International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 10 Issue: 05 | May 2023
p-ISSN: 2395-0072
www.irjet.net
ANALYSIS OF TOOL WEAR IN MILLING USING MACHINE LEARNING TECHNIQUE Sushant J.1, Manas S.2, Abhishek P.3, Amey O.4, Ajaykumar U.5 1Sushant J. Student, MIT School of Engineering, Pune 2Manas S. Student, MIT School of Engineering, Pune
3Abhishek P. Student, MIT School of Engineering, Pune 4Amey O. Student, MIT School of Engineering, Pune
5Ajaykumar U. Associate Professor, MIT School of Mechanical Engineering, Pune
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Predicting tool wear while machining is a
sophisticated optical sensors, such as laser displacement sensors, have been used to monitor tool conditions and identify the status of tool wear in real-time. However, these methods are still challenging to utilize in the sector because to their high cost and challenging assembly. Data-driven solutions are becoming more and more successful at predicting downtime and keeping track of the health of machine tools because to substantial advancements in computing and data science. Large volumes of data are being gathered during machining due to the availability of better sensor technologies, necessitating the need of a robust feature selection approach. The milling machine that was tested. In this paper the approach used for tool monitoring is based on vibration analysis and applying those vibration data (Acceleration Vs Frequency) on machine learning techniques to predict tool wear. FFT used for data acquisition. There are several methods to analyze vibration. Frequency domain works with higher frequency. It is observed that abnormal peaks are obtained at higher frequency.[2] So, acceleration taking as a parameter to vibration analysis. It is observed when machining with different tool conditions the acceleration peak also varies with respect to tool condition. Observations are then collected with higher frequency. New tool, medium worn tool and blunt tool is used for dataset preparation. In this project a logistic regression approach is used to describe the tool wear based on features that is determined by the multinomial logistic regression algorithm. The proposed method sets up a process for supervision and a predictive model on a milling machining process dataset.
difficult aspect. Traditional methods to use process characteristics that affect tool wear are available, however, some parameters are particular to the machining process, and existing prediction models fail. The current work discusses a process supervision system that uses machine learning (logistic regression) to anticipate tool wear. An application for the prediction of tool wear while milling is chosen as a case study to demonstrate the methodology. The next dataset will be created by running the milling operation with the end mill cutter under three different conditions, namely 1. New Tool, 2. Medium Wear Tool, and 3. Blunt Tool, and recording the vibration reading as acceleration and frequency using an FFT analyzer. There are many vibration analysis techniques, but choosing the best one requires evaluating the parameters and surroundings of the milling operation. The frequency domain is used for vibration analysis. Utilizing the Logistic Regression Method, train the acquired dataset and predict accuracy, as well as the tool condition using this prediction. The accuracy of the trained model is 99.1%. Now with the obtained accuracy, it is possible to implement this algorithm in industrial working conditions to accurately predict the conditioning of the tool. Key Words: tool wear prediction, logistic regression, milling machine, vibration Analysis, FFT Analyzer.
1.INTRODUCTION To be competitive, manufacturing sectors prioritize improving both quality and production costs. The 'Quality at Source' idea mandates that quality be inspected at every stage. Poorly completed items are the result of using wornout tools for milling. Due to subpar output, a bad finish results in quality loss and rejection. Costs will go up if a tool is used to prevent a drop in quality.[1] Track the machining quality by foreseeing tool wear, monitoring the Tool Life Cycle has gained popularity in the research community. It is challenging and complex to mathematically describe the machining process. There are attempts to propose Artificial Neural Network (ANN) based approaches to predict that a tool is worn. The above approaches lack a robust method to select features to predict tool wear. Techniques with specific
© 2023, IRJET
|
Impact Factor value: 8.226
2. LITERATURE REVIEW Pooja V. Kamat, et.al.[1] The key findings are that machine learning techniques help in making machines more accurate. This paper discussed a comparative approach to tool wear monitoring using the clustering machine learning technique of K-Nearest Neighbour (k-NN) and deep learning technique of Convolutional Neural Network (CNN). The CNN and AELSTM techniques out-perform k-NN by achieving a higher degree of separability of around 93% and 87%, respectively. The techniques provide improved outcomes in terms of precision, recall, and f1-score, indicating that the models are more accurate at detecting false positives.
|
ISO 9001:2008 Certified Journal
|
Page 1602