|
Original research A HYBRID FUZZY C-MEANS–ANFIS MODEL TO PREDICT PRODUCTION LEAD TIME UNDER MACHINE FAILURE, STOCHASTIC BOTTLENECK AND OTHER UNCERTAINTIESPages 459-466 Abstract
Production lead time is difficult to predict when multiple sources of uncertainty occur simultaneously. Machine failures, stochastic bottlenecks, setup-time variability, processing-time variability, and work-in-process make deterministic estimates inadequate for representing actual production conditions. This study develops a hybrid Fuzzy C-Means (FCM)–Adaptive Neuro-Fuzzy Inference System (ANFIS) model to predict production lead time in a multi-stage manufacturing environment. A simulation model generates 5,000 production observations under stochastic operating conditions. FCM identifies groups of similar production states, and cluster-specific ANFIS models are trained to capture nonlinear relationships between operating conditions and lead time. The proposed model is evaluated against linear regression, artificial neural network, random forest, XGBoost, and standalone ANFIS benchmarks. The FCM–ANFIS model achieves an RMSE of 2.84 hours, MAE of 2.11 hours, and MAPE of 4.37%, yielding the lowest overall prediction error among the compared models. Sensitivity analysis identifies MTTR and bottleneck waiting time as the strongest positive drivers of predicted lead time.
Keywords:
Fuzzy C-Means, ANFIS, production lead time, machine failure, stochastic bottleneck, uncertainty.
|
Amir Azizi 
PDF 