Track 2 Session 3

1:00 to 2:00 p.m. Tuesday, June 15, 2010

RAM Dependence Counts in Production Simulation

For the work described in this presentation, the objective was to simulate new production line alternatives for design optimization. The new line would use the same processes with presumably the same RAM statistics as existing workstations. Mechanics know that a short time-between-failures (TBF) often is followed by a short time-to-repair (TTR) and vice-versa (long-long). A positive correlation of TBF and subsequent TTR reduces the asymptotic variance of workstation availability even though asymptotic availability remains MTBF/(MTBF+MTTR). A positive correlation also reduces the time to steady state. Using the bivariate lognormal TBF and TTR captured more dependence than the alternative bivariate distribution models. Production line simulations also show that dependence reduces throughput uncertainty and time to converge to the asymptotic throughput rate. This result can be exploited by increasing dependence. If it has been a long time between failures, then make more repairs. If TBF is short, look for a quick fix; perhaps the previous repair failed again. Consider opportunistic maintenance to combine repairs when workstations are unavailable. Reliability of production lines could be as important as product reliability.

Key Words: Reliability, Repair Time, Dependence, Correlation, Asymptotic, Availability, Throughput, Statistics, Bivariate Lognormal, Simulation, Variance, Convergence, Cholesky Decomposition, Copula

Larry George

Problem Solving Tools

Livermore, California