V. Dhanush Abhijit ¹, V. Sugumaran ¹, K. I. Ramachandran ²

Affiliations
1 School of Mechanical and Building Science, SMBS, VIT University, Chennai - 600237, Tamil Nadu, IN
2 CEN, Amrita School of Engineering, Ettimadai, Coimbatore – 600127, Tamil Nadu, IN

Abstract

Objectives: Being widely used in most of the industrial machineries, bearings are subjected to wear and tear. Failure of bearings can incur heavy losses in the industries. In order to prevent such mishaps during operation, it is necessary to subject the bearings to a suitable fault diagnosis technique. Methods/Statistical Analysis: Vibration analysis is performed to detect the fault in bearings. For the fault analysis, vibration signals were taken for good, inner race defect, outer race defect and combination of these defects. Since vibration signals are complex and the defect related signature is buried deep within the noise and high frequency resonance, simple signal processing cannot be used for effectively detecting bearing fault. In this paper, discrete wavelets transform were used to detect bearing faults. For wavelet and feature selection, J48 decision tree algorithm was used. For feature classification, Best First Tree (BFT) algorithm was used. Findings: The experimental results indicate biorthogonal wavelets show maximum successful bearing fault detection rate. The classification accuracy was calculated and found to be 96.25%. This result is further refined to get better classification accuracy and the final result was found to be 98%. Application/Improvements: This can be considered to be a part of a preventive maintenance method in order to avoid mishaps in industries. The classification accuracy can be further improved using different algorithms.

Keywords

Biorthogonal, Decision Tree, Fault Diagnosis, Feature Selection, Vibration Signals, Wavelet Selection, Wavelet Transforms.

Full Text

   

Ayush Kimothi ¹, Ameet Singh ¹, V. Sugumaran ¹, M. Amarnath ²

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai Campus, Chennai - 600127, Tamil Nadu, IN
2 Indian Institute of Information Technology Design and Manufacturing Jabalpur, Jabalpur – 482005, Madhya Pradesh, IN

Abstract

Objective: Gearbox being the backbone of transmission system is designed and manufactured very carefully so that there is minimum compliance in the system. However, there are still faults and failures which usually occur in the system. The failure in helical gearbox is more prominent in bearings rather than gears which are the main components of the system. Gearbox is susceptible to failures because of reasons like misalignment, vibration and shocks. In this paper wavelet feature extraction is used along with random forest algorithm to diagnose faults in gearbox. The vibration signals were used for extracting wavelet features. Features were selected using J48 Decision Tree and were classified using random forest algorithm. A detailed study has been done to ensure that the optimum number of features was used and the factor was iterated so that maximum classification accuracy is obtained. The results are presented along with the conclusion. Method Analysis: The classification accuracy is obtained by 3 steps namely, feature extraction, feature selection and feature classification. By obtaining the Decision Tree the most important factors are selected to obtain maximum classification accuracy at minimum number of features to reduce calculations in real time application. The number of features and depth of data is iterated to obtain the maximum classification accuracy. Findings: Through this research random forest algorithm was tested for fault diagnosis of gearbox and a better classification accuracy was obtained. These results can be further used for fault diagnosis in industries for any gearbox related problems. Application/Improvements: An extensive investigation is done by a random forest algorithm which produced better forecasting than the other algorithms. Based on the overall study, random forest was found as the most preferred classification algorithm that achieved the best classification accuracy of 93.08% which is better than the other algorithms.

Keywords

Fault Diagnostics, Gearbox Fault Diagnostics, J48 Decision Tree, Machine Learning, Random Forest, Vibration Signals, Wavelet Feature Extraction

Full Text

   

S. Aravinth ¹, R. Jegadeeshwaran ¹, V. Sugumaran ¹

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai Campus, Chennai - 600127, Tamil Nadu, IN

Abstract

Objectives: The brake system is one of the major components used in automobiles which inhibits motion by absorbing energy from a moving system. So regular monitoring is essential in brake system which ensures not only vehicle safety but also human lives. Methods/Statistical Analysis: In this study, a vibration based fault diagnosis approach has been reported through machine learning approach. A hydraulic brake setup was fabricated and vibration signals under various fault conditions were extracted using accelerometer sensor with suitable frequency. These signals were compared with good range of signals and variation is analyzed through histogram feature extraction, selection and classification of machine learing scenario. Findings: Histogram features were extracted by separation of signals into different bin ranges among which bin with highest accuracy level is further processed through selection process of Decision Tree and 87.78% was the achieved accuracy in fault determination. In Voting Feature Interval (VFI) 85.64% was the accuracy attained in error identification. Application/Improvements: Since Decision Tree gives the better result in fault identification in brake fault diagnosis of this study, it can be further improved by varying the frequency ranges of signals, so each and every variation in signals are noted. Moreover improvement in accuracy level can also be achieved in future by increasing number of samples percondtion of brake system.

Keywords

Decision Tree, Histogram Features, Machine Learning, Vibration Signals, Voting Feature Interval

Full Text

   

Abhijit D. Gaikwad ¹, V. Sugumaran ¹, M. Amarnath ²

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai Campus, Chennai - 600127, Tamil Nadu, IN
2 IIITDM Jabalpur, Jabalpur - 482005 , Madhya Pradesh, IN

Abstract

Objectives: Use of an appropriate fault diagnosis methods alerts in advance about malfunctioning and failure of bearings. Vibration and Sound signals of rotating machines contain the dynamic information about their operating conditions. There are many articles reporting suitability of vibration signals for fault diagnosis applications; however, the transducers ( accelerometers) and data acquisition equipment used for vibration signals analysis are costly. This prevents small scale industries and low cost equipment from using diagnostic tools on affordability ground. On the other hand, transducers used for acquiring sound signals (microphones) are relatively low cost or/and affordable. Hence, there is a need for studying the use of sound signal for fault diagnosis applications. This paper uses sound signals acquired from roller bearings in good and simulated faulty conditions for the fault diagnosis purpose. Methods/Analysis: Sound signals from bearings having defects on inner race and outer race have been considered for analysis. Since the characteristic sound signals of faulty bearings are complex and are struck in the noise and high frequency structural resonance, simple signal processing techniques cannot be used to detect bearing fault. Hence, wavelet features are used for extracting features from sound signals. The energy levels at various levels of wavelet decomposition are used to define features from sound signals. The most contributing features were selected and their classification is done using decision tree algorithm. This paper also discusses the effect of features, effect of various classifier parameters on classification accuracy. Findings: In feature classification of the fault signals the RBIO 2.4 wavelet has given the highest classification accuracy of 96.66%. Out of the 120 total instances, 116 (96.66%) were correctly identified while 4 instances were incorrectly classified with an error margin of (3.33%). Application/Improvements: An extensive investigation has been made by a J48 algorithm which produced better predictive performance than the other algorithms. The training and the optimization of J48 model with their essential parametric measures are reported. Based on the overall study, J48 with variation in number of objects (from 1 to 6) feature was found as the most successful classification algorithm that achieved the best classification accuracy of 96.66%. The classification accuracy of the proposed algorithm has been found better with only 4 misclassified features. The classification capability and the performance evaluation of J48 algorithm with confusion matrix and detailed classification accuracy is reported and discussed for further study.

Keywords

Bearings, Classification Accuracy, Decision Tree, Fault Diagnosis, Feature Selection, Sound Signals, Wavelet Features.

Full Text

   

Shantanu Gupta ¹, Siju K. Abraham ¹, V. Sugumaran ¹, M. Amarnath ²

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai – 600127, Tamil Nadu, IN
2 Indian Institute of Information Technology Design and Manufacturing Jabalpur, Jabalpur - 482005, Madhya Pradesh, IN

Abstract

Objectives: Heart of the transmission system in most machineries are gears for efficient power transmission. Even minor faults in gear can lead to major losses in terms of energy as well as in terms of money. The unwanted by-product while operating gear box are vibration and acoustic signals, which can be used for the condition monitoring and fault diagnosis of the gearbox. This study proposes the usage of machine learning algorithm for condition monitoring of a helical gearbox by using the acoustic signals produced by the gearbox. Methods/Analysis: The acoustic signals were captured using microphone from a gearbox with artificially created fault conditions. A comprehensive study was carried out using different discrete wavelet transformations for feature extraction which was further used in generating J48 decision tree algorithm and subsequently it was employed for selection and classification of the extracted features. Finding: Through this study the classification accuracy obtained is 97.619% by varying the different parameter to achieve the highest accuracy level. Data used in this study is exclusively obtained through experiment and subsequently through J48 decision tree and random tree classification accuracy level is studied to accomplish the highest accuracy. Novelty/Improvements: The comparison of different discrete wavelet transforms of the acoustic signals proved Daubechies 5 Discrete Wavelet Transform is the best suited one to use. The methodology yielded a satisfactory classification accuracy of 97.619%, which is higher than what was obtained by similar experiments with different methodology till date. The results and their analysis is discussed in the study. The performance of this methodology may be further improved by using different classifiers.

Keywords

Acoustic Signals, Gearbox, J48 Decision Tree, Random Tree, Wavelets

Full Text

   

Nirmal Mangal ¹, M. Ramesh Nachiappan ¹, M. Elangovan ², V. Sugumaran ¹

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai - 600127, Tamil Nadu, IN
2 Department of Mechanical Engineering, Amrita School of Engineering, Coimbatore. Amrita Vishwa Vidyapeetham, Amrita University, Coimbatore – 641112, Tamil Nadu, IN

Abstract

Objective: Tool condition monitoring is an important aspect of the modern day manufacturing system. It plays a significant role in increasing the efficiency of machining operation by identifying defects at a very early stage. Tool wear decreases the life of the tool considerably, increases the length of the machining process, also affects the surface finish and the dimensional accuracy of the product. To identify whether the tool is in a good or faulty condition, a monitoring system is essential. Method/Analysis: The fault diagnosis of the single point cutting tool was accomplished with the vibration signals obtained from auniaxial accelerometer attached to the cutting tool in a lathe machine. In this study, three different spindle speeds, feed rates and depth of cuts and four different wear levels of cutting tool are considered. Statistical data obtained from the signals is classified using a decision tree algorithm to get substantial features. The recognized features are considered in classifying data by using Simple CART classifier. Findings: The accuracy of the classifier was found to be 73.38% for the model with all the signals combined. The classification accuracy was observed to improve with the reduction in complexity of the model. The classification accuracy obtained for the model with only varying feed rate and depth of cut was in the range of 81–87 %. On further reduction of the model to have varying depth of cut was found to have a classification accuracy in the range of 81.5–91 %. The model with all the parameters independent yielded classification accuracy in the range of 81–100 %. Applications/Improvements: This study broadly analysed the use of simple CART classifier to diagnose fault in the cutting tool during machining. It can be used to increase productivity and reduce machine downtime. The improvements can be made to this study by considering different feature extraction techniques for more reliability.

Keywords

Decision Tree, Feature Extraction, Simple CART, Statistical Features, Tool Condition Monitoring

Full Text

   

Cynthia Deb ¹, M. Ramesh Nachiappan ¹, M. Elangovan ², V. Sugumaran ¹

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai - 600127, Tamil Nadu, IN
2 Department of Mechanical Engineering, Amrita School of Engineering, Coimbatore. Amrita Vishwa Vidyapeetham, Amrita University, Coimbatore – 641112, Tamil Nadu, IN

Abstract

Objectives: There is a wide range of methods implemented for tool condition monitoring in the erstwhile manufacturing industry to ensure that the process continues uninterruptedly with minimal supervision. This monitoring method reduces the overall maintenance cost of machinery and prevents the occurrence of failure by prediction. This prior detection of tool wear, in turn, reduces the machine downtime and enhances machining efficiency. The progressive wear of a cutting tool can be detrimental to the quality of the machined surface, tolerances, dimensional accuracy and also adversely change the work or tool geometry. So the requirement of a diagnosing system with consistency is vital. Method/Analysis: This study deals with acquiring vibrational signals using accelerometer during turning operation performed on a lathe machine with good and fault simulated single point cutting tool. From the acquired signals, certain statistical features such as standard deviation, kurtosis etc. were extracted and substantial features were recognised using a decision tree algorithm. Those recognised features were deliberated in classifying data using random forest classifier. Findings: The accuracy of classification by the random forest classifier for all the signals combined together yields 74.4%. When considering feed rate and depth of cut as varying parameters yields an accuracy around 84%. Further an accuracy of around 88% was observed when considering depth of cut as varying parameter. When considering every experiment as a separate model yields around 95% classification accuracy. Applications/Improvements: This research work analysed the utilization of random forest classifier to identify the tool wear. It can be used in identifying the tool wear which affects surface finish, dimensional accuracy and tolerance of the part during machining. This work can be improved by analysing with different classifier algorithms to efficiently predict the tool wear.

Keywords

Confusion Matrix, Decision Tree, Feature Extraction, Random Forest, Statistical Features, Tool Condition Monitoring

Full Text

   

Nikhil Pawar ¹, V. Sugumaran ¹, Ameet Singh ¹, M. Amarnath ²

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai - 600127, Tamil Nadu, IN
2 Department of Mechanical Engineering, IIITDM, Jabalpur - 482005, Madhya Pradesh, IN

Abstract

Objectives: Gears are machine elements that transmit motion by successively engaging teeth. In technical terms, gears are used to transmit motion. Fault in gears can lead to major problems which may end up in affecting the gear’s functionality. Hence, fault diagnosis at an initial stage is of utmost importance to reduce losses that might occur. Continuous monitoring of the gears is very necessary. Vibration signals recorded for good and faulty conditions are used for fault detection in the helical gearbox. The fault diagnosis is done using feature extraction, feature selection and feature classification. Firstly, feature extraction was carried out using MATLAB software. Feature selection was done using J48 classifier. The classification accuracies for different conditions were calculated and compared by using K-Star classifier and the results obtained were very promising. Methods/Analysis: Vibration signals were obtained from the experimental set up of the helical gearbox. The recorded signals were then used for feature extraction using MATLAB through different wavelet features. The total number of signals extracted was 448 with each class consisting of 64 signals. The families of wavelets taken into account for fault diagnosis were Haar, Discrete Mayer, Daubechies, Biorthogonal, Reverse Biorthogonal, Coiflet and Symlets. In wavelet selection, signals were split into different frequency components and each component was studied with a resolution matched to its scale. J48 classifier was used to carry out the feature selection process and decision tree was obtained for Sym 8 wavelet. The best combination of nodes was visualized and further feature classification was done on these nodes. By varying the global blends the optimum number of objects was selected to obtain the highest classification accuracy. Finding: The classification accuracy for the built model was 91.74%. The data extracted from the vibration signal is used for the classification purpose. This maximum classification accuracy was obtained with K star algorithm. Novelty/Improvements: Wavelet selection was different from Fourier methods in analyzing physical situations where the signal contains discontinuities and sharp spikes. K Star algorithm was used to carry out the fault diagnosis.

Keywords

Decision Tree, Gearbox Fault Diagnosis, J48 Classifier, K-Star Classifier, Wavelet

Full Text

   

Abhi Srivastava ¹, Ameet Singh ¹, V. Sugumaran ¹, M. Amarnath ²

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai - 600127, Tamil Nadu, IN
2 Department of Mechanical Engineering, IITDM, Jabalpur - 482005, Madhya Pradesh, IN

Abstract

Objectives: Gearbox, being an important component in the mechanism of many industrial machines can have a few faults mostly by fatigue cracking under cyclic contact stressing. Most of the implements presently being utilized in the industries for the gearbox fault diagnosis are dependent upon the vibration signals which are accumulated from the gearbox. Methods: A machine based learning approach has been utilized for the detection of faults with the utilization of vibration signals that have been acquired from helical gearbox setup. The features were extracted from the collected vibration signals using wavelets. The significant features were selected using a Decision Tree algorithm. The selected features from this approach were then classified using random tree algorithm and higher accuracy was achieved. Findings: The random tree algorithm used for the classification of the wavelets which were extracted from the vibration signals of the gearbox resulted in a classification accuracy of 90.4%. This classification accuracy is unique in terms of the vibration signals that have been acquired utilizing the accelerometer from the helical gearbox setup. The higher classification is achieved after feature extraction, selection and classification. Improvements/Applications: The classification accuracy achieved using the random tree algorithm was higher than the previously attained values for the gearbox. The higher accuracy would result in better fault diagnosis for the helical gearbox setup.

Keywords

Decision Tree, Fault Diagnosis, Helical Gearbox, Machine Learning, Random Tree Algorithm, Wavelet Features.

Full Text

   

Siju K. Abraham ¹, V. Sugumaran ¹, M. Amarnath ²

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai - 600127, Tamil Nadu, IN
2 Indian Institute of Information Technology Design and Manufacturing Jabalpur, Jabalpur - 482005, Madhya Pradesh, IN

Abstract

Objectives: Most machineries employ gears for efficient power transmission. Even minor faults with the gear box can lead to severe losses both in terms of energy and money. The vibration and acoustic signals from the gear box, which usually are said to be as an unwanted by-product of the operation, can be used for the condition monitoring and fault diagnosis of the gearbox. This study proposes the usage of machine learning algorithm for condition monitoring of a helical gearbox by using the sound signals produced by the gearbox. Methods/Analysis: The acoustic signals were captured using microphone from a gearbox with artificially created fault conditions. An exhaustive study using different discrete wavelet transformations for feature extraction from the acoustic signals was carried out and subsequently J48 Decision Tree algorithm was employed for selection and classification of the extracted features. Findings: The time domain acoustic signals were converted into frequency time domain data using different discrete wavelet transforms. Of all the wavelet ransforms, the Daubechies 5 Discrete Wavelet Transform was found to be the best suited for the current scenario. The methodology yielded a satisfactory classification accuracy of 97.6% when classified using J48 algorithm. Novelty/ Improvements: The classification accuracy yielded through this methodology is higher than what was obtained by similar experiments with different methodologies till date. The results and their analysis is discussed in the study. The whole methodology when put in a real time sytem will have the capability to monitor the condition and diagnose the faults in the gearbox quickly and effectively. The performance of this methodology may be further improved by using different classifier algorithms.

Keywords

Acoustic Signal, Condition Monitoring, Decision Tree Classifier, Gearbox, Wavelets.

Full Text

   

Ameet Singh ¹, V. Sugumaran ¹, M. Amarnath ²

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai - 600127, Tamil Nadu, IN
2 Indian Institute of Information Technology Design and Manufacturing Jabalpur, Jabalpur – 482005, Madhya Pradesh, IN

Abstract

Objectives: Gear plays an efficient role in power transmission. Minor faults in gears can lead to severe faults. The vibration analysis can be used for determining the causes of the faults which are raised while ongoing operation. This study determines the usage of machine learning algorithm for condition monitoring of helical gearbox. Methods/Statistical Analysis: The vibration signals were taken by using accelerometers from helical gearbox in which artificial faults were incorporated before testing. By using Discrete Wavelet Transform (DWT) feature extraction was done. The feature selection and feature classification was done by using J48 algorithm and subsequent results were observed. Findings: The classification accuracy of helical gearbox using Discrete Wavelet Transform was observed to be 89.28% which itself shows its efficiency. In feature extraction maximum accuracy of 89.06% was obtained by sym 8 wavelet. During feature selection and classification many modifications in algorithm were made i.e. minimum number of object, confidence factor etc. Suitable readings of the modifications were applied and feature classification was done. Improvements: Different Discrete Wavelet Transforms were compared taken from vibration signal proved Sym 8 Discrete Wavelet Transform is the best one to be used in this scenario. The methodology yielded a satisfactory classification accuracy of 89.28%, which is higher than what was obtained by similar experiments with different methodology till date. The results and their analysis are discussed in the study. The performance of this methodology may be further improved by using different classifiers and different wavelets.

Keywords

Condition Monitoring, Discrete Wavelet Transform, J48 Algorithm, Vibration Signals.

Full Text

   

R. Satishkumar ¹, V. Sugumaran ¹

Affiliations
1 School of Mechanical and Building Sciences, VIT University, Chennai Campus, Chennai - 600127, IN

Abstract

Objective: This paper proposes a predictive model to assess the health condition of bearing using classification technique. Method: In the present study, vibration signals were acquired on a daily basis until the bearing is damaged. Initially, feature selection was done with decision tree and predictive model was built using selected features. Now, Random forest classifier was used to build the model to assess the remaining lifetime of the bearing. Distinct data were used to validate the performance of the classifier. Findings: The classification accuracy of the built model was found to be 95.64%. Applications: The proposed model was tested with the data acquired from a bearing experimental set-up wherein run-to-failure test were conducted on bearings at rated load conditions.

Keywords

Bearings, Life Time Assessment, Random Forest Classifier

Full Text

   

R. Satishkumar ¹, V. Sugumaran ¹

Affiliations
1 School of Mechanical & Building Sciences, VIT University, Chennai- Campus, Vandalur - Kelambakkam Road, Chennai - 600 127, Tamil Nadu, IN

Abstract

The overall performance metric of rotating machineries are governed by the reliability of bearings. Bearings are vital components for all moving parts. It has its presence in most of the equipments and machineries. Also, these bearings contribute to most of the failures or breakdowns in an industry. Failures can be reduced to a greater extent by selecting appropriate bearings that suit to the application. Nevertheless, after selection of right bearings, the failure in the bearings tops the list. It becomes complicated when we want to trace out the reasons for failures. Condition monitoring techniques are being deployed in order to increase the uptime of the machineries. Objectives: Strengthening the predictive maintenance by estimating the remaing useful life of bearings. Method: This paper proposes a predictive model to address the remaining life of the bearing that suits to a real time application. This method is validated on a laboratory experiment wherein the bearing is tested till it fails naturally at stated conditions. Findings: Thus obtained results show the model built using Support Vector Regression method proves to be effective in predicting the remaining life of the bearings. Applications/Improvements: The proposed predictive model is validated with the new set of data taken from experiments. This model can be deployed in critical real time applications where the bearing failure affects the performance of the machine. Addittionaly this model can be horizontally deployed for other critical components where continuous monitoring is essential.

Keywords

Remaining Useful Life (RUL), Statistical Methods, Support Vector Regression (SVR)

Full Text

   

B. Rebecca Jeya Vadhanam ¹, S. Mohan ², V. Sugumaran ³

Affiliations
1 Research and Development Centre, Bharathiar University, Coimbatore - 641046, Tamil Nadu, IN
2 CCIS, AL Yamamah University, Riyadh, Kingdom of Saudi Arabia
3 SMBS, VIT University, Chennai Campus, Chennai – 600127, Tamil Nadu, IN

Abstract

Objectives: In this present study, there are various methods and techniques that are reviewed to dig the hidden information from the video frames to process the live stream Television (TV) videos. Video classification is an emerging trend that is intended to classify the Advertisement (ADD) videos from the television programme. Classification of ADD videos from the general programs provides an efficient approach to manage and utilize the ADD video data. Detection of ADD video plays a major role for advertisement content management, advertisement for targeted customers, querying, retrieving, inserting, and skipping the advertisement to view the desired channels. Detection of advertisement frames creates a unique application in the multimedia systems. Methods/Analysis: The process of feature extraction which enables recognition of ADD videos and Non Advertisement (NADD) videos directly from the TV streams are discussed. The features are extracted using Block Intensity Comparison Code (BICC) technique. BICC technique is applied on various block sizes of a frame and the best performing block size 8×8 has been chosen for the experimental study. Decision tree (J48) algorithm and BICC feature are utilized to find out the promising block size of the frame. The best features are identified and selected by decision tree (J48) algorithm. Artificial Immune Recognition System (AIRS) is applied on these features to classify the ADD class and NADD class. The AIRS classification algorithms are motivated by the biological immune system components that include important and unique abilities. These algorithms recreate the specialities of the immune framework like; discrimination, learning, and the memorizing methodology in place are utilized to classification and pattern recognition. AIRS2 algorithm is parallelism, separating the dataset into number segments and handling them exclusively. Findings: In this study, three versions of AIRS algorithms, namely, AIRS1, AIRS2 and AIRS2 parallel are used for classification with BICC feature. AIRS2 parallel classifier performed better compared with AIRS1 and AIRS2. The present study proved the biological immune recognition based AIRS algorithm out performs than various classifiers in terms of reliability and classification accuracy. The classification capability and the efficiency of AIRS2 parallel algorithm with BICC feature has been compared among various classifiers and reported. Application/Improvements: This study is very much helpful and essential for television viewers and the busy current generation to skip the nuisance of advertisements to enjoy watching their favourable shows of various television channels. The proposed work is useful for demands on video and video content management systems. This work can also be extended with novel feature set to improve the classifiers performance for efficient video classification and retrieval systems.

Keywords

Television Live Stream (TV), Advertisement Frames (ADD), Non Advertisement Frames (NADD), Block Intensity Comparison Code (BICC), Decision Tree, Artificial Immune Recognition System (AIRS) Classification

Full Text

   

V. V. Ramalingam ¹, S. Mohan ², V. Sugumaran ³

Affiliations
1 Bharathiar University, Coimbatore - 641 046, Tamil Nadu, IN
2 AL Yamamah University, Riyadh, Kingdom of Saudi Arabia, SA
3 VIT University, Chennai Campus, Chennai - 600048, Tamil Nadu, IN

Abstract

Objectives: In monitoring brain activities, Electroencephalogram (EEG) signals play a significant role. As brain activities are many and highly dynamic in nature, processing of EEG signals is a challenging task. Since classification is more accurate when the pattern is simplified through representation by well performing features, feature extraction and selection play an important role in classification systems such as Clonal Selection Classification Algorithm (CSCA) algorithm. Methods/Analysis: This study is one such attempt to perform the prosthetic limb movements using EEG signals. In this research, the performance of CSCA for prosthetic limb movements of EEG signals has been reported. Findings: In this paper, the EEG signals are acquired for four different limb movements like finger open (fopen), finger close (fclose), wrist clockwise (wcw) and wrist counterclock wise (wccw). These EEG signals can be used to build a model to control the prosthetic limb movements using CSCA algorithm. The statistical parameters were extracted from the EEG signals. The best feature set was identified using J48 decision tree classifier. The well performing features were then classified using CSCA algorithm. The classification performance of CSCA has been reported. Novelty/Improvement: Our work is useful for controlling artificial limb with movements using EEG signals. The signal processing of EEG signals is a complex task and requires sophisticated techniques to yield a better classification accuracy.

Keywords

CSCA, Classification, Electroencephalogram (EEG) Signals, Statistical Features

Full Text

   

A. Joshuva ¹, V. Sugumaran ², M. Amarnath ³, Sang-Kwon Lee ⁴

Affiliations
1 School of Mechanical and Building Sciences (SMBS), VIT University, Chennai Campus, Vandalur Kelambakkam Road, Chennai – 600127, Tamil Nadu, IN
2 School of Mechanical and Building Sciences (SMBS), VIT University, Chennai Campus, Vandalur Kelambakkam Road, Chennai – 600127, Tamil Nadu
3 Department of Mechanical Engineering, Indian Institute of Information Technology Design and Manufacturing,Airport Road, IIITDM Jabalpur Campus, Khamaria, Jabalpur – 482005, Madhya Pradesh, IN
4 Department of Mechanical Engineering, Inha University, KR

Abstract

Objectives: The main objective of this study is to develop a model which can able to predict the remaining life time working of a gearbox using vibration signals. Method: This study is considered as a machine learning problem which consists of three phases, namely feature extraction, feature selection and feature classification. In this research, histogram features are extracted from vibration signals, feature selection are carried out using J48 algorithm and different regression models were built to predict the reaming lifetime assessment of a gearbox. Findings: In this study, the J48 algorithm was used and the regression was found to be 0.8944 for Gaussian model. This is a novel approach to finding the life prediction of gearbox using histogram and regression model. Improvements: This algorithm is applicable for real-time analysis and further the condition monitoring can be carried out using different algorithms with less computation time.

Keywords

Assessment, Fault Diagnosis, Gearbox, Histogram Features, Life Time, Multiple Regression, Sound Signals.

Full Text

   

G. Balamurugamohanraj ¹, K. Vijaiyendiran ¹, P. Mohanaraman ¹, V. Sugumaran ²

Affiliations
1 Department of Mechanical Engineering, Sri Manakula Vinayagar Engineering College,Mannadipet Commune, Madagadipet, Puducherry – 605107, Tamil Nadu, IN
2 School of Mechanical and Building Sciences (SMBS), Vellore Institute of Technology (VIT) University Chennai Campus, Vandalur - Kelambakkam Road, Chennai – 600127, Tamil Nadu, IN

Abstract

Background/Objectives: Modern manufacturing industries aim to increase production rate with less production cost and high quality. To achieve high production rate with minimum cost, machining parameters must be optimized in industries. Therefore, the main objective of this study is to establish the relationship between the influence of cutting parameters and surface roughness in dry boring operation. Methods/Statistical analysis: A full factorial design was used to evaluate the effect of four independent variables (feed rate, spindle speed, depth of cut, tool flank wear). Stainless steel 304 was selected as a work piece due to high hardness, chemical stability and various applications. Carbide tipped tool insert was used for machining. Findings: During machining, Statistical features were extracted from the vibration signal. The extracted statistical features, machining condition and tool flank wear were considered to establish the various surface roughness prediction models. Multilayer perceptron and decision tree models were developed to predict the surface roughness. From these two models, the best suitable prediction model was selected based on maximum correlation coefficient and minimum ischolar_main mean squared error values. Application/Improvements: The selected best model can be used for variety of machining conditions to predict the surface roughness of the machined surface.

Keywords

Cutting Parameters, Decision Tree, Flank Wear, Linear Regression, Vibration Signals.

Full Text

   

Nikhil Pawar ¹, V. Sugumaran ¹, Ameet Singh ¹, M. Amarnath ²

Affiliations
1 School of Mechanical and building Science (SMBS), VIT University, Chennai Campus, Chennai – 600127, Tamil Nadu, IN
2 Department of Mechanical Engineering, Indian Institute of Information Technology Design and Manufacturing, Airport Rd, Jabalpur Campus, Khamaria, Jabalpur – 482005, Madhya Pradesh, IN

Abstract

Objectives: In this paper, machine learning approach, grounded on vibrations, has been used for helical gear box and holds a vital position in the industry. This approach has three steps namely feature extraction, feature selection and feature classification. Firstly, feature extraction was carried out using Matrix Laboratory (MATLAB) software. Feature selection was done using J48 classifier. The nodes with highest classification accuracy were further tested using J48 graft classifier and the results obtained were very promising. Methods/Analysis: Vibration signals were obtained from the experimental set up of the helical gear box. The recorded signals were then used for feature extraction using MATLAB through different wavelet features. The total numbers of signals extracted were 448 with each class consisting of 64 signals. The families of wavelets taken into account for fault diagnosis were Haar, Discrete Mayer, Daubechies, Biorthogonal, Reverse Biorthogonal, Coiflet and Symlets (SYM). In wavelet selection, signals were dissected into various frequencies and each was analyzed with appropriate resolution.J48 classifier was used to carry out the feature selection process and decision tree was obtained for Sym 8 wavelet. The best combination of nodes was visualized and further feature classification was done on these nodes. Findings: Feature classification was carried out by J48 graft algorithm. Using the grafting technique, the classifier achieved the highest accuracy for pruned data for 10 times cross validation. It gave maximum accuracy for pruned data (40%) and the results were satisfactory. Novelty/Improvements: The J48 graft algorithm uses grafting to infer from previous decision trees. This helps in reducing prediction errors.

Keywords

Decision Tree, Gearbox Fault Diagnosis, J48 Classifier, J48 graft Classifier, Wavelet.

Full Text

   

V. V. Ramalingam ¹, S. Mohan ², B. Rebecca Jeya Vadhanam ³, V. Sugumaran ⁴

Affiliations
1 Department of Computer Science and Engineering, S. R. M University, Kattankulathur – 603203, Tamil Nadu, IN
2 CCIS, AL Yamamah University, Riyadh, Kingdom of Saudi Arabia
3 Department of Computer Applications, Faculty of Science and Humanities, S. R. M University, Kattankulathur – 603203, Tamil Nadu, IN
4 School of Mechanical and Building Sciences (SMBS), VIT University, Chennai Campus, Chennai-600127, Tamil Nadu,, IN

Abstract

Objectives: This paper presented a novel approach with a set of Auto Regressive Moving Average (ARMA) features for the best classification of different hand moments in Electroencephalogram (EEG) signals using C4.5 Decision tree algorithm. Methods/Analysis: The characteristics of EEG signals can be represented through the best features is the most prominent and significant role in the classification systems. The classification is more flawless when the specimen is streamlined through the feature extraction and feature selection process. Findings: In this study, there are four kinds of EEG signals recorded from strong volunteers with finger open, finger close, wrist clockwise and wrist counterclockwise. The well performing statistical features are acquired from the EEG signals. C4.5 Decision tree classifier is used to identify the changes in the EEG signals. The yield of the classifier confirmed that the proposed C4.5 Decision tree classifier has potential to classify the EEG signals of the specific hand movements. Improvement: The proposed work is contributed to manage the right hand movements through the EEG signals. The efficient techniques are required to process the complex EEG signals to achieve the better classification result. To improve the classification accuracy, an efficient feature extraction technique may be applied.

Keywords

ARMA Features, C4.5 Decision Tree, Classification, Electroencephalogram (EEG) Signals.

Full Text

   

M. Ramesh Nachiappan ¹, V. Sugumaran ¹, M. Elangovan ²

Affiliations
1 School of Mechanical and Building Sciences (SMBS), VIT University, Chennai Campus, Chennai - 600127, Tamil Nadu, IN
2 Department of Mechanical Engineering, Amrita School of Engineering, Amrita Vishwa Vidyapeetham,Amrita University, Coimbatore – 641112, Tamil Nadu, IN

Abstract

Objective: A variety of tool condition monitoring techniques in modern manufacturing system plays a key role in estimating the tool wear which can save the machine downtime and increase the cutting tool utilization. Tool wear compromises dimensional accuracy and affects the precision, tolerance and surface finish. An active condition monitoring system of tool health is required for superior productivity. Method/Analysis: In this experimental study, the accelerometer was used to acquire the vibration signal during the turning operation in a lathe machine with good and fault simulated single point cutting tool. The signals are acquired for all combinations of spindle speeds, feed rates, depth of cuts and tool wear level. In this study, 3 different spindle speeds, feed rates and depth of cuts, and 4 different tool wear levels were considered. Statistical features were extracted from the acquired signal and substantial features were recognized using a decision tree algorithm. The identified substantial statistical features were considered in classifying signals using logistic model tree classifier. Findings: The classification accuracy obtained for all the signals combined (i.e., variable spindle speeds, feed rates, depth of cuts and tool wear levels) were found to be 74.27%. The classification accuracy achieved was improved through simplifying the model by considering feed rate and depth of cut as variable factor. The accuracy of the classification was found to be in the range of 82-86%. Further, the classification accuracy was found to increase to the range of 82-93%, when considering the depth of cut alone as variable factor. Application/Improvement: The utilization of logistic model tree to identify the tool wear level in a single point cutting tool during turning operation was comprehensively analysed in this study. The performance of the classifier on fault diagnosis of single point cutting tool and its improvement by reducing the complexity of the model was discussed.

Keywords

Decision Tree, Feature Extraction, Logistic Model Tree, Logistic Regression, Statistical Features Tool Condition Monitoring.

Full Text

   

P. S. Anoop ¹, V. Sugumaran ¹, Hemanth Mithun Praveen ¹

Affiliations
1 SMBS, VIT University Chennai campus, Chennai - 600127, Tamil Nadu, IN

Abstract

Objectives: Tyre pressure monitoring systems are automotive electronic systems used to monitor the automobile tyre pressure. The existing systems use pressure sensors or wheel speed sensors. They depend on batteries and radio transmitters which would add up to cost and complexity. Methods/Analysis: This paper proposes a new machine learning approach to monitor the tyre pressure. Vertical vibrations are extracted from a wheel hub of a moving vehicle using an accelerometer and are classified using machine learning techniques. The statistical features are extracted from the vibration signal and the features are classified using K Star algorithm. Findings: A reasonably high classification accuracy of 89.16% was obtained. Application/Improvements: The proposed model can be used for monitoring the automobile tyre pressure successfully.

Keywords

Automobile, K Star Algorithm, Machine Learning, Statistical Features, Tyre Pressure Monitoring System.

Full Text

   

S. Aravinth ¹, K. Rajesh Kanna ², V. Sugumaran ²

Affiliations
1 Bannari Amman Institute of Technology, Sathyamangalam - 638401, Tamil Nadu, IN
2 School of Mechanical and Building Sciences (SMBS), VIT University, Chennai Campus, Chennai - 600127, Tamil Nadu, IN

Abstract

Objectives: The fault diagnosis in reciprocating air compressor system was done through this article using vibration signals from accelerometer for both healthy and faulty conditions. Methods/Analysis: This article presents a condition monitoring strategy for compressor through vibration signals using accelerometer data in identifying five common faults of air compressor these were simulated manually. These vibration signals were processed through machine learning technique, where statistical features were extracted and the features contributing to the maximum classification accuracy were selected. The J48 decision tree algorithm is used in predicting the compressor faults in early stages. Findings: High classification accuracy of 98.33% was obtained for fault detection in compressor system. Application/Improvements: The proposed model can be used for regular monitoring of air compressor.

Keywords

Fault Diagnosis, J48 Algorithms, Reciprocating Air Compressor, Vibration Signals.

Full Text

   

Yasir Khan ¹, Siju Abraham ², V. Sugumaran ³, M. Amarnath ⁴

Affiliations
1 School of Mechanical and Building Sciences, Vellore Institute of Technology, Chennai Campus,Chennai - 600127, IN
2 1School of Mechanical and Building Sciences, Vellore Institute of Technology, Chennai Campus, Chennai - 600127, IN
3 School of Mechanical and Building Sciences, Vellore Institute of Technology, Chennai Campus, Chennai - 600127,, IN
4 Indian Institute of Information Technology Design and Manufacturing Jabalpur, Jabalpur - 482005, Madhya Pradesh, IN

Abstract

Objectives: This study aims at devising a methodology for accurately predicting the different fault conditions of gears in a gearbox using acoustic signals. Statistical Analysis: The acoustic signals are captured for several artificially created fault conditions of different magnitude and the wavelet features are extricated from captured acoustic signals. Subsequently,prominent features are selected by utilizing J48 Decision tree which discerns the most dominant traits among the allocated data obtained from wavelet transform of the acoustic signals followed by Random Forest for the classification of features. Findings: Out of a total of eleven features extracted, six were selected through Decision Tree and Random forest was used for feature classification of acoustic signals using wavelet features. Several iterations were conducted on the wavelet features by varying different parameters and the maximum percentage accuracy was found to be 99.76%. The instances of misclassification of features were minimal in Random Forest and it proved to be an efficient and precise classifier. Hence, Random Forest proved to be an easy to use, fast and accurate classifier that could classify various kinds of wavelet features efficiently. Applications: The methodology can be used to provide accurate real time results about the condition of gear teeth.

Keywords

Acoustic Signals, Decision Tree, Fault Diagnostics, Gear Box, Wavelets

Full Text

   

A. Joshuva ¹, V. Sugumaran ¹

Affiliations
1 School of Mechanical and Building Sciences (SMBS), VIT University, Chennai Campus, Vandalur-Kelambakkam Road,Chennai – 600127, Tamil Nadu, IN

Abstract

Objectives: The main objective of this research is to develop a model which can able to predict the various blade faults occurs in the wind turbine blade while the turbine in operating condition using vibration signals. Method: This study is considered as a machine learning problem which consist of three phases, namely feature extraction, feature selection and feature classification. In this research, statistical features are extracted from vibration signals, feature selection are carried out using J48 algorithm and different parameters of J48 algorithm were optimized to build a better classifier. Findings: In this study, the J48 algorithm was used and the classification accuracy was found to be 85.33% for multiclass problem. This is a novel approach of finding the different problem occurs in wind turbine blade at once. Improvements: This algorithm is applicable for real-time analysis and further the condition monitoring can be made as a portable device with less computation time.

Keywords

Fault Diagnosis, J48 Algorithm, Statistical Feature, Structural Health Monitoring, Vibration Signal, Wind Turbine Blade.

Full Text

   

B. Rebecca Jeya Vadhanam ¹, S. Mohan ², V.V. Ramalingam ³, V. Sugumaran ⁴

Affiliations
1 Research and Development Centre, Bharathiar University, Coimbatore − 641 046, Tamil Nadu, IN
2 CCIS, AL Yamamah University, Riyadh, Kingdom of Saudi Arabia, SA
3 Department of Computer Science and Engineering, Faculty of Engineering and Technology, S.R.M University, Kattankulathur − 603203, Tamil Nadu, IN
4 VIT University, Chennai Campus, Vandalur − 600127, Kelambakkam Road, Chennai, IN

Abstract

Background/Objectives: The main objective of the present study is to do the prerequisite process to develop a viewerfriendly electronic embedded system and business beneficial system to promote their products. This can be achieved by classifying the extracted Advertisement (ADD) videos from the Non-Advertisement (NADD) videos which consists of more visual information. Methods/ Statistical Analysis: The proposed frame work facilitates to identify the advertisement and non advertisement videos from the live stream television videos are discussed. The Block Intensity Comparison Code (BICC) technique is applied to extract the essential features from the ADD and NADD video frames. The frames are divided into various block sizes to select the best performing block size of the frame. The 8x8 frame size has been chosen as the promising block size to conduct the experiments. An extensive experimental analysis has been demonstrated with different classifier and a comparative study also reported. Findings: Decision tree algorithm (C4.5) has been employed to identify the vibrant features and these features are taken as the input to the various decision tree algorithms, namely J48, J48graft, LM tree, Random tree, BF tree, REP tree and NB tree to classify the video genre. A broad investigation has been made by a random tree algorithm which produced better predictive performance than the other algorithms. The training and the optimization of random tree model with their essential parametric measures are reported. Based on the overall study, random tree with BICC feature was found as the most preferred classification algorithm that achieved the 92.08% than the other algorithms. The classification capability and the performance evaluation of random tree algorithm with block intensity comparison code is reported and discussed for further study. Application/Improvements: The performance of the classifier can also be improved with other novel features.

Keywords

Advertisement (ADD) Videos, Block Intensity Comparison Code (BICC) Features, Classification, Non- Advertisement (NADD) Videos.

Full Text

   

R. Jegadeeshwaran ¹, V. Sugumaran ¹

Affiliations
1 School of Mechanical and Building Science, VIT University Chennai Campus, Chennai - 600127,Tamil Nadu, IN

Abstract

Background/Objectives: To study the recent development for monitoring the condition of a hydraulic brake system using statistical learning approaches. Methods/Statistical Analysis: Machine fault diagnosis is one of the condition monitoring approaches used to monitor the condition of machinery. For brake fault diagnosis, many conventional techniques have been reported in literature. In recent days, statistical learning approaches like, naïve bayes, decision tree, bayes net, best first tree, support vector machines, K Star have been successfully used for the fault diagnosis study. Findings: Keeping in mind the end goal to distinguish the most plausible deficiencies prompting to disappointment, numerous strategies in particular, like thermal image mapping, oil particle analysis, acoustic emission signal analysis, vibration analysis have been used for analyzing the data. Among these, vibration signal has been conveniently used for many fault diagnosis study. The same vibration signal can be used for the brake fault diagnosis study. Then these vibration data are processed using shortterm Fourier transform, high-resolution spectral analysis, waveform analysis, wavelet analysis, wavelet transform, etc. The results of such analysis are used to analyze the causes of failures. Recent advancement is the application of statistical approach for analyzing the data. This study presents a brief review about the possibilities for implementing the recent statistical learning approaches for monitoring the condition of the brake system. Application/Improvements: Number of new statistical learning approaches like nested dichotomy, clonal selection classification algorithm, Artificial Immune Recognition System (AIRS) algorithm can be used for the brake fault diagnosis study.

Keywords

Brake System, Condition Monitoring, Fault Diagnosis, Statistical Learning, Vibration Signal.

Full Text