Integrating Feature Engineering with Deep Learning to Conduct Diagnostic and Predictive Analytics for Turbofan Engines

Table 1

Comparison between this research and past studies.

	Feature engineering	Model/regressor	RMSE/MAE
This study	SR, MARS, RF, XGB, PCA, SIR	PCA + RF	17.58/12.36
		PCA + CNN	17.6/12.45
Raid et al. [12]		MLR	19.76
		BPN	18.59
Le son et al. [13]	PCA	Wiener process modeling	28.6
Lim et al. [14]	PCA, KPCA, K means	DNN	15.6
		PCA + DNN	15.8
		KPCA + DNN	15.75
		K means + DNN	15.16
Sateesh Babu et al. [15]	CAE	DNN	37.56
		SVM	20.96
		RVR	23.8
		CNN	18.45
Mathew et al. [6]		KNN	30.79
		CART	28.48
		AB	28.82
		GB	27.45
		RF	24.95
		DNN	29.62
		SVM	48.17
Li et al. [16]	CAE	DNN	13.56
		RNN	13.44
		LSTM	13.52
		DCNN	12.61
Li et al. [17]		LSTM-CNN	11.96
Yurek et al. [11]	Chi-square, mutual information, correlation, Fisher’s score	MLR	34.67
		GB	16.89
		RF	13.63
Zhang et al. [9]	Step differential	CNN	13.59
		CNN-XGB	12.61
Deng et al. [8]	Long-term differential technique	Categorical boosting	15.8
		DNN	17.3
		CART	26.1
		RF	18.2
		AB	24.5
		GB	17.7
Li et al. [10]	CAE	Multiscale CNN	11.44
Remadna et al. [3]	CAE	RNN	19.51
		LSTM	19.68
		BDLSTM	18.59
		CNN-BDLSTM	10.74
Chen et al. [18]	Correlation, consistency	SVM	24.61
		LSTM	20.16
		SVM-LSTM	10.11
Kang et al. [19]	PCA	MLR	25.85
		PCA + MLR	23.62
		PCA + DNN	22.56
		PCA + RF	26.48
		PCA + SVM	28.28
Xiang et al. [20]		Multicellular LSTM	14.53
Li et al. [21]	Deep multiscale feature fusion network	CNN-GRU	12.18