Hui En Lee

In the field of Industry 4.0 and smart manufacturing, intelligent Tool Condition Monitoring (TCM) is essential for enabling predictive maintenance and ensuring product quality in machining operations. This research presents the development of a data-driven TCM system capable of acquiring and analyzing machine data for tool life optimization. The study utilizes hybrid data acquisition architecture, integrating internal machine data (Spindle Load) with external sensing technologies, including vibration sensors and a threephase smart energy meter, combined via an (internet of things) IoT gateway and Node-Red platform. Experimental testbeds were conducted on a FANUC ROBODRILL $\boldsymbol{\alpha}$-D21MiB Plus equipped with 5-Axis Rotary Table using SAE 1020 carbon steel to investigate tool wear progression under Conventional and High-Speed Machining (HSM) strategies. Direct offline measurements were performed using a Vision Measuring Machine (VMM) to categorize tool conditions for preexperiment and post-experiment. Three Machine Learning classification models were evaluated: Random Forest, Gradient Boosting, and Support Vector Machine (SVM). The Random Forest classifier achieved the highest reliability with an accuracy of 73.82 %, effectively distinguishing between operational modes and wear states. The poor performance of SVM model (34.70 %) highlighted the high non-linearity and scatter in the dataset.

Malaysia, Sarawak state in particular produces superior quality black pepper compared to other producing countries. While Sarawak black pepper deserves higher price in the global market, from year 2015 onwards, its price has been plummeting due to global oversupply of pepper. Physical, chemical and biological profiling of Sarawak pepper is conducted to differentiate Sarawak pepper against adulteration and to justify its higher market price. However, the analysis of pepper was performed using time-consuming, non-portable, expensive and invasive laboratory equipment and methodologies. This paper reports an attempt to develop a black pepper geo-tracing system using metal oxide semiconductor (MOS) gas sensors array. The pneumatic system to contain and purge the sample volatile organic compounds (VOCs) is designed, simulated, fabricated and proven to be contain and purge sample VOCs effectively. A total of 200 black pepper samples from Malaysia and India are sampled using non-invasive method for the MOS gas sensors response and the sensors response are analyzed using Principal Component Analysis (PCA). The PCA scatter plots shows that Malaysia and India black pepper can be discriminated based on the MOS gas sensor response towards different VOCs composition in the black pepper samples. The PCA result are then used for geo-tracing of a total of 40 samples (20 from Malaysia; 20 from India), which shows an overall accuracy of 92.5%. The geo-tracing system developed performs rapidly with 260-600 s sampling cycle compared to gas chromatography-mass spectroscopy (GC-MS) equipment which requires at least 30 minutes sampling time, and is portable, cheaper and non-invasive.