PENGEMBANGAN MODEL ESTIMASI BIAYA PADA DESAIN BERBANTUAN KOMPUTER SOLIDWORKS MENGGUNAKAN PEMBELAJARAN MESIN

• Main Author: Christian Hermawan, Nicholas
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77833
• Institution: Institut Teknologi Bandung

CSM (Cipta Sinergi Manufacturing) CV is a company engaged in the field of machinery products manufacturing with orders such as molding, dies, spare parts, and other products using a make-to-order strategy. One of the activities carried out at CV CSM is the estimation of CAD (Computer Aided Design) costs for custom order products. In the process of product designing using CAD, the company estimates design costs through the approach of a percentage of the cost of goods manufactured (COGM) and adjustments based on drafter's modifications. This results in inconsistent and less representative values of design complexity. In response, this research is conducted to design a cost estimation model based on the complexity level of product designs, defined as the number of feature uses at the part, assembly, and drawing levels. This research designs a scheme and data acquisition method through the Solidworks API (Application Programming Interface) to calculate the number of features used in CAD documents. Meanwhile, the cost estimation model in this study is developed using machine learning techniques and refers to the CRISP-DM methodology engineering standards. Four algorithms are utilized: random forest, extreme gradient boosting, multi-linear regression, and k-nearest neighbors. The random forest algorithm, in particular, exhibits the best performance metrics compared to other algorithms, with an R2 of 0.97 on testing data and a Maximum Absolute Percentage Error (MAPE) of 25% on testing data. Based on the best machine learning model, a prototype software for design cost estimation is developed using the Python programming language and the Tkinter library. The resulting model is then used to predict new datasets not used in the model development process, yielding an average error rate of 23.414%. This value signifies that the developed model is capable of accurately predicting design costs in the industry.