Wong , Shi Yuen (2024) Intelligent tool wear condition prediction for CNC milling with modular neural network / Wong Shi Yuen. PhD thesis, Universiti Malaya.
PDF (The Candidate's Agreement) Restricted to Repository staff only Download (108Kb) | |
PDF (Thesis PhD) Restricted to Repository staff only until 31 December 2025. Download (2644Kb) |
Abstract
Cutting tool wear in Computer Numerical Control (CNC) milling is a gradual process which significantly affects product quality. Left unmonitored, risks of tool breakage would increase, leading to losses due to scrap and equipment damage. A Modular Neural Network (MNN), the Dissociation Artificial Neural Network (Dis-ANN), was proposed in this work for tool wear prediction. The Dis-ANN consists of a modular structure constructed out of parallel Artificial Neural Network (ANN) modules (referred to collectively as the Dissociation Unit), connected to an intermediary. The output of each ANN module is dependent on input feature vectors formed from the concatenation of both previous and current feature values, allowing each module to account for feature trends in a limited fashion. Three choices of complementary Dis-ANN optimization methods were also introduced – Selective Brute Force (SBF), Partial Correlation Evaluation (PCE), and Feature Selection with Partial Correlation Evaluation (FS+PCE). In addition to minimizing network redundancy in Dis-ANN, these optimization methods aim to optimize the number of time-steps between previous and current feature values in input feature vectors. The performance of Dis-ANN combined with each optimization method was investigated using the Universiti Malaya-Slot Milling Dataset (UM-SM Dataset), 2010 Prognostics and Health Management (PHM) Data Challenge Dataset, and NASA Ames Milling Dataset. The UM-SM Dataset contains data in the form of images of machined workpiece surfaces and acoustic signals during milling. In order to account for uneven lighting in each workpiece surface image, image features were extracted by processing texture descriptors based on the Grey-level Co-occurrence Matrix (GLCM) of different non-overlapping sections within the same image. For model validation using the 2010 PHM Data Challenge Dataset, Dis-ANN was tested using feature sequences extracted from the dataset under two conditions – with and without the addition of random noise in the feature sequences. Results showed Dis-ANN optimized with SBF or FS+PCE was better at learning complex non-linear relationships in tool wear trends compared with Linear Regression (LR), Support Vector Regression (SVR), and monolithic ANN. The Dis-ANN model optimized by FS+PCE had the possibility of achieving less accuracy than the Dis-ANN model optimized by SBF when the relationships in a dataset are highly complex or noisy. However, FS+PCE required much less computational time compared with SBF. Furthermore, FS+PCE optimized Dis-ANN to be more accurate when handling low-noise datasets. Dis-ANN optimized using PCE appeared to have low robustness to noisy datasets such as the UM-SM Dataset. In addition, further investigations on the performance of Dis-ANN optimized using FS+PCE with the UM-SM Dataset showed that features extracted from image data were beneficial for accurate tool Remaining Useful Life (RUL) predictions. This implied that images of machined surfaces, which can be considered a form of product quality data in industrial applications, have a certain level of utility in Tool Condition Monitoring (TCM), especially when used in conjunction with other sensor data. Moreover, results from model testing using the NASA Ames Milling Dataset showed that a properly optimized Dis-ANN had good generalization when handling machining data obtained from different experiments under different machining conditions.
Item Type: | Thesis (PhD) |
---|---|
Additional Information: | Thesis (PhD) - Faculty of Engineering, Universiti Malaya, 2024. |
Uncontrolled Keywords: | Cutting tool wear; CNC milling; Tool condition monitoring; Modular neural network; Remaining useful life |
Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering |
Divisions: | Faculty of Engineering |
Depositing User: | Mr Mohd Safri Tahir |
Date Deposited: | 13 Sep 2024 02:23 |
Last Modified: | 13 Sep 2024 02:23 |
URI: | http://studentsrepo.um.edu.my/id/eprint/15393 |
Actions (For repository staff only : Login required)
View Item |