System identification techniques for simulating fire detector signals

This paper describes possibilities of modeling fire detector signals for the design of fire recognition algorithms. For this purpose, test experiments including test fires fulfilling EN54 Standards and disturbing events have been carried out. Relative mass loss of fire fuels and sensor responses have been introduced as input-outputs of the process that is modeled. In a second step, the ability to identify executed tests has been investigated using a guessed weighting function and ARX-modeling techniques. The evaluation of these techniques is based on the capability to reproduce measured data. According to the practice, sensor responses are different even by the same test fires, therefore, the complexity to generate different signals from identified responses has been additionally taken into account in the evaluation procedure.