top of page

Reconfigurable Manufacturing Systems Design using Simulation-based Multi-objective Optimization

For manufacturing companies to remain competitive, a reconfigurable manufacturing system (RMS) able to rapidly and economically cope with uncertainties, like unpredictable customer needs and demand variations, is a priority. The RMS design process is a crucial and underdeveloped aspect, affected by the complexity of determining the optimal manufacturing system design.

Multi-objective optimization

Additionally, although simulation-based optimization techniques have been applied to tackle certain RMS challenges, only a few attempts have applied multi-objective optimization methods to deal with several conflicting design objectives simultaneously. Hence, the paper this whitepaper is based on, aims to propose a simulation-based multi-objective optimization method towards the main RMS design areas.

Simulation-based multi-objective optimization towards RMS design areas

RMS design challenges have been widely studied. However, a simulation-based multi-objective optimization approach to simultaneously deal with the three main areas, finding the optimal system configuration with the optimal process plan and minimum number of machines, is something not found in the literature.

Therefore, when considering the flexibility offered by simulation in terms of modelling and representing the complexity, dynamic and stochastic behavior of RMS, unlike prior research that used optimization methods to combine several aspects of the RMS, the paper is, to the best of our knowledge, the first to introduce a simulation-based multi-objective optimization approach towards the introduced RMS design areas. This approach takes into account decision variables that represent the number of machines, the system configuration, and the process planning, for maximum throughput and minimum buffers capacity.

Investigates the design of an RMS for a machining process

The machining process takes 960 seconds for 36 tasks. All machines used have 90% availability with a 5 minutes mean time to repair (MTTR). Assuming budget limitations, only 12 machines can initially be bought but additional machines are planned to be added in the future to cope with volume increases. Furthermore, due to space limitation and the technological constraints of the machining processes, in which some tasks like 3, 17 and 33 need to be performed in three different types of machines.

The initial RMS configuration consists of three stages with two buffers in between the stages. Machines in the same stage perform the same tasks sequence. In each stage, there is space for up to 6 machines and it is assumed that the material handling system can deliver parts to them. It is also assumed that installed machines cannot be removed and need to be used in future configurations. Therefore, the RMS taken into account can vary from 12 to 18 machines distributed in three stages.

The proposed approach uses simulation-based multi-objective optimization to simultaneously identify how to maximize throughput per hour (THP), minimize the lean buffer, with the minimal total number of machines, while providing the optimal process plan for every configuration in which the RMS uses between 12 and 18 machines.

Developed a baseline simulation model

A baseline simulation model has been developed using a DES software called FACTS Analyzer. FACTS Analyzer acts as a DES engine used for the animation and the optimization. The FACTS model serves as the basis for the iterative execution of the combinations of input parameters according to the optimization objectives and constraints while the optimization engine evaluates the outputs to set a new combination of input parameters in the effort to generate the Pareto front.

100 000 iterations

The well-known multi-objective optimization algorithm, NSGA-II has been used to solve this problem with 100 000 iterations. Three conflicting optimization objectives are defined as follows.

Maximize f1=THP(x) : Throughput per hour

Minimize f2=B1+B2 : Total Buffer Capacity - Where: Bi ∈ {5,10,15,…100} i=1,2

Minimize f3=N : Number of Machines - Where: N ∈ {12,13,14,15,16,17,18}

Subject to precedence constraints (36)

A successful approach

The constraints are used for setting the precedence of the tasks, so a task must be only active in one stage. Another objective, maximize the number of machines, has been used during the selection phase in order to improve the diversity of the population.

The approach used has proven to be successful and provides the optimal system configuration with the optimal process plan and minimum number of machines for maximum THP and minimum buffers capacity. Furthermore, this approach addresses efficiently the scalability planning and covers the information to cope with demand changes and scalable productions.

Carlos A. Barrera-Diaz, Tehseen Aslam, and Amos H.C. Ng

54 views0 comments


bottom of page