A control chart based approach to monitoring supply network dynamics using Kalman filtering

In today's global market, a critical issue for companies to survive the increasing competition is how to handle uncertainty in their supply network. In this paper, we explore the application of Kalman filtering to estimate the dynamic states in a supply network. Two state-space models are developed. The first one focuses on processing each individual order which includes both waiting time and value-added processing time. Considering the correlation of consecutive orders, the second one enhances the state-space model by employing autoregressive model of waiting time. To signal potential abnormal events, the estimates from the models are further used in control charts with control limits being updated at each monitoring stage according to the related estimation error. A supply network case example is studied and we conclude in the benchmark model (without using Kalman filtering) and the first state-space model, the data collected from the bottleneck stage turns out to be most valuable for increased accuracy in detecting tardy orders. The second state-space model consistently outperforms both the benchmark model and the first state-space model for robustly early detection of abnormalities at all stages, especially the stages before the bottleneck stage, of the system.