Research Of Fire Alarm Based On The Continuous Bayesian Network

Fire not only brought serious disasters to the natural environment, but also caused serious threats and great losses to human life and property. However, the rapid development of economy and the increasing of people’s social activities have brought more possibility to the occurrence of fire. Therefore, in the current environment, we need to do all kinds of work to prevent fires. Fire is a disaster of losing control either in space or in time with its velocity of propagation. The fighting process of fire will consume a large manpower and material resources and financial resources. In order to minimize the harm and the loss caused by fires, it is important and practical significance to alarm before the fire occur. Alarm before the fire is a prediction process to uncertain event. Bayesian network, a new branch of artificial intelligence is do well in dealing with uncertain question. It has its unique advantages in dealing with uncertain event, and it is applied to solve many conditional questions which depending on many control factors.In view of the traditional fire warning research, most of them is used to collect information from a single sensor and discrete these information. But the limitation of a single sensor to collect information and the insufficient of discrete information will affect the authenticity and reliability of fire warning indirectly. With the advantage of multiple sensor and Bayesian network, a Bayesian network with Gaussian mixture model is proposed to deal with the problem of fire alarm in this article. It is necessary to establish appropriate model of fire alarm which is used to make decision. This method is useful to deal with the information of fire and overcome the problem of lacking information. In addition, we can complete the data collection and data processing with the multiple-sensor information fusion technology avoiding the insufficient. This model has better scientific significance and practical value.The main contents are as follows:(1) Certain the topology structure of the net by analysis the various physical quantities of fire information and the relationship between the factors.(2) For the data acquisition and data processing, we can use the multiple-sensor information fusion technology, and establish Gaussian Bayesian network model for multiple-source information with its advantage of handing continuous characteristics.(3) To determine the usefulness of this model, we use the MATLAB simulation platform to simulate, and explained the effectiveness and accuracy of the network model for fire alarm with the comparison of traditional model.The innovation of this article:Single source of data and discrete data affect the accuracy of fire alarm. The proposed of Gaussian Bayesian network and multiple-sensor information fusion model in this article is useful to handle continuous fire data. Gathering fire information in many different sensors and merge these continuous information by Gaussian Bayesian network not only improved the limitations of a single source, avoiding the loss of information, but also improved the authenticity and validity of the fire alarm.