Advancements in manufacturing often hinge on the ability to monitor and maintain quality during production. A recent study has introduced a groundbreaking method for simultaneously monitoring surface roughness and tool wear, which are both pivotal factors in the machining process. With the rise of artificial intelligence (AI) and sensor technology, this method aims to optimize CNC machining efficiency and product quality.
Developed by researchers led by Wenwen Tian, this novel two-task simultaneous monitoring technique enhances data utilization and reduces computational costs. It incorporates state-of-the-art AI models within the CNC machine tool environment to facilitate real-time tracking of machining quality indicators.
Surface roughness is not just a measurement of product finish; it directly impacts various properties of the finished parts, including wear resistance and sealing effectiveness. Tool wear during machining can lead to less accurate dimensions and compromised part integrity—factors which can cause costly downtime. Therefore, integrating effective monitoring techniques is more than just beneficial; it is necessary for optimizing manufacturing processes.
To validate this new method, the researchers conducted end-face milling experiments using a vertical machining center (VMC850B). The workpiece material was 45 steel, measuring 125 mm by 125 mm by 120 mm. A four-flute flat end mill (UP210-S4-10025) was employed, with the tool set to specific parameters: milling depth at 1.2 mm, milling width at 10 mm, spindle speed at 3800 rpm, and feed speed at 600 mm/min, complemented by air cooling.
During the milling operations, data on vibration, current, and cutting force were simultaneously collected, sampled at 20 kHz. Off-machine, tool wear was measured using a microscope, whereas surface roughness was assessed on-machine with portable equipment, capturing the Ra value as the surface roughness index.
This systematic approach revealed significant results: of the 816 tool strokes conducted, 63 sets of valid sample data were obtained. The researchers utilized Kernel Principal Component Analysis (KPCA) for nonlinear dimensionality reduction, optimizing data processing and measurement accuracy.
Noteworthy was the introduction of the Broad Echo State Two-Task Learning System (BESTTLS), which replaced traditional enhancement layers with dynamically adaptable reservoirs. This enhancement captured the distinct characteristics of each monitoring task and enabled information sharing between them, thereby improving prediction accuracy.
The effectiveness of the BESTTLS was impressive, achieving a Mean Absolute Percentage Error (MAPE) of just 5.75% for surface roughness prediction, alongside achieving 100% accuracy for monitoring tool conditions. When benchmarked against two other systems—the Broad Two-Task Learning System (BTTLS) and the Fuzzy Broad Two-Task Learning System (FBTTLS)—BESTTLS stood out significantly. Its MAPE showed lower prediction errors compared to BTTLS at 7.57% and FBTTLS at 10.14%, highlighting its superiority.
The experimental results validate the developed two-task simultaneous monitoring method's robustness and efficiency. "The findings indicate our method is superior to traditional two-task learning approaches," wrote the authors of the article. The dual focus on surface roughness and tool wear not only enhances quality control during machining but also promises cost savings and reduced production time—a win-win for manufacturers aiming for excellence.
This advancement marks another step forward in leveraging AI for intelligent manufacturing and lays the groundwork for future developments. With this novel monitoring system, the manufacturing industry can expect improved machine health and part quality, contributing significantly to the overarching goal of zero-defect manufacturing.