Improved grey principal component analysis neural network based adaptive thermal comfort model: Application in the enclosed cabin with microclimatic conditions

Predicting the thermal comfort of operators in enclosed cabins under extreme operational conditions is crucial for the enhanced and optimal design of cabin air circulation systems. In this study, an improved supervised machine learning algorithm, namely a Grey Principal Component Analysis (G-PCA) was proposed to evaluate the operators’ thermal comfort. The comprehensive dataset was first attained and constructed from the proposed 32 indicators, which recorded each tested object’s EEG and physiological features, core body temperature, skin temperature, and subjective assessments. The accuracy of the proposed model was demonstrated by comparing it with the results predicted via existing indicator sets (PMV and 7-point skin temperature) and modelling algorithms (BP and PCA BP). The results showed that the prediction accuracy of the PMV model, 7-point skin temperature, the 32 indicators were found to be 79.4%, 82.8%, 96.1%, utilising the G-PCA WNN algorithm, and the prediction accuracy of BP, PCA BP, G-PCA WNN is 82.3%, 89.9%, 96.1%, based on the 32 indicators. A conclusion could be drawn that the proposed 32 indicators can incorporate a more comprehensive range of thermal comfort information, and the improved G-PCA WNN thermal comfort prediction model achieved the best prediction performance among the compared algorithms and models.

openalex: W4391755697