Filtering Optimization For On-line Spectral Data And TDLAS System Integration

Spectroscopic analysis has been widely used in metallurgy, chemical, pharmaceuticals industry, machinery, advanced materials development, aerospace and other fields with the advantages of non-destruction, remote sensing, high sensitivity and high precision. However, the spectral data always contain noise and disturbance signals, so it is significant to do data preprocessing. Savitzky-Golay smoothing filter is the most commonly method used in data processing, and it has been optimized by many scientists since its proposition in the 1964s.Aiming at the optimization and choice of its input parameters while applying Savitzky-Golay smoothing filter and the integration of TDLAS system, this research includes:This paper studied the frequency characteristics of Savitzky-Golay smoothing filter with different input parameters (filter length and fitting order), performed the simulations by matlab, and then got the relationships between primary frequency characteristic indices and filter length and fitting order by data fitting. Based on the relationships, we got the Savitzky-Golay smoothing filter's optimum parameters by making the filter meet the frequency characteristic requirements of the raw spectral data. Simulation data and the experiment data were both improved notably using this method. Moreover, the experiment data acquired from a pulsed laser, where the noise and signal characters varied significantly with time, was dealt with twice varying filter length Savitzky-Golay smoothing technology, and the results verified the method's effectiveness.In addition, TDLAS gas monitoring system was integrated. And all of the program-controlled instruments in the system working in different mode were controlled by a LabVIEW program through GPIB bus. The program could also analyze and deal with spectral data collected by a data acquisition card, forming a highly automated gas monitoring device.The innovation is that this paper gives the relationships between primary frequency characteristic indices and filter length and fitting order, proposes the method of obtaining Savitzky-Golay smoothing filter with optimum parameters, and then solves the much-condemned shortcoming(the blindness of selecting the input parameters).