Reducing simulation time using design of experiments

Purpose – The purpose of this paper is to demonstrate the effectiveness of Six Sigma as an innovative tool in software design optimization. The problem of reducing simulation time for characterizing a type of DC motor is studied in this paper. The case study illustrates how Six Sigma tools such as the design of experiments (DOE) method can be used to improve a simulation process. Design/methodology/approach – A first‐principle model for the motor is used for simulation in MATLAB ® . Each parameter in the model is assumed to have a known distribution. Using the random number generator in MATLAB ® , Monte Carlo analysis is conducted. To reduce simulation time, several factors in the simulation process are identified. A two‐level full factorial DOE matrix is constructed. The Monte Carlo analysis is carried out for each of the parameter set in the DOE matrix. Based on the simulation results and the DOE analysis, the Simulink model is identified as the main contributor to the computational time of the simulation. Several steps are taken to reduce the computational time related to the Simulink model. The improved model requires only one‐fourth of the original computational time. Findings – The paper illustrates that Six Sigma tools can be applied to algorithm and software‐development process for optimization purpose. Statistical analysis can be conducted in the simulation environment to provide valuable information. Practical implications – As an example, the improved simulation process is used to derive statistical information related to the speed vs torque curve and response time as part of the motor characteristics. The findings suggest that with an optimized simulation model, large amount of statistical analyses can be conducted in the simulation environment to provide practical information. This approach can be effectively used in early stage of product design, e.g. during the feasibility study. Originality/value – In industry, most of the DOE are conducted using real‐test data. It is usually time consuming and cost inefficient. This paper combines mathematical modelling and statistical analysis to optimize a simulation model using DOE. The novel approach used in this paper can be applied in many other software optimization problems. It is expected that this approach will broaden the application of Six Sigma in industry.