BP Neural Network-Optimized Risk Evaluation of Dust Explosion for Dry Vacuum Cleaners and Its Experimental Verification

Abstract

Translate Article

Take Notes

ABSTRACT Dry dust extractors effectively capture dust and gas during industrial operation processes but create conditions conducive to environments where both coexist, increasing the risk of dust explosions due to the coexistence of combustible materials and ignition sources. Conventional risk assessment methods – such as scoring systems, risk matrices, expert judgment, and classification – are limited by subjectivity and limited applicability. To overcome these limitations, this study establishes a comprehensive explosion risk indicator system for dry dust extractors, structured into three primary and seven secondary indicators based on dust characteristics and equipment operation. Five types of dust – paper powder, goat-horn powder, water bull-horn powder, aluminum-titanium-alloy powder, and calcium-carbonate powder – were selected for analysis. Explosion parameters were obtained using a 1.2 L Hartmann tube, Godbert–Greenwald furnace, minimum ignition temperature device, and 20 L spherical explosion chamber. A total of 328 operational data samples from dust extractors were collected. A fuzzycomprehensive evaluation model was constructed, and a back propagation (BP) neural network was trained using this data set to establish a risk assessment framework. By integrating experimental data with fuzzy theory, the model provides an objective, data-driven approach for evaluating dust explosion risks in dry dust extraction systems.