| Manufacturing industry is the foundation of the overall national strength.Under the background of raising the country's manufacturing power,the importance of the reliability of CNC equipment has been realized.With the implement of National Science and Technology Major Project of‘High-grade CNC Machine Tool and Basic Manufacturing Equipment',the reliability of CNC equipment in China has been gradually improved,and the demonstration application of domestic equipment on the production line has also begun.In order to meet the requirements of multi-variety,variable-volume,order-based production,modern production lines have increasingly grown in complexity,flexibility,efficiency,and digitization correspondingly.The types and quantities of their components have also rapidly increased.Coupled with the characteristics of the series structure and complex working environment,equipments breakdown frequently,and the entire line performance is extremely sensitive to these failures.In addition,as an important part of production capacity,the availability of production lines is the basis for achieving predictable production.Therefore,no matter for a new design production line or the existing one,managers have an urgent need to predict and master their availability levels.In view of the reasons above,it is necessary to study the availability of production lines.As a typical discrete event dynamic system,the production line has complex and varied behavior,and its availability is reflected in workstations',buffers',and overall line's performance.The existing researches on the availability of manufacturing systems mainly focus on the single-indicator modeling methods,there is still no recognized indicator system,and the existing methods are mainly aimed at specific systems with strong theoretical additions,which often lead to poor implementation in engineering practice.Meanwhile,a single system availability indicator cannot describe the bottleneck of the production line,so it is difficult to guide the implementation of availability guarantee and growth.Due to the lack of a comprehensive and unified availability evaluation method for discrete production line with buffers,this paper combines the tasks of National Science and Technology Major Project and industry needs,takes the unreliable series-parallel asynchronous production line with buffers as the research object,focuses on the relationship between configuration parameters and availability evaluation indicators of production line.In this paper,the analysis of operating status for the production line is taken as the starting point,and an optimized availability evaluation indicator system is established.Three simulation-based methods for obtaining these indicators are proposed,and surrogate model technology is introduced to develop an indicator predicting method.Finally,sensitivity analysis is used to identify the key parameters.The main content of the paper is as follows:(1)Aiming at this issue that there is still no recognized indicator system for production line's availability evaluation,the operation mechanism of production line with buffers is analyzed,the dynamic relationships between various components of the production line(workstations and buffers)are described,and the status analysis of the workstations and the entire line is conducted.According to the hierarchical structure of the production line,the availability evaluation indicator system is divided into three units:single equipment,buffer,and entire line.17 types of availability evaluation indicators are defined,and an indicator system is initially obtained.(2)Due to the large amount of redundant information of the preliminary indicator system,an optimization method combining R cluster analysis with multi-strategy screening model is proposed,and a practical and feasible construction process is given.Among which,R cluster analysis can classify indicators by relevance,and multi-strategy screening model can quantify indicators for each cluster based on subjective,relevant,and informative factors,and finally come to an optimized indicator system.An application in a rough machining production line for crankshaft is given,from which the number of indicators were reduced from 49 to 13 through the clustering and screening,and they can reflect the availability of the production line from different sides.(3)Three simulation methods of production line are proposed for the calculation of availability evaluation indicators.The first method is to adopt Plant Simulation software for production line simulation,and take it as a control group to verify other methods.The second one is discrete event real-time simulation developed under the Matlab platform according to the production line operating mechanism.This method can scan the status of the system at each simulation time unit and make a record,with the help of which it can get a deeper understanding of the dynamic production information of the production line.The last method is a semi-analytical simulation which combines analytical model and Monte Carlo technique.An output-inventory model,which could divide the entire simulation time into series intervals of multiple TBFs and TTRs,is tentatively proposed.It can calculate the output and buffer inventory for each unit in every interval,and based on this,the calculation formula of the average number of work-in-process products is given.The method will reduce the computational load from time unit level to failure number level,so as to significantly improve computational efficiency.Taking six different types of production lines as examples,the above three methods and two analytical methods were compared and verified.At last,for the initial transient problem in simulation,Welch moving average approach is adopted to determine the warm-up period.(4)Since simulation methods for experiment in high-dimensional uncertainty space need a large amount of simulation time,the surrogate model is developed as availability evaluation indicator predicting method to improve computational efficiency.Based on experimental design of Latin hypercube sampling,the surrogate model is introduced into the availability analysis of production lines.Kriging,stochastic Kriging and moving least square method are used as surrogate models to establish the mathematical relationship between the production line configuration parameters and availability evaluation indicators.Example results indicate that the proposed procedure can achieve high prediction accuracy,in which R~2 is larger than 0.95.(5)Sensitivity analysis of availability indicator model is conducted.Based on the model samples,this paper studies the influence of the production line configuration parameters on the availability indicators,and analyzes their uncertainty in the presence of input uncertainty.With respect to the interaction effect of configuration parameters,two global sensitivity analysis methods,which are Fourier amplitude sensitivity test(FAST,for qualitative analysis)and Sobol'method(for quantitative analysis),are introduced and the key parameters are identified.Example results of two methods are consistent in general except the difference of main effect and total effect in several factors.In order to solve the problem of large-scale calculation for global sensitivity analysis with multiple parameters,a sensitivity analysis procedure based on surrogate model or semi-analytic simulation is proposed and this procedure can get similar results with low-cost calculation.Through the study of this paper,an evaluation indicator system for production line availability was established,the acquisition and prediction methods of the indicators were studied,the quantitative relationship between the configuration parameters of the production line and the availability indicators were obtained,and the sensitivity analysis of the model was completed,which could provide support for availability comprehensive assessment,availability allocation,and availability enhancement,etc.of production line in the future. |
PDF Full Text Request