Reliability modeling method based on mechanism data hybrid model
By adopting a reliability modeling method based on a mechanistic data hybrid model, the problems of large errors and high evaluation costs caused by insufficient experimental data are solved. It realizes dynamic correction and efficient evaluation under different data conditions and is applicable to equipment such as pumps and valves.
Patent Information
- Application Number
- CN202510962557.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-14
- Publication Date
- 2025-11-21
AI Technical Summary
Existing reliability modeling methods mostly rely on experimental data or pure mechanistic models, which leads to large errors when experiments are insufficient. They do not consider dynamic factors in the actual environment. Traditional pump reliability assessments do not incorporate the wear mechanism of seals, resulting in an overestimation of lifespan by 20%-30%. The assessment cycle for critical equipment such as nuclear power valves is long and costly.
A reliability modeling method based on a mechanistic data hybrid model is adopted. Through data collection, modeling path selection, parameter calculation and result output, the mechanistic model and the data-driven model are combined. The modeling path is selected according to the sufficiency of experimental data, and the prior data of the mechanistic model are integrated with the experimental data for correction. The Bayesian method is used for dynamic correction.
It enables dynamic correction of reliability models when experimental data is insufficient, improving the model's generalization ability and confidence interval accuracy, reducing evaluation costs, and is applicable to various types of equipment such as pumps, valves, and filters, thereby improving evaluation efficiency and accuracy.
Smart Images
Figure CN120995654A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, specifically to a reliability modeling method based on a mechanistic data hybrid model. Background Technology
[0002] Existing reliability modeling methods mostly rely on experimental data or pure mechanistic models.
[0003] Data-driven models suffer from significant errors when experiments are insufficient, while mechanistic models fail to consider dynamic factors in the actual environment. Traditional pump reliability assessments rely solely on fitting exponential distributions to experimental data, neglecting seal wear mechanisms, leading to an overestimation of lifespan by 20%-30%. Furthermore, existing technologies do not address correction issues arising from insufficient data, resulting in lengthy and costly assessment cycles for critical equipment such as nuclear power plant valves. Considering the need to verify whether pump system reliability indicators meet overall engineering requirements when experimental data is insufficient, and to address the coexistence of equipment unit mechanisms and data, a reliability modeling method based on a hybrid mechanistic-data model is necessary to resolve these technical problems.
[0004] Many systems rely heavily on experimental data or purely mechanistic models. Data-driven models have large errors when experiments are insufficient, while mechanistic models do not consider dynamic factors in the actual environment. Traditional pump reliability assessments only fit an exponential distribution to experimental data without considering the wear mechanism of seals, leading to an overestimation of lifespan by 20%-30%. The problem of correction when data is insufficient has not been addressed, resulting in long assessment cycles and high costs for critical equipment such as nuclear power plant valves. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a reliability modeling method based on a hybrid mechanistic-data model, solving the problem of relying heavily on experimental data or pure mechanistic models. Data-driven models suffer from large errors when experiments are insufficient, while mechanistic models do not consider dynamic factors in the actual environment. Traditional pump reliability assessments only fit an exponential distribution to experimental data without considering the wear mechanism of seals, leading to an overestimation of lifespan by 20%-30%. Furthermore, the problem of correction when data is insufficient is not addressed, resulting in long assessment cycles and high costs for critical equipment such as nuclear power plant valves.
[0006] To achieve the above objectives, the present invention provides a reliability modeling method based on a mechanistic data hybrid model, comprising the following steps:
[0007] Step 1: Data collection. Based on the equipment status and reliability assessment requirements, collect multi-dimensional data, including technical status, usage stage, environmental information, and fault records.
[0008] Step 2: Modeling path selection. The modeling path is selected based on the sufficiency of experimental data. When there is no experimental data, a mechanistic model is used. When there is sufficient experimental data, a data-driven model is used. When the experimental data is insufficient, the prior data of the mechanistic model is combined with the experimental data for correction.
[0009] Step 3: Parameter calculation. Calculate equipment reliability parameters by fitting the distribution type and using correction factors.
[0010] Step 4: Output the results, including the reliability model and index evaluation results.
[0011] Preferably, the data collection and classification includes: classifying fault data into associated faults and non-associated faults.
[0012] Preferably, the associated faults include design defects, component defects, and wear and tear faults; when counting the number of faults, dependent faults, planned disassembly, and minor defects are excluded; and the data is classified into unrepairable unit data and repairable unit data according to the repairability of the equipment.
[0013] Preferably, the modeling path selection includes the calculation process of the mechanistic model without experimental data and the statistical model with experimental data. The calculation process of the mechanistic model without experimental data includes:
[0014] The decomposition equipment is divided into independent components, and the failure rate of each component is calculated; using the formula λ P =λ SE +λ SH +λ BE +λ CA +λ FD The failure rate of the combined components is calculated, and a correction factor C is introduced. TLF C PS C C Adjust the overall failure rate;
[0015] For high-reliability components, empirical estimation is used, and the failure rate of related components is corrected by combining their design parameters;
[0016] The statistical models with experimental data include:
[0017] In the exponential distribution model, through Calculate the failure rate and combine it with confidence intervals. Assess the risks;
[0018] In the Weber distribution model, the shape parameter β and scale parameter α are calculated using the average rank method, and the distribution parameters are determined by linearly fitting ln[-lnR(t)] and lnt.
[0019] Preferably, the modeling path selection further includes a data fusion method, the data fusion method comprising:
[0020] When experimental data is insufficient, the output of the mechanistic model is used as the prior distribution, and the posterior distribution is updated by combining the experimental data with Bayesian methods.
[0021] The general data processing module converts the general data of the mechanism model into the prior distribution type of the reliability parameters.
[0022] Preferably, the calculation of the correction factor includes:
[0023] Load correction factor Where k1 is the material property coefficient;
[0024] Speed correction factor Where k2 is the bearing wear coefficient.
[0025] Preferably, the output reliability model and index evaluation results include:
[0026] Input design parameters: rotation speed 2000 rpm, seal material is fluororubber, bearing type is deep groove ball bearing;
[0027] The failure rate of each component is calculated by using a mechanistic model, and the error rate is obtained by combining the correction factor with the total failure rate.
[0028] Preferably, the method calculates the device reliability model differently depending on whether experimental data is available or not, and uses a general data processing module to generate a priori distribution of the corresponding reliability parameters using this general data.
[0029] This invention discloses a reliability modeling method based on a mechanistic data hybrid model, which has the following beneficial effects:
[0030] This invention models the model under two conditions: no experimental data and experimental data. Based on the sufficiency of experimental data, it designs multiple technical paths, including cases with no experimental data, insufficient experimental data, and sufficient experimental data, thereby achieving decoupling between reliability modeling and the amount of experimental data.
[0031] By integrating data and mechanisms, we can address the problem of insufficient experimental data and improve the generalization ability of models.
[0032] It can be dynamically corrected, and the accuracy of confidence intervals is improved by fusing prior and experimental data through Bayesian methods;
[0033] It has broad industry adaptability, supports various types of equipment such as pumps, valves, and filters, and is highly versatile. Attached Figure Description
[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0035] Figure 1 A flowchart of a reliability modeling method based on a mechanistic data hybrid model, as provided in an embodiment of the present invention;
[0036] Figure 2 The reliability model selection diagram based on the mechanism-data hybrid model, as described in one embodiment of the present invention, is used for reliability model calculation with experimental data. Detailed Implementation
[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0038] This application provides a reliability modeling method based on a hybrid mechanistic-data model, addressing the issues of reliance on experimental data or purely mechanistic models. Data-driven models suffer from large errors when experiments are insufficient, while mechanistic models fail to consider dynamic factors in the actual environment. Traditional pump reliability assessments only fit an exponential distribution to experimental data without considering the wear mechanism of seals, leading to an overestimation of lifespan by 20%-30%. Furthermore, it fails to address the correction problem when data is insufficient, resulting in long assessment cycles and high costs for critical equipment such as nuclear power plant valves. This method verifies whether the reliability indicators of the pump system meet the overall engineering requirements even when equipment experimental data is insufficient, while simultaneously considering the coexistence of equipment unit mechanisms and data.
[0039] This invention discloses a reliability modeling method based on a mechanistic data hybrid model.
[0040] Example 1: According to the appendix Figure 1-2 As shown, it includes the following steps:
[0041] Step 1: Data collection. Based on the equipment status and reliability assessment requirements, collect multi-dimensional data, including technical status, usage stage, environmental information, and fault records.
[0042] Step 2: Modeling path selection. The modeling path is selected based on the sufficiency of experimental data. When there is no experimental data, a mechanistic model is used. When there is sufficient experimental data, a data-driven model is used. When the experimental data is insufficient, the prior data of the mechanistic model is combined with the experimental data for correction.
[0043] Step 3: Parameter calculation. Calculate equipment reliability parameters by fitting the distribution type and using correction factors.
[0044] Step 4: Output the results, including the reliability model and index evaluation results.
[0045] Furthermore, the data collection and classification includes: dividing fault data into associated faults and non-associated faults.
[0046] Furthermore, associated failures include design defects, component defects, and wear and tear failures; when counting failure counts, dependent failures, planned disassembly, and minor defects are excluded; data is categorized into unrepairable unit data and repairable unit data based on equipment repairability. The data to be collected includes: the technical and production quality status of the equipment or component; the stage the equipment or component is in; testing or usage conditions; design information; environmental information; equipment modification or replacement information; equipment or component usage information: including usage time, failure time, number of failures, corrective actions, etc.; all testing or usage data during the design and usage phases; and relevant information from some similar products as needed.
[0047] Furthermore, the modeling path selection includes the calculation process of the mechanistic model without experimental data and the statistical model with experimental data. The calculation process of the mechanistic model without experimental data includes:
[0048] The decomposition equipment is divided into independent components, and the failure rate of each component is calculated; using the formula λ P =λ SE +λ SH +λ BE +λ CA +λ FD The failure rate of the combined components is calculated, and a correction factor C is introduced. TLF C PS C C Adjust the overall failure rate;
[0049] For high-reliability components, empirical estimation is used, and the failure rate of related components is corrected by combining their design parameters. Before conducting data evaluation, it is necessary to first determine the data type of the reliability data. Only after determining the data type can data evaluation be carried out. According to the actual situation in the project, the data type can be divided into data types for unrepairable units and data types for repairable units, based on whether the unit is repairable; and reliability distribution type.
[0050] The statistical models with experimental data include:
[0051] In the exponential distribution model, through Calculate the failure rate and combine it with confidence intervals. Assess the risks;
[0052] In the Weiber distribution model, the shape parameter β and scale parameter α are calculated using the average rank method, and the distribution parameters are determined by linear fitting of ln[-lnR(t)] and lnt. After determining the data type, data fitting methods can be used to fit the reliability data to determine the distribution type that the reliability data follows. Different failure types follow different distribution types. In power plant data processing, it is necessary to analyze the characteristics of the equipment itself and operating experience to select a reasonable failure distribution type. Commonly used reliability distribution types include the exponential distribution and the Weiber distribution, which are widely used in power plant equipment reliability assessment.
[0053] Specifically disclosed, the modeling path selection also includes a data fusion method, which includes: when experimental data is insufficient, using the output of the mechanism model as the prior distribution, updating the posterior distribution by combining the experimental data with a Bayesian method; and converting the general data of the mechanism model into the prior distribution type of the reliability parameters through a general data processing module.
[0054] After data collection, the equipment's current state and the number of data samples are analyzed. If the equipment has completed reliability testing, a data-driven model is used for calculations. If testing has not been completed, a mechanistic model is used. Mechanistic model calculations without test data include: decomposing the equipment into independent components and calculating the failure rate of each component; using the formula λ... P =λ SE +λ SH +λ BE +λ CA +λ FD The failure rate of the combined components is calculated, and a correction factor C is introduced. TLF C PS C C Adjust the overall failure rate; for high-reliability components, use empirical estimation and combine it with their design parameters to correct the failure rate of related components.
[0055] Specifically disclosed, the calculation of the correction factor includes:
[0056] Load correction factor Where k1 is the material property coefficient;
[0057] Speed correction factor Where k2 is the bearing wear coefficient.
[0058] It should be particularly emphasized that the output reliability model and index evaluation results include:
[0059] Input design parameters: rotation speed 2000 rpm, seal material is fluororubber, bearing type is deep groove ball bearing;
[0060] The failure rate of each component is calculated by using a mechanistic model, and the error rate is obtained by combining the correction factor with the total failure rate.
[0061] For equipment reliability model calculations without experimental data, when there is no experimental data, a mechanism-based equipment reliability prediction model is selected. Taking a pump as an example, when conducting the analysis, it is necessary to summarize various parameters of the pump in conjunction with the design documents, including information such as type, structure, performance, size, speed, seal, and component type, and input this information into the prediction model.
[0062] The overall failure rate of a pump is the combination of the failure rates of its individual components. The failure rates of centrifugal pumps and positive displacement pumps can be calculated using the following formula.
[0063] λ p =λ SE +λ SH +λ BE +λ CA +(λ FD ·C TLF ·C PS ·C C )
[0064] in,
[0065] λ P = Overall pump failure rate (unit: failures per million operating hours)
[0066] λ SE = Overall failure rate of all pump seals (unit: failures per million operating hours)
[0067] λ SH = Overall failure rate of pump shaft (unit: failures per million operating hours)
[0068] λ BE = Overall failure rate of all pump bearings (unit: failures per million operating hours)
[0069] λ CA = Overall failure rate of pump casing (unit: failures per million operating hours)
[0070] λ FD = Overall failure rate of drive units (unit: failures per million working hours)
[0071] C TLE=Load Correction Factor
[0072] C PS =Speed correction factor
[0073] C C =Pollutant Correction Factor
[0074] The failure rate of each component is calculated using its own failure rate model. Taking bearings as an example, when there are multiple bearings in a pump, the calculation uses an additive method, and the failure rate of a single pump bearing is calculated using the following formula:
[0075] λ SH =106 / N
[0076] Where N: number of failure cycles under the applied stress level.
[0077] For high-reliability components such as the pump housing, estimation methods can be used. The pump housing is a highly reliable component, and its impact on the overall pump reliability is mainly reflected in its influence on other less reliable components. For example, the average service life of a pump housing may typically reach 10 years, while the lifespan of seals or bearings may only be one to two years. However, the type of pump housing has a significant impact on the lifespan of bearings and seals because the design of the pump housing affects the load on the pump bearings, and the load is determined by the fluid flow pattern. In the reliability prediction model, the failure rate of the pump housing can be estimated as 0.001 failures per million rotations.
[0078] For non-component-related correction parameters, the calculation formulas for the load correction factor and speed correction factor are given here. Other formulas are too cumbersome to be detailed here, but are built into the model. When assessing pump failure rates, the appropriate pump parameters and configurations are selected, and the model can be matched to its structural characteristics, failure mechanisms, performance requirements, etc., to calculate its reliability prediction results.
[0079] Similarly, for other equipment such as valves and filters, a similar modeling and calculation process can be used to obtain mechanism-based reliability predictions.
[0080] Statistical models with experimental data include: in the exponential distribution model, through... Calculate the failure rate and combine it with confidence intervals. Risk assessment; In the Weiber distribution model, the shape parameter β and scale parameter α are calculated using the average rank method, and the distribution parameters are determined by linear fitting ln[-lnR(t)] and lnt.
[0081] Equipment reliability model calculation with experimental data
[0082] For different types of experimental data, calculations are performed using reliable calculation algorithms.
[0083] When experimental data is available, the classic evaluation method for unit equipment with an exponential distribution is taken as an example. To increase the amount of data for evaluation and reduce the risk of evaluation, reliability evaluation can be conducted by selecting reliability test data, second-party reliability test data, and test data of similar equipment.
[0084] (1) Exponential distribution model
[0085] Input data: test time T, number of failures r, task time t0, confidence level c;
[0086] The formula for calculating the test time is as follows:
[0087]
[0088] Point estimation: Based on maximum likelihood estimation, the point estimates of the exponentially distributed failure rate λ, mean lifetime θ, and reliability R(t0) are as follows:
[0089]
[0090] Interval estimation: Timed truncation case
[0091] The one-sided confidence lower limit of θ is
[0092]
[0093] The one-sided confidence lower bound of R(t0) is:
[0094]
[0095] in —2r+2 degrees of freedom The c-quantile of the distribution.
[0096] (2) Estimation of the Weiber distribution model
[0097] Input data:
[0098] Timed truncation test: number of test samples n, truncation time, failure time t1, t2...tr, r is the number of failures that occurred during the test;
[0099] Fixed-number truncation test: number of test samples n, number of failures r, failure time t1, t2...tr
[0100] Task time: t0, confidence level c
[0101] Point estimation: Shape and scale parameters are calculated using the average rank method. The following formula is used for calculations using the average rank method:
[0102]
[0103] R*(t k )=1-F*(t k )
[0104] In the formula:
[0105] A k —The average rank of the faulty samples
[0106] k — the sequence number of the faulty sample
[0107] A k-1 —The average rank of the previous faulty sample
[0108] ΔA k —Average rank increment
[0109] i — the sequential number of all samples, arranged according to their exit time.
[0110] t k —Running time before failure of the i-th sample
[0111] The two-parameter Weibull distribution has the following failure rate function, failure distribution function, and reliability function:
[0112]
[0113] because
[0114]
[0115] If we take ln[-lnR(t)] as the vertical axis and lnt as the horizontal axis, then the above equation can be expressed as a straight line with a slope of β and an intercept of (-βlnα). This allows us to use a graphical method to find α and β.
[0116] Mean life θ point estimation
[0117]
[0118] Point estimation of reliability R(t0)
[0119]
[0120] Interval estimation of reliability:
[0121] For task time t0, the one-sided lower bound of the reliability R given confidence level c is calculated using the following formula:
[0122]
[0123] in
[0124]
[0125] x = N (1-c) The (1-c) quantile of the standard normal distribution.
[0126] A4 = 0.049q - 0.314 + 0.622q -1
[0127] A5 = 0.2445(1.78-q)(225+q)
[0128] A6 = 0.029 - 1.083ln(1.325q)
[0129] q = r / n
[0130] In the actual modeling process, according to the description in the modeling methodology report, it is possible to calculate various distributions, including the Weiber distribution and the log-normal distribution. This part is traditional content in reliability engineering and will not be elaborated here.
[0131] As an example, when applying this method to the reliability assessment of a centrifugal pump, the following steps are taken: Input design parameters: speed 2000 rpm, seal material fluororubber, bearing type deep groove ball bearing; calculate the failure rate of each component using a mechanistic model, and derive the error rate by combining a correction factor with the total failure rate. Specifically, the failure rate of each component is calculated using the mechanistic model: seal λ SE =15×10 -6 , bearing λ BE =25×10 -6 ,shell λ CA =0.001×10 -6 ;
[0132] Combined with correction factor C TLF =1.2, C PS =1.1, total failure rate λ P =42.1×10 -6 The error between the measured data and the actual data is less than 5%.
[0133] As an example, centrifugal pump reliability modeling
[0134] Data input: design parameters, environmental parameters;
[0135] Mechanism model calculation:
[0136] Seal failure rate λ SE =15×10 -6 bearing λ BE =25×10 -6 shell λ CA =0.001×10 -6 ;
[0137] Correction factor calculation: C TLF =1.2, C PS =1.1, C C =1.05;
[0138] Total loss rate λ P = (15 + 25 + 0.001) × 1.2 × 1.1 × 1.05 = 42.1 × 10 -6 .
[0139] Comparative test data: Measured failure rate 40.5 × 10 -6 With an error of only 3.9%, it outperforms traditional models.
[0140] Meanwhile, this application uses a general data processing module to generate the prior distribution of the corresponding reliability parameters by calculating the equipment reliability model with and without experimental data, since the reliability model calculations differ with those with and without experimental data.
[0141] Example 2: According to the appendix Figure 1-2 As shown, it includes the following steps:
[0142] Step 1: Data collection. Based on the equipment status and reliability assessment requirements, collect multi-dimensional data, including technical status, usage stage, environmental information, and fault records.
[0143] Step 2: Modeling path selection. The modeling path is selected based on the sufficiency of experimental data. When there is no experimental data, a mechanistic model is used. When there is sufficient experimental data, a data-driven model is used. When the experimental data is insufficient, the prior data of the mechanistic model is combined with the experimental data for correction.
[0144] Step 3: Parameter calculation. Calculate equipment reliability parameters by fitting the distribution type and using correction factors.
[0145] Step 4: Output the results, including the reliability model and index evaluation results.
[0146] It is particularly important to emphasize that the method differs in the calculation of equipment reliability models with and without experimental data. The general data processing module utilizes this general data to generate the prior distribution of the corresponding reliability parameters.
[0147] General Valve Bayes Correction
[0148] Prior data: Mechanistic model prediction failure rate 8×10 -6 ;
[0149] Test data: Two failures occurred in 10 sets of data, with a total test time of T = 5000 hours;
[0150] Posterior calculation:
[0151] pass Hours, reliability R L (1000)=e -1000 / 1428.6 =0.51; failure rate after Bayesian correction: 7.2 × 10⁻⁶ -6 The confidence interval width was reduced from [5.5; 10.1] to [6.8; 7.6].
[0152] The calculation of the correction factor includes:
[0153] Load correction factor Where k1 is the material property coefficient;
[0154] Speed correction factor Where k2 is the bearing wear coefficient.
[0155] Compared with existing technologies
[0156]
[0157] Determining equipment reliability parameters based on mechanistic data mixing
[0158] According to the technical approach designed in this application, when there is no experimental data, the reliability result of the pump equipment is obtained through the equipment reliability prediction model based on the mechanism model; when there is sufficient experimental data, the reliability parameter result of the pump equipment is obtained through the data-driven pump equipment reliability assessment model; when experimental data exists but is insufficient, and similar equipment is insufficient to support the data-driven pump equipment reliability model, the reliability analysis result of the pump equipment is obtained by using the pump equipment reliability prediction based on the mechanism model as prior data and correcting it with experimental data.
[0159] The reliability parameters obtained from the mechanism-based pump equipment reliability prediction model are general data, which do not directly provide the prior distribution type and parameters, but rather provide the data. In this case, the general data processing module needs to use this general data to generate the prior distribution of the corresponding reliability parameters.
[0160] The foregoing has shown and described 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 to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.
Claims
1. A reliability modeling method based on a mechanism data hybrid model, characterized in that, The method comprises the following steps: Step 1, data collection, according to the equipment state and reliability evaluation requirements, collect multi-dimensional data, including technical state, use stage, environmental information and failure record; Step 2, modeling path selection, select the modeling path based on the sufficiency of experimental data, use mechanism model when there is no experimental data, use data-driven model when experimental data is sufficient, and fuse prior data of mechanism model and experimental data for correction when experimental data is insufficient; Step 3, parameter calculation, calculate the reliability parameters of the equipment through distribution type fitting and correction factor; Step 4, result output, output the reliability model and index evaluation result.
2. The reliability modeling method based on mechanism data hybrid model according to claim 1, characterized in that, The data collection and classification comprises: dividing the failure data into associated failure and non-associated failure.
3. The reliability modeling method based on a mechanism data hybrid model according to claim 2, characterized in that, The associated failure comprises design defect, component defect and wear-out failure; when counting the number of failures, exclude dependent failure, planned disassembly and slight defect.
4. The reliability modeling method based on mechanism data hybrid model according to claim 1, characterized in that, The modeling path selection comprises the calculation process of mechanism model under no experimental data and statistical model under experimental data, and the calculation process of mechanism model under no experimental data comprises: The decomposition equipment is divided into independent components, and the failure rate of each component is calculated; using the formula λ P =λ SE +λ SH +λ BE +λ CA +λ FD The failure rate of the combined components is calculated, and a correction factor C is introduced. TLF C PS C C Adjust the overall failure rate; For high-reliability components, use empirical estimation, and modify the failure rate of associated components in combination with their design parameters; The statistical model under experimental data comprises: In the exponential distribution model, the failure rate is calculated by combining the confidence interval to assess the risk. In the Weibull distribution model, the shape parameter β and the scale parameter α are calculated by the average rank method, and the distribution parameters are determined by linear fitting ln[-lnR(t)] and lnt.
5. The reliability modeling method based on mechanism data hybrid model according to claim 1, characterized in that, The modeling path selection also comprises a data fusion method, which comprises: When the experimental data is insufficient, use the mechanism model output as the prior distribution, and update the posterior distribution by combining the experimental data through the Bayesian method; Convert the general data of the mechanism model into the prior distribution type of the reliability parameter through the general data processing module.
6. The reliability modeling method based on mechanism data hybrid model according to claim 1, characterized in that, The calculation of the correction factor comprises: Load correction factor where k1 is a material characteristic coefficient; Speed correction factor where k2 is a bearing wear coefficient.
7. The reliability modeling method based on mechanism data hybrid model according to claim 1, characterized in that, The output of the reliability model and the index evaluation result comprises: Input the design parameters, the rotating speed is 2000 rpm, the sealing material is fluorine rubber, and the bearing type is deep groove ball bearing; Calculate the failure rate of each component through the mechanism model, and obtain the error rate by combining the correction factor and the total failure rate.
8. The reliability modeling method based on mechanism data hybrid model according to claim 1, characterized in that, The modeling method generates the prior distribution of the corresponding parameters through the general data processing module using these general data.