An audit-based prediction model for aviation maintenance safety

Human error has become the major cause of aviation accidents around the world for decades. Airlines regard one of the main sources of human failures as the technical personnel of their maintenance department. In order to explore the causal relationship between maintenance error and flight safety, a reliable human error taxonomy and validated safety prediction methodology is much needed. This study integrated the human error concepts in aviation industry and developed a classification framework for aviation maintenance audit systems (HFACS-MA). The reliability of the HFACS-MA was examined through measuring the inter-rater agreement between independent participants. Audit records from an aviation authority were classified via the HFACS-MA to quantify the human error rates of aviation maintenance systems. A safety prediction model was developed based on the analysis of human error using both multiple regression and neural network method. The prediction results reached a considerable level with correlation coefficients around 0.6, and showed that the neural network method had better prediction performance than the regression method. This study proved the causality between human error and flight safety, and showed that the safety prediction model could be used to facilitate the application of human factors concepts in proactive accident prevention. Key words: Human Error, Maintenance, Audit, HFACS-MA, Regression, Neural Network