Long short-term memory neural network for statistical process control of autocorrelated multiple stream process with an application to HVAC systems in passenger rail vehicles

Rail transport demand in Europe has increased over the last few years, and passen ger thermal comfort has been playing a key role in the fierce competition among different transportation companies. Furthermore, European standards settle opera tional requirements of passenger rail coaches in terms of air quality and comfort level. To meet these standards and the increasing passenger thermal comfort de mand, data from on-board heating, ventilation and air conditioning (HVAC) systems have been collected by railway companies to improve maintenance programs in the industry 4.0 scenario. Usually, a train consists of several coaches equipped with a dedicated HVAC system, and the sensor signals coming from each HVAC sys tem produce multiple data streams. This setting can thus be regarded as a multiple stream process (MSP). Unfortunately, the massive amounts of data collected at high rates makes each stream more likely to be autocorrelated. This scenario calls for a new methodology capable of overcoming the simplifying assumptions on which traditional MSP models are based. This work is intended to propose a new control charting procedure based on a long short-term memory neural network trained to solve the binary classification problem of detecting whether the MSP is in control or out of control, i.e., to recognize mean shifts in autocorrelated MSPs. A simulation study is performed to assess the performance of the proposed approach and its prac tical applicability is illustrated by an application to the monitoring of HVAC system data, made available by the rail transport company Hitachi Rail based in Italy.