Predicting uncompensable heat stress with embedded, wearable sensors

The use of heavy protective clothing (such as by EOD operatives) brings problems related to the build-up of heat within the clothing, potentially endangering the health of the wearer and their activities. This paper presents a method of autonomously predicting the onset of thermally dangerous conditions such as Uncompensable Heat Stress in EOD operatives. The method is based on a Dynamic Bayesian Network, trained using Gaussian Kernel Density Estimators based on experimental data. An accuracy of 88.5% was achieved on unseen data when predicting the occurrence of heat stress up to two minutes in the future. The method is intended to be generally applicable to wearers of protective clothing in thermally challenging environments.