Model development based on evolutionary framework for condition monitoring of a lathe machine

Garg, Akhil; Vijayaraghavan, Vineeth; Tai, Kang; Singru, Pravin M.; Jain, Vishal; Krishnakumar, Nikilesh

doi:10.1016/j.measurement.2015.04.025

Cited by 19 publications

(9 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The validation and performance of the LR, MLP, RBF, and ANFIS models were compared using R 2 , VAF, and RMSE as follows:

R^{2} = 1 - true(\frac{\sum_{i = 1}^{n} {false(z - z^{'} false)}^{2}}{\sum_{i = 1}^{n} z_{i}^{2}} true)

VAF = true(1 - \frac{var (z - z^{'})}{var (z)} true) \times 100

RMSE = \sqrt{\frac{1}{N} \sum_{i = 1}^{N} {(z - z^{'})}^{2}}

where z is the measured value, z' is the predicted value, and N is the total number of observations . R 2 is a descriptive measure between zero and one and indicates the ability of a parameter to predict another parameter.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Prediction of output energies for broiler production using linear regression, ANN (MLP, RBF), and ANFIS models

Amid

Gundoshmian

2016

Env Prog and Sustain Energy

View full text Add to dashboard Cite

In this study, artificial neural networks and fuzzy inference systems and the combination of these two are employed to develop predictive models to input and output parameters for broiler production. The data was randomly collected from 70 broiler farms in northwestern Iran. The energy used in broiler production was determined to be fuel, feed, and electricity; these were selected as input parameters for the models. The corresponding output energies (broilers and manure) were used as output variables. Linear regression, the multi‐layer perceptron and the radial basis function as the methods of artificial neural networks, and the adaptive neuro‐fuzzy inference system were simulated and compared for prediction of broiler and manure energy. A comparison of the results showed that radial basis function recorded higher coefficient of determination (0.996 and 0.995 for broiler and manure) and lower root mean square error (0.016 and 0.02 for broiler and manure) values to make it the best predictor of outputs. This was followed in order by the adaptive neuro‐fuzzy inference system, linear regression, and multi‐layer perceptron models. © 2016 American Institute of Chemical Engineers Environ Prog, 36: 577–585, 2017

show abstract

“…The validation and performance of the LR, MLP, RBF, and ANFIS models were compared using R 2 , VAF, and RMSE as follows:

R^{2} = 1 - true(\frac{\sum_{i = 1}^{n} {false(z - z^{'} false)}^{2}}{\sum_{i = 1}^{n} z_{i}^{2}} true)

VAF = true(1 - \frac{var (z - z^{'})}{var (z)} true) \times 100

RMSE = \sqrt{\frac{1}{N} \sum_{i = 1}^{N} {(z - z^{'})}^{2}}

Section: Methodsmentioning

confidence: 99%

“…where z is the measured value, z' is the predicted value, and N is the total number of observations [29,30]. R 2 is a descriptive measure between zero and one and indicates the ability of a parameter to predict another parameter.…”

Section: Anfis Modelmentioning

confidence: 99%

Prediction of output energies for broiler production using linear regression, ANN (MLP, RBF), and ANFIS models

Amid

Gundoshmian

2016

Env Prog and Sustain Energy

View full text Add to dashboard Cite

show abstract

“…It can boost statistical and machine learning model performance, despite not being compulsory for deep learning models given these can extract new representative features that fit the problem automatically. The most common techniques can be grouped into next groups: feature extraction as statistical features in time [200] and frequency [55,200,210] domains that extract time/frequency relations of features; based on projection to new space like principal component analysis [26,45] which reduce dimensionality while keeping relevant information; concatenation and fusion methods [87] create new features by combining available ones; feature selection [155] reduces dimensionality discarding features of low variance, redundant and uncorrelated to target, given these increase complexity while not supplying additional information.…”

Section: 22mentioning

confidence: 99%

Deep learning models for predictive maintenance: a survey, comparison, challenges and prospect

Serradilla,

Zugasti,

Zurutuza

2020

Preprint

View full text Add to dashboard Cite

Given the growing amount of industrial data spaces worldwide, deep learning solutions have become popular for predictive maintenance, which monitor assets to optimise maintenance tasks. Choosing the most suitable architecture for each use-case is complex given the number of examples found in literature. This work aims at facilitating this task by reviewing state-of-the-art deep learning architectures, and how they integrate with predictive maintenance stages to meet industrial companies' requirements (i.e. anomaly detection, root cause analysis, remaining useful life estimation). They are categorised and compared in industrial applications, explaining how to fill their gaps. Finally, open challenges and future research paths are presented.

show abstract

“…If the amplitude reaches rough zone, preventive maintenance should be done. The artificial neural network and evolutionary algorithm are used to flag the user that the bearing needs to be changed [34,35].…”

Section: Failure Prediction From Fft Graphsmentioning

confidence: 99%

Bearing failure prediction using Wigner-Ville distribution, modified Poincare mapping and fast Fourier transform

Singru¹,

Krishnakumar²,

Natarajan³

et al. 2018

J. vibroeng.

View full text Add to dashboard Cite

This study outlines the experimental investigation methods of condition monitoring to predict bearing failures using the experimental vibration signatures. The purpose of condition monitoring is to maximize the machine availability and utility of the machine components. Bearings being one of the most common component in any rotating machinery, it is vital to study the health of the bearing and can predict bearing failure, its location and severity. This prevents machine downtime, monetary loss and unfortunate accidents. A test rig was fabricated to get the vibration signatures of bearings. Prediction of bearing failure relies on the presence of the bearing characteristic frequencies-inner race frequency, outer race frequency, ball pass frequency and fundamental train frequency-and its harmonics in the vibration signal acquired. These frequencies are present in the vibration signature due to the interaction of surfaces of different bearing components that have defects in them. Both time and frequency domain numerical signature analysis were performed on the vibration signatures acquired. Simple frequency domain method like Fast Fourier Transform (FFT), chaotic vibration method like modified Poincare mapping and time-frequency domain Wigner-Ville distribution (WVD) were used in detecting bearing failure. Using the FFT analysis method, it is hard to predict the failures, hence better signal processing methods like modified Poincare mapping and WVD are used. Also, it is observed that the chaotic vibration signatures found in the lower-order mechanical systems like bearings. With the chaotic analysis methods like, Poincare Mapping and Wigner-Ville Distribution, the location and the severity of the bearing failure can be predicted.

show abstract

Model development based on evolutionary framework for condition monitoring of a lathe machine

Cited by 19 publications

References 36 publications

Prediction of output energies for broiler production using linear regression, ANN (MLP, RBF), and ANFIS models

Prediction of output energies for broiler production using linear regression, ANN (MLP, RBF), and ANFIS models

Deep learning models for predictive maintenance: a survey, comparison, challenges and prospect

Bearing failure prediction using Wigner-Ville distribution, modified Poincare mapping and fast Fourier transform

Contact Info

Product

Resources

About