A Robust Fusion Algorithm of ts-rbf Fuzzy Neural Network for Infrared Flame Recognition

Abstract

The invention provides a TS-RBF fuzzy neural network robust fusion algorithm applied to infrared flame identification, belonging to the technical field of infrared flame identification. The specific process is: collecting the time-domain signal data of different flames and interference sources, and preprocessing them to obtain frequency-domain signal data; extracting characteristic information from the time-domain and frequency-domain signal data of the waveform to obtain the characteristic vector of the sample , form a sample set; divide the sample set into training set, validation set and test set; build a TS-RBF fuzzy neural network; set the initial values ​​of the parameters of the fuzzy neural network, use the training set to train the fuzzy neural network; The trained fuzzy neural network is used for verification and model selection; the test set is input into the trained fuzzy neural network, and the result is used as the final evaluation of the model. The method can effectively resist the situation of data loss, data distortion and signal saturation in the sampling data of non-flame detection channels in flame detection.

Description

technical field [0001] The invention belongs to the technical field of infrared flame recognition, and in particular relates to a TS-RBF fuzzy neural network robust fusion algorithm applied to infrared flame recognition. Background technique [0002] Flame detectors based on infrared pyroelectric sensors are widely used in the flame detection of modern industrial hydrocarbons, and are an important part of the automatic operation of industrial production systems and a necessary safety device. Hydrocarbon flames and most non-flame interferences have a fixed wavelength range in the infrared spectrum, so they can be analyzed and identified in a variety of ways. It is well known that flame identification becomes more complex and difficult when processing sensor data from real industrial environments, especially with multiple sensors, each operating at a different wavelength. There are some common difficulties in traditional multi-channel infrared sensor signal processing schemes...

AI Technical Summary

Problems solved by technology

[0004] 1. The traditional TS-RBF fuzzy neural network often requires a large number of hidden layer nodes (fuzzy rules) for excellent local approximation performance

[0005] 2. In the traditional TS-RBF modeling of infrared flame detection, there may be some features that have a large dispersion and do not satisfy the Gaussian characteristic, and the Gaussian membership function is used to calculate the membership degree, which is obviously inaccurate

[0006] 3. After defuzzification is added to the RBF fuzzy neural network to improve the generalization ability of the model, the model cannot suppress the output of outliers, and the model is not effective in dealing with uncertainty, and the outliers cannot be grouped into one category

[0007] 4. In infrared flame detection, due to hardware problems, system aging, transmission failure, strong external interference and other factors, data loss, data distortion, and signal saturation in the sampling data of non-flame detection channels will cause severe changes in some characteristics, which will cause the entire Fuzzy System Collapse

Images