Wyniki wyszukiwania - BazTech

1

Milling cutter fault diagnosis using unsupervised learning on small data: A robust and autonomous framework

Patange Abhishek, Soman Rohan, Pardeshi Sujit, Kuntoglu Mustafa, Ostachowicz Wieslaw

Eksploatacja i Niezawodność

|

2024

|

Vol. 26, no. 1

art. no. 178274

EN

Tool condition affects the tolerances and the energy consumption and hence needs to be monitored. Artificial intelligence (AI) based data-driven techniques for tool condition determination are proposed. Unfortunately, the data-driven techniques are data-hungry. This paper proposes a methodology for classification based on unsupervised learning using limited unlabeled training data. The work presents a multi-class classification problem for the tool condition monitoring. The principal component analysis (PCA) is employed for dimensionality reduction and the principal components (PCs) are used as input for classification using k-means clustering. New collected data is then projected on the PC space, and classified using the clusters from the training. The methodology has been appliedforclassification of tool faults in 6 classes in a vertical milling center. The use of limited input parameters from the user makes the method ideal for monitoring a large number of machines with minimal human intervention. Furthermore, due to the small amount of data needed for the training, the method has the potential to be transferable.

2

Towards sustainable and intelligent machining: energy footprint and tool condition monitoring for media-assisted processes

Dogan Hakan, Jones Llyr, Hall Stephanie, Shokrani Alborz

Journal of Machine Engineering

|

2023

|

Vol. 23, No. 2

16--40

EN

Reducing energy consumption is a necessity towards achieving the goal of net-zero manufacturing. In this paper, the overall energy footprint of machining Ti-6Al-4V using various cooling/lubrication methods is investigated taking the embodied energy of cutting tools and cutting fluids into account. Previous studies concentrated on reducing the energy consumption associated with the machine tool and cutting fluids. However, the investigations in this study show the significance of the embodied energy of cutting tool. New cooling/lubrication methods such as WS2-oil suspension can reduce the energy footprint of machining through extending tool life. Cutting tools are commonly replaced early before reaching their end of useful life to prevent damage to the workpiece, effectively wasting a portion of the embodied energy in cutting tools. A deep learning method is trained and validated to identify when a tool change is required based on sensor signals from a wireless sensory toolholder. The results indicated that the network is capable of classifying over 90% of the tools correctly. This enables capitalising on the entirety of a tool’s useful life before replacing the tool and thus reducing the overall energy footprint of machining processes.

3

Time-frequency Representation -enhanced Transfer Learning for Tool Condition Monitoring during milling of Inconel 718

Zhou Yuqing, Sun Wei, Ye Canyang, Peng Bihui, Fang Xu, Lin Canyu, Wang Gonghai, Kumar Anil, Sun Weifang

Eksploatacja i Niezawodność

|

2023

|

Vol. 25, no. 2

art. no. 165926

EN

Accurate tool condition monitoring (TCM) is important for the development and upgrading of the manufacturing industry. Recently, machine-learning (ML) models have been widely used in the field of TCM with many favorable results. Nevertheless, in the actual industrial scenario, only a few samples are available for model training due to the cost of experiments, which significantly affects the performance of ML models. A time-series dimension expansion and transfer learning (TL) method is developed to boost the performance of TCM for small samples. First, a time-frequency Markov transition field (TFMTF) is proposed to encode the cutting force signal in the cutting process to two-dimensional images. Then, a modified TL network is established to learn and classify tool conditions under small samples. The performance of the proposed TFMTF-TL method is demonstrated by the benchmark PHM 2010 TCM dataset. The results show the proposed method effectively obtains superior classification accuracies for small samples and outperforms other four benchmark methods.

4

Intelligent cyber-physical monitoring and control of I4.0 machining systems – an overview and future perspectives

Hassan Mahmoud, Sadek Ahmad, Attia M. Helmi

Journal of Machine Engineering

|

2022

|

Vol. 22, No. 1

5--24

EN

Rapid evolution in sensing, data analysis, and industrial internet of things technologies had enabled the manufacturing of advanced smart tooling. This has been fused with effective digital inter-connectivity and integrated process control intelligence to form the industry I4.0 platform. This keynote paper presents the recent advances in smart tooling and intelligent control techniques for machining processes. Self-powered wireless sensing nodes have been utilized for non-intrusive measurement of process-born phenomena near the cutting zone, as well as tool wear and tool failure, to increase confidence in the process and tool condition monitoring accuracy. Cyber-physical adaptive control approaches have been developed to optimize the cycle time and cost while eliminating machined part defects. Novel artificial intelligence AI-based signal processing and modeling approaches were developed to guarantee the generalization and practicality of these systems. The paper concludes with the outlook for future work needed for seamless implementation of these developments in industry.

5

Data augmentation techniques for transfer learning improvement in drill wear classification using convolutional neural network

Kurek Jarosław, Antoniuk Izabella, Górski Jarosław, Jegorowa Albina, Świderski Bartosz, Kruk Michał, Wieczorek Grzegorz, Pach Jakub, Orłowski Arkadiusz, Aleksiejuk-Gawron Joanna

Machine Graphics and Vision

|

2019

|

Vol. 28, No. 1/4

3--12

EN

This paper presents an improved method for recognizing the drill state on the basisof hole images drilled in a laminated chipboard, using convolutional neural network (CNN) and data augmentation techniques. Three classes were used to describe the drill state: red - for drill that is worn out and should be replaced, yellow - for state in which the system should send a warning to the operator, indicating that this element should be checked manually, and green - denoting the drill that is still in good condition, which allows for further use in the production process. The presented method combines the advantages of transfer learning and data augmentation methods to improve the accuracy of the received evaluations. In contrast to the classical deep learning methods, transfer learning requires much smaller training data sets to achieve acceptable results. At the same time, data augmentation customized for drill wear recognition makes it possible to expand the original dataset and to improve the overall accuracy. The experiments performed have confirmed the suitability of the presented approach to accurate class recognition in the given problem, even while using a small original dataset.

6

Classifiers ensemble of transfer learning for improved drill wear classification using convolutional neural network

Kurek Jarosław, Antoniuk Izabella, Górski Jarosław, Jegorowa Albina, Świderski Bartosz, Kruk Michał, Wieczorek Grzegorz, Pach Jakub, Orłowski Arkadiusz, Aleksiejuk-Gawron Joanna

Machine Graphics and Vision

|

2019

|

Vol. 28, No. 1/4

13--23

EN

In this paper we introduce the enhanced drill wear recognition method, based on classifiers ensemble, obtained using transfer learning and data augmentation methods. Red, green and yellow classes are used to describe the current drill state. The first one corresponds to the case when drill should be immediately replaced. The second one denotes a tool that is still in a good condition. The final class refers to the case when a drill is suspected of being worn out, and a human expert evaluation would be required. The proposed algorithm uses three different, pretrained network models and adjusts them to the drill wear classification problem. To ensure satisfactory results, each of the methods used was required to achieve accuracy above 90% for the given classification task. Final evaluation is achieved by voting of all three classifiers. Since the initial data set was small (242 instances), the data augmentation method was used to artificially increase the total number of drill hole images. The experiments performed confirmed that the presented approach can achieve high accuracy, even with such a limited set of training data.

7

Intelligent machining: real-time tool condition monitoring and intelligent adaptive control systems

Hassan M., Sadek A., Attia M. H., Thomson V.

Journal of Machine Engineering

|

2018

|

Vol. 18, No. 1

5--17

EN

Unmanned manufacturing systems has recently gained great interest due to the ever increasing requirements of optimized machining for the realization of the fourth industrial revolution in manufacturing ‘Industry 4.0’. Real-time tool condition monitoring (TCM) and adaptive control (AC) machining system are essential technologies to achieve the required industrial competitive advantage, in terms of reducing cost, increasing productivity, improving quality, and preventing damage to the machined part. New AC systems aim at controlling the process parameters, based on estimating the effects of the sensed real-time machining load on the tool and part integrity. Such an aspect cannot be directly monitored during the machining operation in an industrial environment, which necessitates developing new intelligent model-based process controllers. The new generations of TCM systems target accurate detection of systematic tool wear growth, as well as the prediction of sudden tool failure before damage to the part takes place. This requires applying advanced signal processing techniques to multi-sensor feedback signals, in addition to using ultra-high speed controllers to facilitate robust online decision making within the very short time span (in the order of 10 ms) for high speed machining processes. The development of new generations of Intelligent AC and TCM systems involves developing robust and swift communication of such systems with the CNC machine controller. However, further research is needed to develop the industrial internet of things (IIOT) readiness of such systems, which provides a tremendous potential for increased process reliability, efficiency and sustainability.

8

Wykrywanie zużycia ostrza na podstawie analizy różnic pomiędzy zdjęciami z oświetleniem z różnych kierunków

Błażejak K.

Mechanik

|

2017

|

R. 90, nr 10

834--836

PL

W trakcie budowy systemu opartego na metodach wizyjnych najważniejszy jest dobór oświetlenia. Prawidłowo ustawione upraszcza analizę zdjęć. Opracowano prototyp systemu wykrywania obszaru starcia powierzchni przyłożenia noży tokarskich, ukazującego różnice pomiędzy kolejnymi zdjęciami tego samego ostrza oświetlonego z różnych kierunków.

EN

In developing a vision-based system, the most important thing is choosing the lighting. Properly set up simplifies image analysis. The prototype of the system for detecting wear of flank face, which detects the differences between successive images of the same tool illuminated from different directions, has been developed.

9

Diagnostyka stanu narzędzi i procesu skrawania

Jemielniak K.

Mechanik

|

2017

|

R. 90, nr 7

504--510

PL

Automatyczna diagnostyka stanu narzędzi i procesu skrawania jest oparta na pomiarach wielkości fizycznych skorelowanych z tym stanem. Z każdego sygnału da się wyznaczyć bardzo wiele miar i nie można przewidzieć, które z nich będą przydatne w określonym przypadku. Miary te muszą być zatem wybierane automatycznie, a następnie integrowane w jedno oszacowanie stanu, np. z wykorzystaniem metod sztucznej inteligencji.

EN

Automatic tool condition monitoring is based on the measurements of physical phenomena which are correlated with this condition. There are numerous signal features (SFs) that can be extracted from the signal. As it is really not possible to predict which signal features will be useful in a particular case they should be automatically selected and combined into one tool condition estimation. This can be achieved by various artificial intelligence methods.

10

Algorytm diagnostyki zużycia ostrza oparty na wielu sieciach neuronowych

Bombiński S., Kossakowska J.

Mechanik

|

2017

|

R. 90, nr 3

220--223

PL

Porównano różne sposoby określania zużycia ostrza – z wykorzystaniem sieci neuronowych RBF, metody hierarchicznej oraz standardowego zliczania czasu pracy. Analizę sygnałów z procesu skrawania przeprowadzono dla trzech różnych zestawów badań doświadczalnych. Wyniki otrzymane w przypadku zespołu sieci neuronowych są zbliżone do wyników z algorytmu hierarchicznego – jest to potencjalnie bardzo skuteczna metoda szacowania zużycia ostrza.

EN

Presented is a comparison of different methods of estimating tool wear – obtained for group of RBF neural networks, hierarchical methods and the standard time counting. The analysis of the signals from the machining process carried out for three different experiments, clearly demonstrating the effect of presented methods. The results obtained for group of RBF neural networks are similar to results obtained for hierarchical methods.

11

Możliwości zastosowania sieci neuronowej RBF do diagnostyki zużycia ostrza w systemie online

Bombiński S., Błażejak K., Nejman M.

Mechanik

|

2016

|

R. 89, nr 8-9

1056--1057

PL

Porównano wyniki dwóch metod szacowania zużycia ostrza: uzyskane dla sieci neuronowej RBF oraz metodą hierarchiczną. Analizę sygnałów z procesu skrawania przeprowadzono dla trzech różnych eksperymentów, jednoznacznie wykazując skuteczność obu metod.

EN

Presented are the results of a comparison of two different methods of estimating tool wear: obtained for RBF neural network and hierarchical methods. The analysis of the signals from the machining process carried out for three different experiments, clearly demonstrating the effect of both presented methods.

12

Automatyczny wybór reprezentatywnych fragmentów sygnałów do diagnostyki stanu narzędzia

Błażejak K., Bombiński S., Nejman M., Jemielniak K.

Mechanik

|

2016

|

R. 89, nr 10

1416--1417

PL

W systemach diagnostyki stanu narzędzia skrawającego istotnym problemem jest wybór fragmentów sygnałów, na podstawie których należy prowadzić diagnostykę. W ramach prac stworzono algorytm, który w pełni automatycz sposób ny wybiera fragmenty sygnału, które są reprezentatywne dla stanu narzędzia i pozwalają na prowadzenie obliczeń online.

EN

For the tool condition monitoring systems important issue is the choice of segments of signals on the basis of which the diagnostics should be carried out. This article presents algorithm for fully automatic selection parts of signals which are representative for the tool condition and allow to carry out the calculations online.

13

Monitoring of tool vibration for magnetorheological fluid controlled bar during turning of hardened AISI4340 steel

Paul P. S., Jazeel M., Varadarajan A. S.

Archive of Mechanical Engineering

|

2015

|

Vol. LXII, nr 2

237--255

EN

In recent times, the concept of hard turning has gained awareness in metal cutting as it can apparently replace the traditional process cycle of turning, heat treating, and finish grinding for assembly of hard, wear-resistant steel parts. The major apprehension in hard turning is the tool vibration, which affects the surface finish of the work piece, has to be controlled and monitored. In order to control tool vibration in metal cutting, a magnetorheological fluid damper which has received great attention in suppressing tool vibration was developed and used. Also an attempt has been made in this study to monitor tool vibration using the skewness and kurtosis parameters of acoustic emission (AE) signal for the tool holder with and without magnetorheological damper. Cutting experiments were conducted to arrive at a set of operating parameters that can offer better damping characteristics to minimize tool vibration during turning of AISI4340 steel of 46 HRC using hard metal insert with sculptured rake face. From the results, it was observed that the presence of magnetorheological damper during hard turning reduces tool vibration and there exist a strong relationship between tool vibration and acoustic emission (AERMS) signals to monitor tool condition. This work provides momentous understanding on the usage of magnetorheological damper and AE sensor to control and monitor the tool condition during turning of hardened AISI4340 steel.

PL

W ostatnich latach, w obróbce skrawaniem wzrasta zainteresowanie koncepcją toczenia na twardo, ponieważ może ono zastąpić tradycyjny proces toczenia, utwardzania i szlifowania stosowany przy wykonywaniu twardych, odpornych na zużycie cześci metalowych. Głównym problemem przy twardym toczeniu są wibracje narzędzia, które muszą być monitorowane i kontrolowane, gdyż wpływają na wykończenie powierzchni elementu obrabianego. W celu kontrolowania wibracji narzędzia przy obróbce skrawaniem autorzy zastosowali tłumik z płynem o właściwościach reologicznych sterowanych polem magnetycznym. Podjęto także próbę monitorowania wibracji na podstawie parametrów skośności i kurtozy sygnałów akustycznych (AE) emitowanych przez uchwyt narzędzia, mierzonych w warunkach bez tłumika i z tłumikiem magnetoreologicznym. Przeprowadzono szereg eksperymentów z toczeniem stali AISI4340 o twardości 46 HRC przy użyciu narzędzia z płytką z twardej stali, o geometrycznie kształtowanym ostrzu, firmy Taegu Tec. Otrzymano zbiór parametrów roboczych, wyznaczając na ich podstawie lepsze charakterystyki tłumienia i osiagając minimalizację wibracji narzędzia. Wyniki eksperymentów wskazują, że obecność tłumika magnetoreologicznego redukuje wibracje i że istnieje silna zależność miedzy wibracjami narzędzia i wartością skuteczną sygnału emisji akustycznych (AERMS). Praca przyczynia się do znacznie lepszego zrozumienia funkcji tłumika magnetoreologicznego i czujnika emisji akustycznych przy monitorowaniu stanu narzędzia przy toczeniu utwardzonej stali AISI4340.

14

Evaluation of sensor-based condition monitoring methods as in-process tool wear and breakage indices - case study: drilling

Tsanakas J. A., Botsaris P. N., Amirids I. G., Galeridis G. G.

Diagnostyka

|

2012

|

nr 2(62)

3-13

EN

Today, effective unmanned machining operations and automated manufacturing are unthinkable without tool condition monitoring (TCM). Undoubtedly, the implementation of an adaptable, reliable TCM and its successful employment in industry, emerge as major instigations over the recent years. In this work, a sensor-based approach was deployed for the in-process monitoring and detection of tool wear and breakage in drilling. In particular, four widely reported indirect methods for tool wear monitoring, i.e. vibration signals together with thermal signatures, spindle motor and feed motor current measurements were obtained during numerous drillings, under fixed conditions. The acquired raw data was, then, processed both statistically and in the frequency domain, in order to distinguish the meaningful information. The study of the latter is influential in identifying the trend of specific signals toward tool wear mechanism. The efficiency of this information as a tool wear and/or breakage index is the feature that determines the effectiveness and reliability of a potential indirect TCM approach based on a multisensor integration. The paper concludes with a discussion of both advantages and limitations of this effort, stressing the necessity to develop simple, fast condition monitoring methods which are, generally, less likely to fail.

15

Intelligent cutting tool condition monitoring in milling

Zuperl U., Cus F., Balic J.

Journal of Achievements in Materials and Manufacturing Engineering

|

2011

|

Vol. 49, nr 2

477--486

EN

Purpose: of this paper is to present a tool condition monitoring (TCM) system that can detect tool breakage in real time by using a combination of neural decision system, ANFIS tool wear estimator and machining error compensation module. Design/methodology/approach: The principal presumption was that the force signals contain the most useful information for determining the tool condition. Therefore, ANFIS method is used to extract the features of tool states from cutting force signals. The trained ANFIS model of tool wear is then merged with a neural network for identifying tool wear condition (fresh, worn). Findings: The overall machining error is predicted with very high accuracy by using the deflection module and a large percentage of it is eliminated through the proposed error compensation process. Research limitations/implications: This study also briefly presents a compensation method in milling in order to take into account tool deflection during cutting condition optimization or tool-path generation. The results indicate that surface errors due to tool deflections can be reduced by 65-78%. Practical implications: The fundamental limitation of research was to develop a single-sensor monitoring system, reliable as commercially available system, but much cheaper than multi-sensor approach. Originality/value: A neural network is used in TCM as a decision making system to discriminate different malfunction states from measured signals.

16

Real-time tool condition monitoring in milling by means of control charts for auto-correlated data

Colosimo B. M., Moroni G., Grasso M.

Journal of Machine Engineering

|

2010

|

Vol. 10, No. 2

5--17

EN

Real time monitoring of tool condirions and machining processes has been extensively studies in tne last decades, but a wide gap is stiil present between research activities and commercial tools. One of the factors which currently limit the utilization of these systems is the low flexibility of off-the-shelf solutions: in most cases they need dedicated off-line training sessions to acquire the reference patterns and thresholds, and/or the need for several input data to be defined a priori by a human operator. Instead of exploiting off-line learning sessions and a prior defined thresholds, this paper proposes an approach for automatic modelling of a cutting process and real-time monitoring of its stability that is based only on data acquired on-line during the process itself. This approach avoids any a-priori assumption about expected signal patterns, and it is characterized by an innovative implementation of well known Statistical Process Control techniques. In particular, with regard to milling processes, the paper proposes the utilization of cross-correlation coefficient between repeating signal profiles as the feature to be monitored, and an EWMA (Exponentially Weighted Moving Average) control chart for auto-correlated data as monitoring tool.

17

Tool wear monitoring based on wavelet transform of raw acoustic emission signal

Jemielniak K., Kossakowska J.

Advances in Manufacturing Science and Technology

|

2010

|

Vol. 34, nr 3

5-17

EN

The paper presents a new efficient method of evaluation of relevancy of signal features extracted from the wavelet coefficients of raw AE signal while rough turning of Inconel 625. Several meaningful signal features were automatically extracted from band pass signals using 22 different wavelets and used for tool condition monitoring. Accuracy of tool condition evaluation was employed as main criterion for selection of the most indicative wavelets and decomposition level of Discreet Wavelet Transform (DWT) and Wavelet Packet Transform (WPT).

PL

W artykule przedstawiono analizę przydatności miar sygnałów pasmowych uzyskanych za pomocą Transformaty Falkowej (WT). Zastosowano Dyskretną Transformatę Falkową (DWT) oraz Pakietową Transformatę Falkową (WPT). Do wykonania WT użyto 22 typy falek. Analizie poddano surowy sygnał emisji akustycznej rejestrowany podczas toczenia zgrubnego Inconelu 625. Automatycznie selekcjonowane miary sygnałów pasmowych użyto do monitorowania stanu narzędzia. Dokładność oszacowania zużycia ostrza na ich podstawie stanowiła kryterium oceny przydatności poszczególnych falek i transformaty falkowej.

18

Potentials and limitations of tool condition monitoring for MoS/sub 2/- and TiN-coated drills with acoustic emission in dry machining

Krimmel O., Tonnessen K., Zhong Wen Li

Prace Naukowe Instytutu Technologii Maszyn i Automatyzacji Politechniki Wrocławskiej. Konferencje

|

2000

|

Vol. 76, nr 35

138-149

EN

Tool condition monitoring is a means to increase the reliability and efficiency of machining through tool breakage detection and wear monitoring. This paper provides the relevant knowledge about the state of the art in monitoring drill condition with acoustic emission to support a successful choice of sensors for a monitoring system. In drilling acoustic emission (AE) sensors are successfully applied for breakage detection. Attempts have been made to monitor wear with acoustic emission. The theory for the effect of wear on acoustic emission is reviewed. A set of experiments has been evaluated in order to illustrate characteristics of the AE-signal in drilling. Experiments for 4 and 10 mm drills with MoS/sub 2/- and TiN-coating have been chosen to show the effect of the coating on acoustic emission. The soft MoS/sub 2/ coating provides a dry lubrication layer whereas the hard TiN coating protects the tool through wear resistance. Additionally, the feed force, which is the common signal used for drill wear monitoring, has been recorded during the experiments to illustrate the difference in the validity of the signals.

19

Laboratory versus industrial cutting force sensor in tool condition monitoring

Jemielniak K., Balazinski M.

Prace Naukowe Instytutu Technologii Maszyn i Automatyzacji Politechniki Wrocławskiej. Konferencje

|

2000

|

Vol. 76, nr 35

109-116

EN

Research works concerning the utilisation of cutting force measurements in tool condition monitoring usually present results and deliberations based on laboratory sensors. These sensors are too fragile to be used in industrial practice. Industrial sensors employed on the factory floor are less accurate, and this must be taken into account when creating a tool condition monitoring strategy. Another drawback of most of these works is that constant cutting parameters are used for the entire tool life. This does not reflect industrial practice where the same tool is used at different feeds and depths of cut in sequential passes. The paper presents a comparison of signals originating from laboratory and industrial cutting force sensors. The usability of the sensor output was studied during a laboratory simulation of industrial cutting conditions. Instead of building mathematical models for the correlation between tool wear and cutting force, an FFBP artificial neural network was used to find which combination of input data would provide an acceptable estimation of tool wear. The results obtained proved that cross talk between channels has an important influence on cutting force measurements, however this input configuration can be used for a tool condition monitoring system.