Applying data mining techniques to analyze the causes of major occupational accidents in the petrochemical industry

Abstract Accidents that occur in the petrochemical industry frequently result in serious social issues. Behind every occupational accident, there are safety management problems requiring investigation. This study collected 349 cases of major occupational accidents in the petrochemical industry between 2000 and 2010 in Taiwan for analysis. Using descriptive statistics, we elucidated the factor distribution of these major occupational accidents. The data mining classification and regression tree (CART) was used to examine the distribution and rules governing the factors of the disasters. This study found that for equipment such as pipelines and control valves, devising high-quality safety and protective devices/maintenance/renewal plans and pipeline setups/design plans can effectively prevent accidents such as fires, explosions, and poisoning caused by material leakage, as well as employees being caught in/rolled up in machinery. Furthermore, implementing safety management measures, such as worker safety educational training, and enforcing standards for inspections, operations, and risk assessments personnel, has become an important factor in accident prevention. This study suggests the use of the following measures: for abnormal conditions such as pipeline cracking/damage or rusting, high-temperatures caused by material leakage into the inner protective layer of pipelines should be prevented. Considering overlapping pipelines, rusting issues caused by pipelines touching each other should be avoided, and maintenance and repair should be performed to ensure the safety of work environments. These measures can eliminate the risk of work injuries and resulting social issues.