PENGEMBANGAN MODEL ESTIMASI WAKTU PERMESINAN CNC MENGGUNAKAN JARINGAN SARAF TIRUAN

• Main Author: Ahmad Thoriq, Dimas
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/74390
• Institution: Institut Teknologi Bandung

PT X is a make-to-order (MTO) machining product manufacturing company. PT X needs to estimate each order’s lead time in order to estimate the cost at an early stage of the order cycle, due to the unique nature of orders in the MTO industry. These time and cost estimations would then be used to negotiate their proposed fee to the customer. PT X currently utilizes CAM software to estimate their CNC machining time, which turns out to produce a recognizable figure of deviation from the actual CNC machining time. This research aims to develop a CNC machining time estimation method using a machine learning approach, an artificial neural network model, to utilize the abundant machining data available in PT X. The development of the CNC machining time estimation model uses an Artificial Neural Network (ANN) model as the proposed model and a Multiple Linear Regression model for benchmarking purpose. The model development adopts a popular cross-industry standard for data mining projects, the CRISP-DM framework. The ANN model proved to be superior in accuracy and reliability against the benchmark model, thus being deployed in the proposed software prototype during the implementation test. The test result using 62 rows of testing data shows that the proposed ANN model is capable of estimating unseen data in PT X quite accurately, recording RMSE of 196.35 seconds with 147,49 seconds of absolute error standard deviation. This level of performance is equal to reducing 72% of the RMSE produced by the current method of estimation in PT X during the implementation test. Several machining parameters such as cut length and stepover showed to be significant towards the CNC machining time.