Tool condition monitoring of friction drilling process using adaptive neuro-fuzzy inference system

Friction drilling is a new progressive hole-making method. The interaction of the rotating conical tool and the thin workpiece produces heat allowing penetration of the tool and soften the work-piece forming a hole and bush in one process. The process is environmentally friendly since bush formation creates no material wastage and requires no coolant fluid during the machining process. However, performing the machining process with a worn tool can increase the friction between the tool-workpiece and the late replacement of the worn tool may cause an unpredictable machine breakdown. The focus of the present study is to develop an AI-based expert system for tool condition monitoring (TCM) in the friction drilling process. Thus, the TCM was developed by detecting the machining signals through signal processing and pattern recognition. Subsequently, the tool condition was predicted by the artificial intelligence (AI) approach. A tungsten carbide tool was used in this experiment of friction drilling on medium carbon steel AISI 1045. As preliminary experiments, to determine optimal processing parameters in the friction drilling process by considering multi-performance characteristics (i.e., bush length and roundness error), an effective grey relational analysis (GRA) approach has been used. Tool wear characteristics were quantified of friction drilling by analyzing the changes in tool shape and weight reduction. TCM in the friction drilling process was developed based on the vibration signal collected through accelerometer sensors of the machining signals through a low-pass filter. Three approaches AI-model such as Artificial Neural Network (ANN), Fuzzy Logic (FL), and Adaptive Neuro-Fuzzy Inference System (ANFIS) used to boost the efficiency of the prediction system to anticipate the state of the tool in terms of the tool length and angle. The outcomes of the established models were compared in terms of prediction accuracy to find the best performing model. Therefore, real-time condition monitoring took part to verify the TCM system for the friction drilling process. The GRA obtained 3000 rpm of spindle speed and 50 mm/min of feed rate the best combination of processing to achieve a greater bush length and lower roundness error. The tool wear characteristic can be confirmed that the abrasive wear revealed in the conical region with circular grooves. The adhesive wear was observed at the tool centre and conical regions, and oxidation wear was identified with a dark burned appearance at the tool surface. The development of the AI-model model shows excellent performance, which the R-squared correlation shows the ANFIS model was 97.2% and 97.1% for tool length, and the angle at the training phase seen an increase to 98.4% and 98.2% in the testing phase. It was verified in the real-time TCM experiments that the ANFIS-based expert system was successfully developed and utilized in monitoring the tool condition by categorizing the level of condition into three distinct categories, i.e., good, half-life, and worn-out conditions.

Text FK 2021 107 - IR.pdf Download (1MB)

Actions (login required)

View Item