Improved Inversion Algorithm Based on Pattern Recognition and Mend for Coal Dust Measure

To increase dust measurement precision and real-time capacity, an improved inversion algorithm of pattern recognition and mend was presented. The pattern classification was performed according to the dust diffraction angle spectrum. Simulation results indicated the minimum recognition time reduced to 0.05 times of that before. On the basis of classification, a number of mend patterns were supplemented reasonably. The optimal pattern was recognized in universe and mended in local area according to the principle of the minimum of variance sum. Then the dust content could be inversed with the total light energy ratio of real-time signal to optimal pattern. The error of total and respiring coal dust declined from 6% to 3% and from 9% to 3.5%, respectively. Underground operations testify the sensor precision achieves 95%. Practice proves the improved inversion algorithm enhances the measurement precision and real-time capacity of dust sensor remarkably.