Developing automatic recognition system of drill wear in standard laminated chipboard drilling process

Abstract

Abstract. The paper presents an automatic approach to recognition of the drill condition in a standard laminated chipboard drilling process. The state of the drill is classified into two classes: "useful" (sharp enough) and "useless" (worn out). The case "useless" indicates symptoms of excessive drill wear, unsatisfactory from the point of view of furniture processing quality. On the other hand the "useful" state identifies tools which are still able to drill holes acceptable due to the required processing quality. The main problem in this task is to choose an appropriate set of diagnostic features (variables), based on which the recognition of drill state ("useful" versus "useless") can be made. The features have been generated based on 5 registered signals: feed force, cutting torque, noise, vibration and acoustic emission. Different statistical parameters describing these signals and also their Fourier and wavelet representations have been used for defining the features. Sequential feature selection is applied to detect the most class discriminative set of features. The final step of recognition is done by using three types of classifiers, including support vector machine, ensemble of decision trees and random forest. Six standard drills of 12 mm diameter with tungsten carbide tips were used in experiments. The results have confirmed good quality of the proposed diagnostic system. Key words: diagnostic expert systems, neural networks, wavelet packets, wear monitoring.Developing automatic recognition system of drill wear in standard laminated chipboard drilling process [5,7].The key issue in diagnostics of cutting tool state is the selection of sensors. Usually force sensors, electrical power, acoustic emission, vibration and acoustic pressure are applied in TCM [4,8]. The best results in metal working are achieved using force sensors.The disadvantage of this kind of measurement in normal production is that the force or torque transducers are relatively expensive and there are difficulties in mounting these sensors to the cutting tool or work piece. Therefore vibration sensors -easy to install, but at the same time less accurate because of background noise -are most commonly used in wood industry [5]. Vibration measurement is easy, since an accelerometer can be mounted close to the spindle bearing and no modifications of machine tools are needed [9].Quite similar to vibration is the sound signal, which can also be used in assessment of drill state. Mechanical vibration of the cutting tool, machine holder and drill are partly transferred to airborne vibration. It means that part of the information contained in vibration can also be obtained from sound measurement. The acquisition of the sound can be done easily using a microphone. However, the sound pressure sensors (microphones), are even more susceptible to background [7,8]. The sound measurement in an extended range of frequencies from 20 kHz to 80 kHz is usually applied [4]. This range is called the acoustic emission. The effectiveness of acoustic emission...

Tool State Recognition During Milling Process using Time Series Signals and Multiple Input CNN Architecture.

“…Multiple steps require high levels of precision. At the same times, poor or ill-timed decisions about tool maintenance and/or exchange can result in a faulty product, not meeting the overall requirements and cause potential loss for the manufacturing company [Hu et al, 2019, Jegorowa et al, 2021, Kurek et al, 2019a, Osowski et al, 2016.…”

Decision Confidence Assessment in Multi-Class Classification

“…For example, the main focus may include evaluating the state of elements without interrupting the actual manufacturing process, as presented in [ 1 ]. Most basic and commonly used approaches, such as the one presented in [ 2 ], measure different signals, such as vibration, noise, acoustic emission, cutting torque, feed force, and others, in order to evaluate the tool state. Similar approaches were used in [ 3 ], where data were extracted both from signal and frequency domains, along with wavelet coefficients, all in order to evaluate the obtained elements automatically, checking how relevant each item was to the selected problem.…”

Multiclass Image Classification Using GANs and CNN Based on Holes Drilled in Laminated Chipboard

“…The judgement of the state of a drill is not simple, and relying only on an expert’s eye would be quite risky. A traditional approach to this problem is collecting and measuring multiple signals produced by the drill, like the feed force, the cutting torque, the noise, the vibration, or the acoustic emission and then estimating its quality based on these signals [ 1 ]. This approach gives acceptably accurate results, as it was shown in previous works [ 2 , 3 , 4 ], but it requires the usage of multiple sensors.…”

“…This approach gives acceptably accurate results, as it was shown in previous works [ 2 , 3 , 4 ], but it requires the usage of multiple sensors. Many pre-processing operations have to be performed on collected data, such as calculating a number of statistical parameters of recorded signals or generating Fourier representations for specific feature selection [ 1 ].…”