Research On Methods For Reliability Modeling And Assessment Of Heavy-Duty CNC Machine Tools

Heavy-duty CNC machine tools are now indispensable manufacturing equipment in many manufacturing sectors, such as aerospace, power equipment, shipbuilding, rail transportation and ocean engineering. The technical level, performance index and reliability status of heavy-duty CNC machine tools are getting more and more critical, especially for the continual prosperity of manufacturing industry, and the national security of our country. Under the guidance and support of national plans and projects, including the “Planning on Adjusting and Revitalizing the Manufacturing Industry” and the “Long-term Science and Technology Development Plan(2006-2020)”, great progresses have been made on the research and development of heavy-duty CNC machine tools. Various new types of heavy-duty CNC machine tools have been introduced, which successfully fill the gap of lacking heavy-duty CNC machine tools in many key industries. However, behind the breakthrough of heavy-duty CNC machine tools, the reliability of these machine tools is under great challenges. Various reliability issues raises among these newly introduced heavy-duty CNC machine tools, telling the very truth that we are quantitatively large but not competitively strong. Reliability engineering of heavy-duty CNC machine tools is consequently becoming an imperative research topic.Reliability modeling and assessment are fundamental techniques within reliability engineering. Accurate reliability modeling is the prerequisite for reliability design, reliability test and reliability improvement. Precise reliability assessment can help the quantitative control and health management of machine tools. Due to various characteristics of heavy-duty CNC machine tools, which make them different from traditional manufacturing machine tools, various reliability modelling and assessment methods can hardly be used for heavy-duty CNC machine tools. Accordingly, it is critical to investigate the technique for reliability modeling and assessment of heavy-duty CNC machine tools.Motivated by a practical project supported by the “High-end CNC Machine Tools and Basic Manufacturing Equipment” major science and technology project, this paper devotes to the research on methods for reliability modeling and assessment of heavy-duty CNC machine tools. Major research contributions and innovative outcomes are summarized as follows.(1) Development of a method for reliability modeling and assessment based on failure time data by taking account the effect of maintenance, failure-relevance and model uncertainty. Reliability of heavy-duty CNC machine tools is affected by various factors from the life cycle stages of design, manufacturing, usage and maintenance. The assumption of independently identically distributed failure time data is invalid in the modeling process. To solve this critical issue, a thorough investigation is carried out on the failure time data modeling under different maintenance situations, generating a serial of failure time models. A model for failure-relevance quantification is then proposed and incorporated into these models to construct the models for failure time data considering the effect of maintenance and failure-relevance. A reliability assessment method based on the Bayesian model averaging method is proposed to handle the problem of model uncertainty within the failure time data analysis.(2) Development of a system reliability modeling and assessment method with multilevel heterogeneous data sets by leveraging Bayesian method and Bayesian network. The analysis of multilevel heterogeneous data sets is generally challenged by the modeling of data with different data types, handling overlapping data, as well as integrating objective and subjective information. A comprehensive Bayesian framework for the integration of multilevel heterogeneous data sets is presented. The pass-fail data, lifetime data, and degradation data at different system levels are combined coherently for system reliability analysis. An information fusion framework is constructed based on the Bayesian method and the Bayesian network. Within this framework, the technique for Bayesian integration of multiple independent heterogeneous data sets, the method for dealing with overlapping data set with Bayesian network, and the method for fusing objective and subjective information using Bayesian network are proposed. A comprehensive method for multilevel heterogeneous data modeling and reliability assessment is constructed for heavy-duty CNC machine tools.(3) Development of stochastic process based models for degradation processes with constant, monotonic and S-shaped degradation rates. Other than a constant degradation rate, various patterns of time-varying degradation rates of degradation processes are often encountered in degradation modeling of heavy-duty CNC machine tools. By introducing a model of degradation rate, and incorporating it together with the random effect model into inverse Gaussian process models, the problem of degradation modeling with time-varying degradation rate and unit-specific degradation process is resolved. Meanwhile, to facilitate the integration of degradation observations from both manufacturers and users of heavy-duty CNC machine tools, a general Bayesian integration framework is constructed. Degradation inference, reliability assessment, and residual life prediction are introduced within the Bayesian integration framework.(4) Development of a method for degradation modeling under dynamic environmental and operating conditions. Degradation processes of heavy-duty CNC machine tools are generally characterized as multiple degradation processes with dynamic operating conditions. To model this kind of degradation processes, inverse Gaussian process model and Wiener process model are combined with Copula function to construct a general multiple degradation process model under dynamic environmental and operating conditions. Two types of dynamic covariates, including environmental conditions and operating profiles, are treated separately and incorporated into the proposed multiple degradation process model. To facilitate information integration and reliability analysis, Bayesian method is used to implement parameter estimation and degradation analysis. Reliability assessment for products’ population, degradation prediction for missing observation points and for future observation points, as well as residual life prediction for individual products are investigated. Based on these methods, a solid foundation on degradation modeling and reliability analysis is constructed for further condition monitoring and health management of heavy-duty CNC machine tools.