Applying a Hybrid Data Mining Approach in Machining Operation for Surface Quality Assurance

Conventionally, the quality of a machined product has been measured based on the specifications, once the machining process is complete. However, this post-process inspection has several shortcomings: (1) it is difficult to isolate the causes of the defect; (2) the manufacturing cost has already been incurred when a non-conformance is detected; (3) rework of any scope increases the manufacturing cost and can be very difficult to accomplish; and (4) there could be a significant time lag between the detection of the defects and subsequent corrective actions. Today, efforts of manufacturers are shifting from the postprocess inspection to improved monitoring of the manufacturing processes, utilizing sensors and other measurement devices, to effectively control the process. Improvements in machining precision can only be accomplished by the development of manufacturing systems that are capable of monitoring processes. Process monitoring reduces scrap, rework, lead-time, and conventional non value-added inspection activities, thereby, increases the system’s productivity. The monitoring has to be based on sound, reliable process control algorithms. Computer numerical control (CNC) of machine tools do help to produce consistent part quality. However, in most cases, CNC machines don’t utilize sensor data to compensate for anomalies generated by the cutting processes (e.g., tool wear, chatter, incorrect machine setup, etc.). If sensors such as cutting force, vibration and spindle motor current were integrated into CNC machine tools, the control functions should be able to interpret and respond to sensory data as the process continues. However, when many process variables need to be considered, it becomes rather difficult to predict quality attributes in machining (i.e., surface roughness). To solve the aforementioned prediction problems, especially with the consideration of negative information and data to improve prediction accuracy, two data mining approaches have been developed. Here, negative information or