Research On Performance Prediction And Control Of Production Line Based On Variability

With the continuous development of automated and flexible production technology,the performance design and control method of production line has become more flexible.However,the unavoidable problems of machine random failure time,random repair time,non-conforming product rework or departure midway,and the coupling process rate are also more serious.These problems bring a lot of unpredictable and dynamic uncertainties to the production line,which makes the performance of production line fluctuate frequently,and increases the complexity of performance analysis and control.This requires the production line to have high utilization rate,short cycle time,and low cost of production line in variability.The corresponding performance prediction and control models require to be implemented in the variability,but the traditional methods are difficult to meet these comprehensive requirements.Therefore,this dissertation takes a series-parallel production line as the research object.The key technologies in-cluding production line performance prediction,sensitivity analysis,and control strate-gy are studied from the variable perspective,and some innovative and applied research results have been obtained.The main research work and innovations in this dissertation are as follows:(1)Aiming at the problem of coupling of processing rates between multi-stations and exponential growth of state space caused by the Markov model,an approximate it-erative method based on M/M/1/m(Kendall notations:The first‘M'indicates that the reaching process is a Poisson process or a negative exponential Markov.The second‘M'means that the processing process is a Poisson process or a negative exponential Mar-kov.‘1'indicates a single workstation or machine.‘m'indicates the maximum number of products in the single queueing system)queueing model for general process rate of workstation is proposed.Firstly,the processing rate between independent single work-station and multi-workstation are analyzed according to the processing rate of Markov model on an equivalent workstation.Then,the processing characteristics,such as star-vation,blocking and non-conforming products,are studied.The equivalent single work-station analysis method is applied to estimate the performance parameters such as blocking/starvation probability of the workstations.Finally,the general processing rate of series-parallel production line is derived based on the approximate iterative method,and the detailed parameter design steps and criteria are given.The proposed approxi-mate iterative method can effectively avoid the impact of variability on the disturbance of the prediction results,and improve the shortcomings of the traditional Markov per-formance prediction model in dealing with more than 3 machines in series with too large state space and complex modeling analysis.It has a widely future application.(2)In order to solve the problem that the random input parameters of model are difficult to design and tradition mathematical queueing models are not suitable for per-formance modeling due to the random characteristics of the input variability parameter type in the analysis model of the production line,a performance sensitivity analysis method based on Arena is proposed.The detailed design steps in the Arena simulation are given,and the relationship between the variability parameters and the influence of various parameters on the sensitivity of production performance in production process is studied.The general definition of coefficient of variation(CV)under different random probability distributions is given.Therefore,the CVs of Gaussian,Triangular,and Gamma probability distribution are calculated according to the variability theory,re-spectively.Under the condition of not changing the mean value of the random variabil-ity,which is arbitrary probability distribution,the influence of the CVs of the produc-tion line c_l,the CV of the time interval of the product c_a,production batch Ba and the buffer size Bu on the production line's throughput TH and the quantity of the product processed N_f are analyzed.By comparing the internal relationship between the variabil-ity parameters and the overall performance of the production line,the problem of un-clear primary and secondary design of object simultaneously relating to multi-parameter is avoided in the performance control of production planning.(3)Aiming at the problem of optimal control strategy for series-parallel produc-tion line performance,an optimal control model of M/M/n/m-FCFS queuing network performance is established,which takes the number of parallel machines and buffer size as control parameters.A hybrid meta heuristic algorithm(HMHA)is designed to find the optimal performance and corresponding production conditions of the production line,and the detailed solution idea and steps of the HMHA are given.The control strategy of buffer size and parallel machines in workstations are experimentally studied.The cor-rectness and feasibility of the theoretical analysis and model design are verified by the Arena simulation,which provides an important reference for improving the production performance.This dissertation analyzes and solves the problems of the series-parallel production line above.These problems include the performance prediction of coupling characteris-tics of the processing rate among the workstations and the exponential growth of state space caused by Markov model,arbitrary probability distribution of variability parame-ters for performance sensitivity analysis,and the optimal control strategy of parallel machine quantity in workstations and the corresponding buffer size.The contribution of the dissertation can be a helpful supplement to performance prediction and control the-ory research of the series-parallel production lines,and it can also enlighten managers or operators in actual production.