Prediction of work order standard production times with iterative model corrections for high variaty production environments

In order to analyze the capacity and efficiency in a production workstation and to detect the accepted abnormal situations, first of all, standard production methods and standard production time must be defined. In this study, a workstation that manufactures for the aviation industry, where it is difficult to determine a standard production time, that processes a wide variety of parts for about 80 different projects and whose part group is constantly changing, is discussed. Within the scope of the study, a method has been developed for the prediction of work order production times. In the proposed approach, the selection criterion of the workstation to be studied and the current situation analysis to be made before the study are explained, and the modeling of the production time prediction with time study is discussed. Observations within the scope of time study are collected using video recording method. By analyzing the observations, the times for the determined standard production activities are obtained. Work order production time prediction model is created and explained. An additive model was created for the prediction of the work order standard production time, and two different statistical methods are used to predict the standard times obtained from the observations in the prediction of the model parameters and the times that vary according to the various characteristics of the manufactured part. The part properties information used in the statistical analysis methods are obtained from the tables in the database of the enterprise and the data is pre-processed before the analysis. In this context, "Multivariate Linear Regression Method" and "CART Method" are tested comparatively and the results are shared. The "Adjusted Shewhart Control Chart" is used to monitor the difference (residues) between the prediction times obtained using the work order standard production time prediction model and the production realizations and to improve prediction performance. Analyzes for monitoring are performed using the "spcadjust" library in "R Studio" software, and sample tracking graphics are given. The signals encountered by monitoring the performance of the established initial forecasting model using the adjusted control limits are examined. Due to the large diversity of parts, there are no observations on all parts, but a new prediction model is established by taking observations of parts that are not adequately represented in the model and caused high estimation error in signal examinations. After this stage, the future production realizations are compared with the initial model and the new model established, and the results are interpreted. With the iterative modeling proposed in the approach, it is aimed to increase the prediction performance by updating the model and control limits and the results are given with a sample application. Time and cost savings are achieved in work studies by updating the model with new observations as needed in production environments with a wide variety of parts.