Improving efficiency in hazard mapping: combining physics-based and kriging models for estimating noise exposure

Estimation of hazard exposure for workers is a key requirement in improving occupational safety. Personal exposure may be estimated with wearable equipment worn by each worker, or a network of sensors throughout the workplace that create a hazard map. Hazard mapping utilizes data from the sensors to interpolate exposure levels at any given location. An experiment was conducted comparing three different interpolation methods: a traditional Kriging model, a physics-based model, and a hybrid model that combines the two. Noise data were collected in four different settings, and the interpolation models were tested on real levels of noise in over 10,000 model simulations. The hybrid model performed the strongest in terms of accuracy in almost all scenarios. In the few scenarios where another model performed better, the hybrid model performed within 10% accuracy of the better model. The hybrid model performed especially well in scenarios where fewer data points were present, showing that the hazard mapping process can have its cost reduced by implementing the model. The hybrid model may also be used in future work for hazard mapping with other hazards besides noise, such as gases and aerosols.