A Remaining Useful Life Prognosis of Turbofan Engine Using Temporal and Spatial Feature Fusion

Abstract

The prognosis of the remaining useful life (RUL) of turbofan engine provides an important basis for predictive maintenance and remanufacturing, and plays a major role in reducing failure rate and maintenance costs. The main problem of traditional methods based on the single neural network of shallow machine learning is the RUL prognosis based on single feature extraction, and the prediction accuracy is generally not high, a method for predicting RUL based on the combination of one-dimensional convolutional neural networks with full convolutional layer (1-FCLCNN) and long short-term memory (LSTM) is proposed. In this method, LSTM and 1- FCLCNN are adopted to extract temporal and spatial features of FD001 andFD003 datasets generated by turbofan engine respectively. The fusion of these two kinds of features is for the input of the next convolutional neural networks (CNN) to obtain the target RUL. Compared with the currently popular RUL prediction models, the results show that the model proposed has higher prediction accuracy than other models in RUL prediction. The final evaluation index also shows the effectiveness and superiority of the model.

Shared Temporal Attention Transformer for Remaining Useful Lifetime Estimation

“…Raw sensor data is pre-processed to generate a 1D-health indicator matrix passed to a hybrid CNN-LSTM-NN RUL predictor in [21]. The extracted spatial and temporal features from CNN and LSTM networks respectively are fused and passed to another CNN layer in [22] for predicting RULs of C-MAPSS sub-datasets FD001 and FD003. Another hybrid network of parallel CNN and LSTM paths is proposed in [23], to reduce the influence of CNN extracted features on series-connected LSTMs.…”

“…17, it is to observe that the computation time of DAG [23] is the least since it consists of fewer trainable parameters. However, the performance score from the proposed models is sufficiently higher [12] 231 3366 251 2840 CNN [16] 1287 13570 1596 7886 CNN+Attention [18] 198 1144 251 2072 HI CNN-LSTM-NN [21] 303 3440 1420 4630 FCLCNN [22] 204 than all the models presented in the literature. Furthermore, although the STAT model contains almost the same number of parameters as in AGCNN [25], and FeaR-STAT consists of virtually the same as in DCNN [17], both the proposed models complete training faster and performs better than the compared literature.…”

Towards Interpretable Deep Learning: A Feature Selection Framework for Prognostics and Health Management Using Deep Neural Networks

“…The second case study used in this work corresponds to the NASA Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) datasets [ 77 ], widely used as benchmark for RUL prediction methods [ 29 , 30 , 31 , 78 , 79 , 80 , 81 ]. They are four datasets containing simulated data of turbofans degradation time series.…”

“…As mentioned before, the C-MAPSS datasets are used for RUL prediction. Typically, best results are obtained when using CNNs and/or LSTMs [ 31 , 79 , 80 , 81 ]. Here, the data are organized into time windows for the model to learn to associate a certain evolution of feature values to a RUL value.…”

پیش‌بینی روند زوال و عمر مفید باقی‌مانده ی یاتاقان غلتشی با کمک شبکه ی عصبی بازگشتی حافظه ی طولانی کوتاه مدت

“…}a |xDUy; OWQ w |Q}opmW VN@ 'sr=U w |O=a xO=iDU= =@ 'u}= Q@ xwqa [3] "Ovm |R=UpOt R}v = Q Qta |v}@V}B |xrUt |=yC}a]k |=yVwQ R= |Qo}O |xDUO |@Ya |xm@W xrtH R= |awvYt Vwy |=yVwQ R= CU= TQDUO QO u=k=D=} l} |xDWPo R= xm |}=yxO=O uOQ@ Q=m x@ =@ 'QwLtxO=O |=ypOt uDW=O x@ |R=}v CQwY u}= QO w OQm |R=UpOt = Q p=wR |=ywor= u=wD|t R= xO=O |QU OvJ =} l} \UwD VRwt; x@ =OD@= QO =yxm@W u}= "Ow@ Oy=wNv |[=} Q 'xm@W VRwt; R= TB &OvOvtR=}v u}W=t u=ty |r@k |xJN}Q=D =} x@=Wt |=yu}W=t u=WxOv=t}k=@ Qta =D OwW|t xO=iDU= x@=Wt |=yu=k=D=} |wQ |R=UxO=}B |= Q@ u; R= [5w3] "OwW |v}@V}B R= |Q=}U@ \UwD 1 QwNW}B |@Ya |xm@W 'G}= Q |@Ya |=yxm@W u}@ QO xOW xDiQo Q=m x@ |WDre |=yu=k=D=} xOv=t}k=@ O}it Qta |v}@V}B |= Q@ u=kkLt |xYNWt |QU l} =D CU= R=}v =OD@= QO Q= R@= u}= R= xO=iDU= Qw_vt x@ [3] [6] u= Q=mty w Q}Om "OW=@ QwLtxO=O VwQ 'QwNW}B |@Ya |xm@W |=y|OwQw u=wvax@ C=W=aDQ= |xJN}Q=D |y=oW}=tR; |xO=O |QU l} |wQ = Q u=k=D=} |xOv=t |k=@ O}it Qta |v}@V}B "OvOQm |R=UxO=}B QFwt Q=Okt '?U=vt |W=aDQ= |=yxYNWt |x}yD Qw_vt x@ [5] u= Q=mty w O= Ry@ |OwQw u=wvax@ w xOQm l}miD u}}=B w q=@ |Uv=mQi |=yxR=@ QO = Q |W=aDQ= p=vo}U Ow@y@ ?Hwt Q=m u}= xm OvO=O u=Wv =yv; &OvDiQo Q_v QO QwNV}B |@Ya |xm@W "CU= xOW u}W}B |=ypOt x@ C@Uv CkO Qta '|@Ya |xm@W w uw}UQoQ |xrUt l} ? }mQD =@ R}v [7] u= Q=mty w u=wJ=}S |=yxYNWt =OD@= =yu; "OvOQm |v}@V}B |@ wN CkO =@ = Q =yu=k=D=} |xOv=t}k=@ O}it TBU w xOQm VR= Q@ |[=} Q |x]@= Q l} =@ = Q p=vo}U QFwt Q=Okt w |orwJ |W=aDQ= = Q =yu=k=D=} Qta QwNW}B |@Ya |xm@W |OwQw u=wvax@ =yu; R= xO=iDU= =@ =@ xm |DvU |@Ya |=yxm@W |=yO= Q}= R= O=kDv= =@ [8] u= Q=mty w nvB "OvOR u}tND |xm@W l} ? }mQD =@ 'OvR=OQB|t xYNWt l} G= QNDU= x@ , =iQY G}=Q |=yVwQ x@ 5 COtx=Dwm |vqw] |x_i=L |@Ya |xm@W l} w 4 |WJ}B j}ta |@Ya "OvDi=} CUO xOv=t |k=@ Qta |v}@V}B QO |QDy@ CkO |vqw] |x_i=L |@Ya |xm@W R= xO=iDU= Qw_vt x@ [9] u= Q=mty w |=S u} QDy@ w OvOQm G= QNDU= |W=aDQ= |=yp=vo}U R= = Q |O=}R |=yxYNWt 'COtx=Dwm OvO=O xm@W x@ |OwQw u=wvax@ 'CW=O u=k=D=} p=wR =@ |QDW}@ |oDU@ty xm = Q =yu; x@ 7 s= Q; u=}O= Qo C} wkD w 6 u=@}DWB Q=OQ@ Qou}tND pOt wO x@ C@Uv C}=yv QO w "OvDi=} CUO |QDy@ CkO R= = Q j}ta |=yxYNWt |WJ}B |@Ya |xm@W ltm =@ [10] u= Q=mty w w=t |x_i=L |@Ya |xm@W |OwQw u=wvax@ w xOQm G= QNDU= s=N |W=aDQ= p=vo}U |xrLQt QO \ki Qta |v}@V}B 'xr=kt u}= QO "OvOQm xO=iDU= COtx=Dwm |vqw] "CU= xOW |R=UxO=}B 'CU= QwyWt 8 `} QU p=wR x@ xm u=k=D=} p=wR |}=yv |vqw] |x_i=L |@Ya |xm@W ?wN OQmrta x@ xHwD =@ 'Q}N= |=yp=U QO hrDNt |=yxRwL u}kkLt '|v=tR |=y|QU w =yxO=O |v}@V}B Qt= QO COtx=Dwm Ov=CQ=@a =yu; R= |NQ@ xm Ovvm|t xO=iDU= |irDNt |=yxv}tR QO xm@W u}= R= |=yQ= R=@ QO Ct}k OvwQ |v}@V}B [11] '= Rv=wriv; Ovv=t |}=y|Q=t}@ OvwQ |v}@V}B %R= [13] "Civ |=yx=J O}rwD OvwQ |v}@V}B w [12] [20] "u=k=D=} |oO=Di= Q=m R= w ?…”