally.

[0070] (3) Static stiffness test. As shown in Table 2, the test results were compared with the machine tool design manual and found that they did not exceed the design requirements.

[0071] Table 2 Static stiffness test results

[0072]

[0073] (4) Torque and power tests: The maximum torque of the target CNC milling machine under the S1 working mode was measured to be 154 N·m. The maximum torque of the spindle under the S1 working mode given in the machine tool design manual is 150 N·m, which does not exceed the design requirements.

[0074] (5) Typical specimen cutting test: In accordance with GB / T 23567.1-2009 "Reliability Assessment of CNC Machine Tools: Part 1 General Principles" and its extended standards, the target CNC milling machine was used to cut standard parts. The roundness and verticality of the center hole, the parallelism, verticality, inclination and straightness of the quadrilateral and rhombus, the roundness, the coaxiality of the hole and other indicators were checked according to the tolerance values ​​given in the standard. All indicators were within the tolerance range given in the standard, and the discrete fault data during operation were recorded.

[0075] (6) Strengthen the idle running test, consider the processing technology of the target milling machine, increase the moving speed, rotation speed, tool change and drilling and milling number of processes of the target milling machine, write the corresponding CNC program, and record the discrete fault data during the operation.

[0076] (7) Safety and reliability test: refer to relevant standards to carry out safety protection device inspection, operation function inspection, surge resistance inspection, limit stroke inspection and other test items of the target CNC milling machine, and the test process is qualified.

[0077] The above 7 test data are compiled into a book, and the discrete failure data of 2 failures that occurred during the reliability test phase are collected. 试验阶段 ={102.5,219}, which is used as the basis for expert judgment. According to literature research, the question that experts are most often asked to answer is the quantile values ​​of the MTBF of CNC machine tools at 1 / 4 and 3 / 4. Therefore, experts are invited to judge the quantile values ​​of the MTBF of the target machine tools at 0.25 and 0.75. The specific meaning of the quantile value is that for the reference machine tool, the probability that its failure interval time is less than 0.25 times the MTBF is 25%, and the probability that its failure interval time is less than 0.75 times the MTBF is 75%.

[0078] The calculation process of the two quantile values ​​is to derive the inverse function of time t based on formula (1), as shown in formula (5).

[0079]

[0080] Wherein, t is the MTBF of the target machine tool at the current percentile value; P is the input percentile value, which is 0.25 and 0.75 in this patent.

[0081] Since expert evaluation involves subjective factors, it is necessary to fully consider the expert's professional level, working time, academic level, scientific research title, professional field status and other factors to set a reasonable weight coefficient. Four industry experts were invited to evaluate the quantile value of the target machine tool. To facilitate the understanding of potential technicians, it is assumed that the professional level, working time, academic level, scientific research title, and professional field status of the four experts are the same, that is, the weight value is 25%. According to the MTBF of similar CNC milling machines, the MTBF estimation interval of the target machine tool at the quantile value of 0.25 and the MTBF estimation interval at the quantile value of 0.75 are obtained, as shown in Table 3.

[0082] Table 3 MTBF estimation intervals at 0.25 and 0.75 for target CNC milling machine

[0083]

[0084] The estimated interval at the 0.25 quantile is [152.5, 185], and the estimated interval at the 0.75 quantile is [390, 458.75]

[0085] Step 3: Estimation interval combination and inverse parameter estimation interval;

[0086] The upper and lower limits of the estimated interval of the quantile value 0.25 and the upper and lower limits of the estimated interval of the quantile value 0.75 are combined respectively {[153,390],[153,459],[185,390],[185,459]}. Substitute the combined upper and lower limits into formula (5) to solve the scale parameter η and shape parameter α of each combined distribution function, which are {[321.1,16806],[365.3,1.4314],[334,2.1085],[380.1,1.7305]} respectively, and the estimated intervals of the Weibull distribution parameters α and η are obtained as {[1.43,2.1085],[321.1,380.1]}.

[0087] Step 4: Solve the posterior parameter distribution in the reliability test phase based on the prior information;

[0088] The estimated intervals {[1.43, 2.1085], [321.1, 380.1]} of the Weibull distribution parameters α and η obtained in step 3 are used as the prior information of the Bayesian method. Combined with the discrete failure data in the reliability test phase, the Markov Chain Monte Carlo (MCMC) method based on Gibbs sampling is used to solve the posterior distribution parameters of the target machine tool. Since this method involves deep statistical theory, in order to facilitate practical engineering use, step 3 can be completed using the free and open Win BUGS software. The technicians can solve the parameters by simply learning through the search engine. The distribution parameter α of the reliability test phase is obtained. L =1.708 and η L =322.5.

[0089] Step 5: Use F statistics to test whether the discrete failure data in the reliability test phase and the discrete failure data of similar machine tools obey the same failure distribution;

[0090] Steps 1 and 4 obtain the shape parameter α of the discrete fault data of similar machine tools and the discrete fault data of the reliability test phase S and α L , using statistics Check whether the discrete failure data of similar machine tools and the discrete failure data of the reliability test phase obey the same failure distribution. Assume that H 0 The failure distribution of the two stages is the same, assuming that H 1 The failure distributions of the two stages are different, that is,

[0091] H 0 :α S =α L ,H 1 :α S ≠αL .

[0092] According to the number of discrete failure data of similar machine tools 14 and the number of discrete failure data of the reliability test phase 2, the F 0.95 (14,2)=3.74 and F 0.05 (14,2)=1 / F 0.95 (14,2)=0.2674, and we get the interval acceptance domain W of the statistic. T ={0.2674<F<3.74} and the rejection region W F ={F≤0.2674 or F≥3.74}. Substituting relevant parameters into the calculation, the F statistic is 0.8149, which does not fall into the rejection region W F In the above example, we accept the null hypothesis H 0 .

[0093] Step 6: Introduce environmental factors to complete the conversion of discrete fault data at different stages

[0094] For machine tools whose life follows Weibull distribution, the environmental factor expression of the discrete failure data in the similarity stage and the discrete failure data in the reliability test stage converted to the discrete failure data in the field stage is shown in formula (7).

[0095]

[0096] Where η is the scale parameter of the distribution function of the similarity stage, reliability test stage, and field stage.

[0097] The conversion process mainly includes two steps: 1) Calculate the environmental factors of discrete failure data in the reliability test phase and discrete failure data of similar machine tools The discrete failure data in the reliability test phase is transmitted through t 一次扩充 =K 1 ×Data 试验 Convert to field data to complete the first sample expansion Data 第一次扩充 ={46,83.5,241,255,333,516}; 2) The first expanded fault data sample set is recalculated through the method in step 4 to calculate the distribution parameter α after the first expansion E =1.823 and η E =351.3; then use the test in step 5 to check whether the distribution parameters of the first expansion distribution are consistent with the failure distribution of similar failure data. The F statistic is 0.7957, which does not fall into the rejection region W F The environmental factor K is further calculated by formula (7): 2 =2.5642, the sample set after the first expansion is passed through t 二次扩充=K 2 ×Data 第一次扩充 Complete the second conversion and expansion of discrete fault data sample set 第二次扩充 ={215,255,261,263,318,333,354,407,474,518,558,654,854,1011,1116,1229,1493,1547}.

[0098] Step 7: Use the least square method to perform reliability assessment on the discrete fault data after quadratic conversion;

[0099] Combined with the second expanded discrete fault data, use formula (1) to formula (4) described in step 1 to re-solve the distribution parameters of the target machine tool, α F =1.6713,η F =743.4451 and MTBF≈664h indicators, the relative error is 2.11%, and because the target CNC milling machine will also be affected by environmental and human factors during operation, the error is within the normal acceptable range.

[0100] From the analysis results, when there is little data on the target machine tool, based on the historical discrete failure data of similar machine tools and the discrete failure data of the target machine tool during the reliability test phase, the MTBF result obtained by the method proposed in this patent is equivalent to the actual value, and solves the defect of conservative estimation of the MTBF value of the machine tool by the previous method.

Claims

1. A reliability assessment method for CNC machine tools based on multi-source discrete data fusion, characterized in that: This method uses multiple experts to determine the percentile value of the mean failure interval of the target CNC machine tool based on the test data of the reliability test phase and the mean failure interval of similar CNC machine tools, and inversely solves the parameter distribution interval of the target CNC machine tool; then, the environmental factor is introduced to convert the data of the reliability test phase and the discrete failure data of similar CNC machine tools into the discrete failure data of the field phase; finally, the least squares method is used to solve and evaluate the distribution parameters of the discrete failure data of the field phase, thereby completing the fusion of multi-stage source data of CNC machine tools and solving the problems of small number of CNC machine tool samples and inaccurate evaluation.

2. A method for evaluating the reliability of a CNC machine tool based on multi-source discrete data fusion according to claim 1, characterized in that: The steps include: Step 1: Collect discrete fault data from the historical operation of similar CNC machine tools, and solve the distribution parameters of similar CNC machine tools by the least squares method; Step 2: Based on the test results of the reliability test phase and the MTBF of similar CNC machine tools obtained in step 1, multiple experts evaluate the quantile values ​​of the MTBF of the target CNC machine tool at 0.25 and 0.75; Step 3: Estimation interval combination and inverse parameter estimation interval; Step 4: Solve the posterior parameter distribution in the reliability test phase based on the prior information; The estimated intervals of the Weibull distribution parameters α and η obtained in step 3 are used as the prior information of the Bayesian method. Combined with the discrete failure data in the reliability test phase, the posterior distribution parameters of the target CNC machine tool are solved by the Markov chain Monte Carlo MCMC method based on Gibbs sampling. Use Win BUGS software to complete step 3, and simply learn through the search engine to solve the parameters; Step 5: Use F statistics to test whether the discrete fault data and discrete similar fault data in the reliability test phase obey the same fault failure distribution; Step 6: Introduce environmental factors to complete the conversion of discrete fault data at different stages; For CNC machine tools whose life follows Weibull distribution, the environmental factor expression of discrete failure data in the similar stage and discrete failure data in the reliability test stage converted to discrete failure data in the field stage is shown in formula (7); Where η is the scale parameter of the distribution function of the similarity stage, reliability test stage, and field stage; The conversion process includes two steps: 1) Calculate the environmental factor K1 of the discrete failure data in the reliability test phase and the discrete failure data of similar CNC machine tools, and convert the failure data in the reliability test phase into 一次扩充 =K1×Data 试验 Convert to the field data to complete the first sample expansion; 2) The first expanded fault data sample set is recalculated through the method of step 4 to recalculate the distribution parameter α after the first expansion E and η E ; Then, use the step 5 to check whether the distribution parameters of the first expansion distribution are consistent with the failure distribution of similar fault data; after passing the F statistic test, calculate the environmental factor K2 by formula (7) to complete the second conversion expansion of the discrete fault data sample set; Step 7: Use the least squares method to perform reliability assessment on the discrete fault data after quadratic conversion.

3. The reliability assessment method of a CNC machine tool based on multi-source discrete data fusion according to claim 2 is characterized in that: In step 1, most CNC machine tools obey the Weibull distribution, and CNC machine tools obey the Weibull distribution, and its specific expression is shown in formula (1); In the formula, t represents the time variable, t≥0; η represents the scale parameter, η≥0; α represents the shape parameter, α≥0; The least squares method is used to estimate and solve the parameters of the assumed distribution model. The specific solution steps are as follows: After transposing formula (1), we can obtain Taking logarithms twice on both sides of the equation gives make Get the univariate linear regression equation Y l =αX l -αlnη, we can get At the significance level γ = 0.05, the test formula (2) is used to test the goodness of fit of the assumed Weibull distribution model. If D n <D n,γ , then the test passes and the null hypothesis is accepted, otherwise the null hypothesis is rejected; D=max|F s (t)-F(t)| (2) Where F(t) represents the cumulative failure distribution function of the assumed distribution; F s (t) represents the empirical distribution function, which is established by formula (3); Where, i = 1, 2, ..., N; N represents the number of fault intervals; Finally, the MTBF value of similar CNC machine tools is solved by formula (4); 4. The reliability assessment method of a CNC machine tool based on multi-source discrete data fusion according to claim 2 is characterized in that: In step 2, experts are invited to judge the quantile values ​​of the MTBF of the target CNC machine tool at 0.25 and 0.75; the specific meaning of the quantile value is that for the reference CNC machine tool, the probability that its failure interval time is less than 0.25 times the MTBF is 25%, and the probability that its failure interval time is less than 0.75 times the MTBF is 75%; The calculation process of the two quantile values ​​is to derive the inverse function of time t based on formula (1), as shown in formula (5); Where, t is the MTBF of the target CNC machine tool at the current percentile value; P is the input percentile value, and the inputs are 0.25 and 0.75; Due to the subjective factors in expert evaluation, it is necessary to fully consider the expert's professional level, working time, academic level, scientific research title, and professional status to set the weight coefficient, and finally obtain the MTBF estimation interval of the target CNC machine tool at the quantile value of 0.25 and the MTBF estimation interval at the quantile value of 0.

75.

5. The method for evaluating the reliability of a CNC machine tool based on multi-source discrete data fusion according to claim 4 is characterized in that: In step 3, the upper and lower limits of the estimated interval of the quantile value 0.25 are combined with the upper and lower limits of the estimated interval of the quantile value 0.75, and the combined upper and lower limits are respectively substituted into formula (5) to solve the shape parameter and scale parameter of each combined distribution function, and obtain the estimated intervals of Weibull distribution parameters α and η.

6. The method for evaluating the reliability of a CNC machine tool based on multi-source discrete data fusion according to claim 4, characterized in that: Steps 1 and 4 obtain the shape parameter α of the discrete fault data of similar CNC machine tools and the discrete fault data of the reliability test phase S and α L , using statistics Check whether the discrete fault data of similar CNC machine tools and the discrete fault data of reliability test stage obey the same fault failure distribution; H0 means the fault failure distribution of the two stages is the same, H1 means the fault failure distribution of the two stages is different, that is H0:a S =a L ,H1:a S ≠a L According to the number of similar CNC machine tool discrete fault data r S and the number of discrete failure data in the reliability test phase r L , by consulting F 0.95 (r1,r2) and F 0.05 (r1,r2)=1 / F 0.95 (r1, r2), we get the interval acceptance domain W of the statistic T ={F 0.05 (r1,r2)<F<F 0.95 (r1,r2)} and the rejection region W F ={F≤F 0.05 (r1,r2) or F ≥ F 0.95 (r1,r2)}; if the statistic does not fall into the rejection region W F In the case of accepting the original hypothesis H0, if it falls into the rejection region W F That is, do not accept the null hypothesis H0 and accept the hypothesis H1.

7. The method for evaluating the reliability of a CNC machine tool based on multi-source discrete data fusion according to claim 3 is characterized in that: In step 7, combined with the second expanded discrete fault data, the distribution parameter α of the target CNC machine tool is re-solved using formulas (1) to (4) described in step 1. F , η F and MTBF indicators.

Images