Text mining and association rules-based analysis of 245 cement production accidents in a cement manufacturing plant.

Abstract

Accidents such as collapses, fires, explosions and mechanical injuries occur frequently in cement manufacturing plants. Understanding the causes of past accidents is essential to prevent future incidents and reduce safety risks. Hence, this article analyzes cement accident cases based on a unified report analysis framework. By integrating text mining technology, the article identifies patterns in cement production accidents and establishes a cement accident causation analysis model to support safety management decisions. First, 245 accident records were categorized using the latent Dirichlet allocation model to identify causal factors. Subsequently, a systematic accident causal analysis based on the 24Model was proposed to establish a unified report framework. An improved Apriori algorithm was then developed for multidimensional, multilayer correlation rule mining in cement enterprises, enhancing text mining efficiency. By applying this algorithm, the study quantitatively analyzed correlations between accident types, causative factors and their interactions. Finally, targeted safety management recommendations were formulated.