Statistical Detection Methods In The Fire Early Warning

Fire not only deprives human's lives and estates, but also destroys environment and nature resources which support all persons in the word. Therefore, it's great important to research the law of fire in achieving accurate and reliable early detection and alarming.The major task of fire detecting technology today is advance warning and reducing mis-reporting. However advance warning and reducing mis-reporting is just a contradiction, which brings forward higher demand for fire detecting technology,.Gases are suitable for early fire detection because they're produced at the same time with the fires and they spread quickly. Based on a great deal of research, we get the conclusion that CO is the best target gas to detect early fires. Researchers use a specific experiment system for early fire warning research based on FTIR technology to make a lot of experiments to simulate fires of limited space. The spectrums of CO are collected dynamically and the concentration is got by spectral analysis. With the utilization of window function and least square method, we distill the process-characteristic parameters of CO with the rejection of random element. The concentration which contains the process-characteristic information of the fire can be achieved.On the base of the signal detection, the problem of fire-warning can be analyzed. Being aimed at the character of non-stationary random process, a detecting model of fire detecting system can be established. To judge whether the fire is on or not, the linking of NP criteria and improved sequent detecting can be used at it.The NP criteria is suitable for fire warning for it can be used without ensuring the priori probability or the cost function. The improved sequent detecting not only avoid the shortcoming of misreporting if it's judged by one-time observing, but also its fixed observing dimension can improved the property of detection. With the renewed data in the form of moving window, the aim of continuously fire warning can be achieved.It's proved the new method can provide an accurate early alarm without misreporting. Comparing to conventional fire detector, this new method will provide more time for fire preventing.