Application of Fractal in Machine Downtime Control

Unexpected machine failures directly impact machine availability, which leads to phantom capacity, high maintenance cost, and risks to factory profitability and competitiveness. The collateral damage far exceeds the original downtime itself. This paper proposes a fractal based method for machine downtime trend analysis and performance control. AR (Auto Regression) model is applied for data pre-process, rescaled range (R/S) analysis and Lo's modified rescaled range (R/S) analysis are used for Hurst exponent calculation, Hurst exponent, as well as high order Hurst exponents is employed for mathematical model building for machine performance tendency alarm. A numerical case study in a chipset factory demonstrates that this fractal based tendency alarm method can give statistical insight for potential unstable or downtrend hazard of machine performance, support decision making for additional preventive actions, and contribute to machine unscheduled downtime control.