Using multi-sensor data fusion to predict dangerous states of LNG transport tank

The goal of this work was to automatically predict the dangerous states of hazardous chemicals´ transport tank using multi-sensor data fusion. Eight kinds of sensors, which are gas sensor, temperature sensor, humidity sensor, pressure sensor, liquid level sensor, acceleration sensor, angle sensor and switch sensor, were used in the monitor system of a tank for LNG (liquefied natural gas) transportation on road. Data from the tank during transporting LNG in 20 days were analyzed to obtain the statistics of sensors´ signals. Based on the JDL data fusion model, different mans were applied to process data in different fusion levels, such as weighted average, least-squares estimation, Kalman filtering, and neural network. The data fusion system firstly automatically judges the transportation pattern of the tank with the characteristic parameters of the sensors, and then predicts the dangerous states of the tank, including leakage, traffic accident, etc., with special judge criteria under different transportation patterns. The algorithm is trial used for LNG transport tank monitoring, and results show that the predicted results were in accordance with the real states of the tank as far as now.