Skip to main content

Table 1 The categorization of artificial intelligence-based methods

From: A robust principal component analysis-based approach for detection of a stator inter-turn fault in induction motors

Method based

Main weaknesses

Fuzzy logic [22]

1. Necessity to large accessible data

2. Unpromising efficiency in inaccurate inputs

3. complete dependence on expertise and human intelligence

Expert systems [23]

1. Insufficient accuracy

2. Complexity

3. Low flexibility

Evolutionary algorithms [24]

1. Require long time

2. Unsuitable for online fault diagnosis

Artificial neural network [25]

Conventional neural networks

RBF [26]

1. Inability to self-learning without feature extraction

2. Necessity to large accessible data

MLP [26]

 

SVM [27]

1. Necessity to large accessible data

2. Sluggish Performance due to large data requirement

 

k-NN [28]

1. Necessity to large accessible data

2. Slow performance in real time

3. Outlier-sensitivity

 

ANN [29]

1. Necessity to large accessible data

2. Overfitting

Modern neural networks

Decision trees and random forest [30]

1. Necessity to large accessible data

2. The longest training period

2. Overfitting

 

Native Bayes [31]

1. Necessity to large accessible data

2. Suitable only for independent features

 

Deep learning

CNN [32]

1. large data requirement

2. Need long time for training

 

RNN [33]

 

AE [34]

 

LSTM [35]