Design of a Bisquare-based fire monitoring and warning system for training rooms

In response to the issue of over-range abrupt change during long-term fire monitoring in the training rooms, a temperature fitting prediction method based on the Bisquare clipping and filtering algorithm was developed, along with a fire monitoring and warning system for training room. The Bisquare algorithm is designed to detect abrupt changes in the current sampling temperature in real time, and utilize the rolling window historical data to predict an optimal replacement value for the abnormal temperature value. The accuracy of the Bisquare clipping and filtering algorithm was verified through the fire standard temperature rise curve, the fire model curve, and long-term fire monitoring experiments in the training room. The fitting error during normal fire monitoring in the training room was constrained to less than 0.05℃, and the maximum fitting error during the fire monitoring remained below 0.6℃.Compared with the traditional least squares and Kalman filtering methods, the average improvement in fitting prediction accuracy was 12.1%, which effectively solves the problem of abrupt change data that may occur in the long-term fire monitoring process, and improves the reliability of fire monitoring and warning in the training room.