Who learns when workers are trained? A case of safety training of maintenance contractors' workers for a major petrochemical plant shutdown

This study evaluated the safety training for petrochemical plant maintenance contractors' workers for a major plant shutdown at a refinery and the correlation between knowledge gained and levels of education. The training was divided into four courses, which were Introduction to Safety, Hydrogen Sulphide, Confined Space and Permit to Work, which were given before starting the job. It is normal practice in any oil, gas and petrochemical plant that contract workers (ranging from 700 to 3000 in Malaysia) are engaged to perform maintenance and modification jobs during any plant shutdown. The first and second levels of Kirkpatrick's evaluation training were chosen to evaluate the training exercise. Seven elements were identified for the first level, namely course objectives, course material, trainer, visual aid, feedback, learning and reaction. A survey was developed consisting of 49 questions and the data were analysed using SPSS. For the second level, post-tests were given to the respondents for each course and the scores were tabulated. The correlation between knowledge gained and level of education was significant for hydrogen sulphide only. It is recommended that, for a large group, it would be more effective to group the trainees according to their level of education. For better understanding of a cause-effect relationship in training, mathematical modelling and simulation studies were conducted and a strong correlation between training effectiveness and percentage age grouping was obtained. The curves obtained are typical for the whole exercise, which gave perfect straight lines for Introduction to Safety, Hydrogen Sulphide, Confine Space and Permit to Work, respectively. The curves' characteristics confirmed the model derived. Subsequently, detailed correlation using the model was performed on all E-i = f(A, X, L, P, R-p, R-R, T), which were plots of training effectiveness vs Ei as percentages, gave good quantitative parameters for further simulations in the future. Each model describes a common characteristic for all four courses.