Health and Safety Situation Awareness Model and Emergency Management Based on Multi-Sensor Signal Fusion

Abstract

Disasters that are uncertain and destructive pose severe threats to life and property of miners. One of the major precautious measures is to set up real-time monitoring of disaster with a number of different sensors. Single sensor which features weak, unstable, and noisy signal is prone to raise misjudgment leading to non-linearly correlated data coming from different sensors. This paper unfolds with a theoretical introduction to the situation awareness of data from sensors in the Internet of Things, covering theories including the Internet of Things, multi-sensor data fusion, and situation awareness. Subsequently, we construct a framework for the situation awareness system based on multi-sensor fusion in the open-pit mine Internet of Things. The data coming from multiple sensors are pre-processed with wavelet transform, data filling, and normalization. In addition, information entropy theory is introduced to weight the data varying with attributes. An RF-SVM-based model is constructed to accomplish data fusion and determine situation levels as well. The output of the RF-SVM-based model is input as an ELM model. The fusion results at the first 10 time points are used to forecast the situation level at next point, so that the proposed disaster forecast approach in this paper is practiced. To test the stationarity and validity of the approach, MATALAB is employed to run a simulation of the data of a given open-pit mine. The results show that the RMSE of the model remains below 0.2 and TSQ is no greater than 1.691 after we run 50 times, 100 times, and 200 times iteration. It convinces that forecast results made by the model are valid, indicating that the multi-sensor signal fusion which is effective and efficient provides support to disaster situation forecast and emergency management in the mine.