An adaptive neuro-fuzzy inference system (anfis) model for assessing occupational risk in the shipbuilding industry

• Published in: Safety science Vol. 63; pp. 226 - 235
• Main Authors: Fragiadakis, N.G., Tsoukalas, V.D., Papazoglou, V.J.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier India Pvt Ltd 01.03.2014; Elsevier
• Subjects: Applied sciences; Artificial neural networks; Biological and medical sciences; Exact sciences and technology; Fuzzy inference; Ground, air and sea transportation, marine construction; Marine construction; Medical sciences; Occupational accident; Occupational medicine; Public health. Hygiene-occupational medicine; Risk assessment; Shipbuilding industry; Fuzzy inference; Occupational accident; Shipbuilding industry; Risk assessment; Artificial neural networks; Fuzzy system; Adaptive system; Inference; Risk; Neural network; Occupational exposure; Modeling; Industry; Risk evaluation; Models; Shipbuilding; Occupational medicine
• ISSN: 0925-7535; 1879-1042
• DOI: 10.1016/j.ssci.2013.11.013

•Occupational accident data can be used to create an ANFIS model.•ANFIS algorithm can effectively predict Occupational Accident Risk values.•Working conditions have a strong effect on occupational risk values. In this research an adaptive neuro-fuzzy inference system (ANFIS) has been applied to study the effect of working conditions on occupational injury using data of professional accidents assembled by ship repair yards. The data were statistically processed in order to select the most important parameters. These parameters were day and time, specialty, type of incident, dangerous situations and dangerous actions involved in the incident. The selected parameters proved, due to statistical processing, to be correlated to the observed frequency of four injury categories, namely negligible wounding, slight wounding, severe wounding and death. For each parameter a Gravity Factor (GF) was calculated based on the percentage of injury categories resulting to the incident each of the above mentioned parameter was involved. These GF values and the resulting risk value based on the accident data were used as input values to train the ANFIS model. Trapezoidal and Gauss membership functions were used for the training of the data. The model combined the modeling function of fuzzy inference with the learning ability of artificial neural networks. A set of rules has been generated directly from the statistically processed reported data. The model’s predictions were compared with a number of recorded data for verifying the approach.