Job Labour-Hours Forecasting for a Pipe Spool Fabrication Shop Using Data Mining

Due to the uniqueness of each product, the fabrication of pipe spool is a complex and variable process. Typically, in fabrication shop, the production goals are set based on the diameter inches of welding to be produced from the mix of various types of pipe spools termed as “production mix”. The dynamic elements of spool production and the variability in engineer to order, however, make it challenging to develop reliable predictions of job labour-hours required for any given production mix. This research suggests a way for predicting the labour-hours needed in a pipe spool fabrication facility for any production mix using data mining techniques. The study framework has two sections: one for analysing historical fabrication data and the other for developing forecasting model. Firstly, the study used machine learning techniques to analyse 10 years of historical data and correlated the quantity of labour-hours utilized for mix of pipe spools with varying characteristics, such as material types, weld types, and weld processes. The forecasting model was then developed using five machine learning algorithms: multiple linear regression, support vector machine, random forest, decision tree, and adaboost to forecast labour-hours required for any job. The models were tested using test data set, and the results demonstrate that, in comparison with methods currently employed in the field, data mining is more useful in providing low error estimations. The study's models should help guide and promote the implementation of more accurate labour-hours forecasting techniques in pipe spool production facilities.