In Order To Improve Equipment Efficiency As The Goal Of Preventive Maintenance Optimization,

Maintenance techniques are being investigated extensively as a method of prolonging the lifecycle of equipment, cutting down cost and enhancing production capacity. On one side, lose caused by unreasonable maintenance far exceeds the maintenance cost. One the other side, the sophisticated new equipments entitled as cluster tools in semiconductor industry bring the complexity to maintenance planning. However, these challenges call for the necessity of maintenance plan study for such cluster tools.This thesis is devoted to optimize the maintenance plan of bottleneck tools in semiconductor factory. Followed by a break down analysis of overall equipment efficiency and its mapping relationship with equipment loses, an improved application method of the delay-time-based maintenance model is proposed to optimize the overall-equipment-efficiency-oriented maintenance scheduling, which mainly addresses the following problems:1. With the goal of improving the overall equipment efficiency, a maintenance strategy as minimizing downtime is identified based on a comprehensive break down analysis of the overall equipment efficiency, and its relationship with equipment loses.2. An improved application method of the delay-time-based maintenance model is proposed by appling model sensitivity analysis after the estimation of model parameters, with which the active maintenance concept is novelly combined with the delay-time to contribute to the optimal solution. The model sensitivity analysis enables the estimation of model stability and identification of top limiters, which support the best optimization strategy.3. To actively decrease the most sensitive parameter of the maintenance model, a data-driven approach for failure prediction is investigated to support predictive maintenance that aims to reduce equipment unscheduled downtime. The Rescaled Range analysis based Hurst exponent is employed for prediction modeling based on machine failure records.4. To make an effort to the control of other sensitive parameters, the defect rate and maintenance activity duration are re-estimated based on objective statistics and experiential data. Then, the best preventive inspection interval is identified by optimizing the delay-time-based maintenance model. The comparison analysis of different solutions given by different application approaches shows the improved method proposed in this thesis can reduce unscheduled downtime and enhance equipment efficiency more significantly.