Robotic Manufacturing System Throughput Prediction Methodology Development Based on Failure Data of an Automotive Assembly

US automotive industry has seen unprecedented worst market condition in recent years. According to Harbour report (2008), even though the car companies were able to improve labor hours per vehicle (HPV) target year after year which is reduced to 20 hours, but their profit margin is not. The average labor cost is reduced from ;Based on through literature review, there is a need for the development of throughput prediction methodology for the automotive robotic manufacturing process. The primary purpose of this dissertation research is to evaluate the past plant failure data, model the series, select and improve appropriate prediction model methodologies to improve body shop throughput based on past plant failure data. The simulation model developed in this dissertation could be applied to individual equipment failure prediction. It is proved through research findings that the throughput could be improved using simulation methodology using plant MTTR and MTBF data.;This dissertation seeks to establish the feasibility of using ARMA and ANN models to improve the body shop robotic manufacturing process throughput. The key findings are that the throughput can be successfully predicted using ARMA and ANN models for the automotive assembly process based on plant failure data. Quantitative comparative research studies have been done for both ARMA and ANN model prediction accuracy using Mean absolute percentage error (MAPE) as measure. Based on the above ARMA-ANN models, this dissertation developed an optimization prediction model for plant throughput based on failure data. The optimization model prediction was verified and validated with actual plant throughput data. Based on this dissertation results, the application of the developed optimization throughput prediction model would yield high economic benefits to the company. The proposed approach could be applied in similar industrial applications.