Fire Signal Detection Based On Fuzzy Clustering

The fire has played a enormously promotive role in the human civilization and progress of the society. As soon as the fire is out of hand, it will jeopardize the lives, properties, and natural resources, and lead to the calamity. People have been regarding the fire as the accidental one which the incident has happened suddenly and isolatedly for a long time, The method to put out a fire is to take action immediately when the fire is happened. The study on fire is only confined to study the fire law in way that is through summarizing and analyzing firsthand fire information. However, loses of the fire and the difficulty to put out a fire increasing day by day impels people to think further, It is to depend on improving the technology, equipment detection to save life and property simply in order, controlling the fire effectively, or further investigating the mechanism and law of the fire, and solving problem on the basis of fire science.The fire is a kind of combustion phenomenon which the law has a deterministic aspect. It can be investigated and known progressively through simulation, and the fire is a kind of calamity phenomenon which the law has one side of randomness at the same time. The fire parameter gathered by the sensor is the unknown in advance or undetermined. It not only changes with fire character, but also people are very difficult to describe accurately with the mathematical language.Fire detection proves to be very difficult signal processing comparing with the other signal models. It requires that signals can adapt to the changes of different environmental situations while dealing with the algorithm.Originally in the article, to the concrete character of the fire, It is introduced the method of fuzzy mathematics and carrying on cluster analysis. Because many things is fuzzy, and the demarcation line between things is often not very clear either , the fuzzy classfying method gear to actual circumstances evenmore.Because initial value of ordinary fuzzy C-means(FCM) clustering algorithm is uncertain and the result of ordinary FCM has many local minimum, in the article many algorithms are combined to improve the fuzzy clustering, such as the algorithm of annealing, PSO algorithm, and the maximum tree method.The new method is carried on with Matlab software to verify the quality of the algorithm. According to result of emulation tests, this article is provided with new effective methods. These methods don't collect sensitive variance to data, remaining stable and getting the optimum result to the overall situation. In experiment, these methods are used to classify the datum at random and fire datum to make satisfied result.Because the price of a lot of type of sensors is reduced constantly, people consider carrying on the same course with more sensors. We can have photoelectricity, ion, three or even more kinds of detection methods compound together, It possesses the function that realize and assesses to all kinds of fire parameters synthetically. When the data rearch the ultimate limit value, the detector could report to the user accurately with parallel ability.The design,installation,and transport and maintenance of the field bus system are more superior and saving the quantity of the hardware and investment. Different alarm devices integrate basing on field bus protocol to form a fire warning system. Different computer centers links together to form a warning system of fire based on field bus protocol. The whole system carries on the omni-directional fire detection and fire alarm.