Добавил:

ivanov666 Опубликованный материал нарушает ваши авторские права? Сообщите нам.

Вуз:

Башкирский Государственный Аграрный Университет

Предмет:

[НЕСОРТИРОВАННОЕ]

Файл:

патенты / Elevator Risk...ysis and SHAP_ZENG Qian-Xin...HuanYANG Yong

.pdf

Скачиваний:

Добавлен:

29.09.2024

Размер:

1.24 Mб

Скачать

☆

2024-09-25 16:07:02https://link.cnki.net/urlid/11.2854.TP.20240924.1626.011

ISSN 1003-3254, CODEN CSAOBN	E-mail: csa@iscas.ac.cn
Computer Systems & Applications [doi: 10.15888/j.cnki.csa.009685]	http://www.c-s-a.org.cn
© .	Tel: +86-10-62661041

SHAP

1, 1, 1, 2

1( , 528225)

2( , 528041)

: , E-mail: huan_yang2019@163.com

: ,. , ; ,, ,. , SHAP , . ,Cox , DeepSurv, .

: ; ; ; ; SHAP

: , , , . SHAP . . http://www.c-s-a.org.cn/1003-3254/9685.html

Elevator Risk Prediction Based on Deep Survival Analysis and SHAP

ZENG Qian-Xin1, WANG Pan1, YANG Huan1, YANG Yong2

1(School of Software, South China Normal University, Foshan 528225, China)

2(Foshan Branch, Guangdong Institute of Special Equipment Inspection and Research, Foshan 528041, China)

Abstract: This study proposes a comprehensive solution that combines deep survival analysis, data segmentation, and data imputation to address the issue of statistical predictive maintenance for elevators, which is characterized by low frequency and irregular time periods. This study establishes both dynamic and static survival vectors to capture factors influencing major fault risks. Additionally, to tackle left censoring in recorded data, this research employs data imputation and explores the impact of different imputation methods and segmentation strategies on the accuracy of deep survival models. Finally, this study utilizes SHAP to analyze deep survival models in elevators to reveal the dynamic influence of various factors on fault risks. The results indicate that a model combining rough data segmentation with Cox imputation demonstrates strong predictive capability and accuracy. The DeepSurv model excels in predictive capability and stability. The contribution of factors such as elevator age and lifting height to major fault risks can shift under specific conditions.

Key words: predictive maintenance; deep survival analysis; elevator; censored data; SHAP

	19 ,	,
.		. ,
, , ,		. ,
.

: (2023CT06); (2020A1515110783); 2020: 2024-04-21; : 2024-05-20, 2024-06-12; : 2024-06-17

				http://www.c-s-a.org.cn		** *
[1–3].					.
, [4]. ,
, . ,					,
,					. ,
,					. Yang [10]
,
.					,
					,
,					,
,					.
,					,
			Cox				[11]
	.	,				. Xiao
[5,6],					(random survival forest, RSF)
[7].					Cox (Cox proportional hazard model)
,					,
,					,
, ,					, . Chen
[8].					[12]
	,
( ).					, . [12] Cox
4 : (1) ,					,
; (2)					, 3 LSTM .
; (3) ,					Snider [13] ,
; (4) SHAP					Weibull ,
.					45 ,
: (1) ,					. ,
,					, ,
; (2) ,					,
; (3)					. [14]
. ,					Cox + ,
.
					.
1					1.2
1.1					,
	,				,
.					. Cox
,					Weibull Aalen .
. ,
: (1) (2)					.

[9]. , .

**** ** *

http://www.c-s-a.org.cn

. 3 : (1): ,. (2) :, ,. (3) :,.

, ,,. , .

率建模时, , CoxWeibull . ,, Cox . Cox, ,, . ,,, .

DeepSurv [15]. DeepSurv dropout., ℓ2 ,Adam , .

Cox-CC Cox-Time[16] DeepSurv

. t, CoxPH, CoxPH. DeepHit [17],., DeepHit, ., . Ishwarn[18] , .

, ,.

1.3 SHAP

SHAP (Shapley additive explanation)

.Shapley , , SHAP, .

SHAP ,,, SHAP . SHAP.

1 .

,,; 3 ,; , DeepSurv; ,SHAP .

. ,. TBF (time between failures) .,.

, 2. 1 ,;, .

(X; d; e) . ,

X ,; d ,, , ; e ,

	http://www.c-s-a.org.cn		** *
.		N , M
,		, " = ij	i 2 N; j 2 M ,
. ,		{ ,j	ij } i j
,		. eij′ i j′
:		.












			DeepSurv
	TBF=0		DeepSurv
	TBF=0
	TBF=Age
	TBF=Age

TBF=CoxPH

CoxPH

	CoxPH	TBF=
	CoxPH

SHAP

TBF11	TBF12	TBF13
ELE1
	(A)
	TBF15
TBF11	TBF13
ELE1
TBF12	TBF14	TBF16
	(B)

,, TBFi j . 3. TBFeij , .14 , , 14.

D = (X; T; E), XX′ X . T TBF, . E ,.

TBF11

ELE1

E11 τ11 E12 E13 τ12

TBF11

ELE2

E12

TBF11 TBF11

ELE3

τ13

.3 , ,

** *	http://www.c-s-a.org.cn
. 1, event11 event12		, .
, 1; event13		. ,
, 0;		TBF, d = Age.
			ˆ

. , ,

TBF11 , 4 .

(concordance index) (Brier score)

[19], ,

TBF11

[19].

. ,

TBF12

TBF13

ELE1

, .

ELE1

, . :

C-index = P{Sˆ (Ti jxi ) < Sˆ (T j xj

) Ti < T j ; Di = 1}

AUC , C-index=0.5 ,

. 3

, ; C-index=1

: 0

; C-index=0 ,

TBF: ,

. TBF:

, (Brier score)

, .

: Cox

. ,

, Cox .

, Cox ,

, ,

, , ,

0.5 , TBF,

0, ,

>Age .

. :

, Cox d

xi ) 1

∑ 6

xi )

1 Ti t; Di = 1

1 S

Ti > t

Brier score =

ˆ ( )

ˆ (T )

i=1

G Ti

(

G ) i

, G (Ti) Kaplan-Meier

, Brier score < 0:25 .

, 0.

http://www.c-s-a.org.cn

**** ** *

(integrated Brier score),

, :

integrated Brier score = 1 ∫ t2 BS (s) ds t2 t1 t1

4.1

.,, . ,,,.

3 ,,,, , ,, 14 .

6 324 , ,

1 300 , 2018–2023 .. 2.12.3 ,, TBF. 1 TBF.

1 TBF



	1611	6190
	60.1	83.85
	174.55	177.19
	1714.74	1714.73
(%)	11.92	7.09
(%)	21.97	76.69

DeepSurv

. ,

DeepSurv 4 , CoxCC

Logistic DeepHit (random survival

forest, RSF) .

DeepSurv CoxCC Logistic DeepHit

(multilayer perceptron, MLP)

, , .. 8:2, TBF. 5, C-index, .

PyTorch Pycox

Scikit-survival SHAP , 48 GB

NVIDIA RTX A6000 ., 2 3 .

SHAP , SHAP ,. ,. DeepSurv CoxCC

Cox ,; DeepHit Logistic, ; RSF. , DeepHit Logistic

, :,1,, .

, Logistic DeepHit, SHAP .

SHAP, , SHAP, . ,, SHAP, .

4.3

2 3 DeepSurv, CoxCC Logistic DeepHit RSF CoxPH. 4.

** *			http://www.c-s-a.org.cn
		2
	-TBF=0	-TBF=0	-TBF=Age	-TBF=Age	-TBF=Cox	-TBF=Cox

DeepSurv	0.729	0.824	0.680	0.738	0.723	0.786
CoxCC	0.731	0.821	0.695	0.724	0.707	0.773
Logistic	0.701	0.760	0.695	0.651	0.707	0.732
DeepHit	0.724	0.797	0.487	0.464	0.687	0.742
RSF	0.701	0.812	0.722	0.770	0.735	0.756
CoxPH	0.675	0.748	0.690	0.712	0.723	0.786

		3
	-TBF=0	-TBF=0	-TBF=Age	-TBF=Age	-TBF=Cox	-TBF=Cox
DeepSurv	0.071	0.173	0.105	0.215	0.082	0.150
CoxCC	0.068	0.165	0.100	0.205	0.088	0.154
Logistic	0.136	0.559	0.164	0.438	0.163	0.562
DeepHit	0.126	0.499	0.159	0.426	0.129	0.507
RSF	0.061	0.095	0.066	0.139	0.062	0.106
CoxPH	0.089	0.131	0.073	0.251	0.095	0.152

, ,,. TBF=0 ,Cox . TBF= Age , Cox . ,, Cox ,Cox ,.

,0.25,, , -TBF=0

CoxCC CoxPH, Logistic DeepHit

, 0.25. ,.

, TBF=0TBF=Cox, TBF=0 .TBF=Age . , 5(c)5(d) TBF=Age ,, 0.25,. TBF=Age, ,.

. ,, . ,, .

, , TBF=0, TBF=

Cox , DeepSurv CoxCC

RSF , ., ,; TBF=0 TBF=Cox, TBF=Age.

4.3.2SHAP

6 TBF=Cox. 6, TBF=Cox ,SHAP ., , . ,SHAP , , ,, .[20] , :,, .

				http://www.c-s-a.org.cn						** *
									0.4
0.3									0.3
				0.2					0.3
				0.2
0.2									0.2
				0.1					0.1
0.1									0.1
				0					0
0	500	1 000	1 500	0	1 000	2 000 3 000	4 000	5 000	0	500	1 000	1 500
		TBF ( )				TBF ( )					TBF ( )
	(a) -TBF=0				(c) -TBF=Age					(e) -TBF=Cox

0.8									0.8
				0.6					0.8
				0.6
0.6									0.6
0.4				0.4
				0.4
									0.4
0.2				0.2					0.2
0.2									0.2
0				0					0
0	500	1 000	1 500		0	2 000	4 000	6 000	0	500	1 000	1 500
		TBF ( )					TBF ( )				TBF ( )
	(b) -TBF=0					(d) -TBF=Age				(f) -TBF=Cox
			DeepSurv			CoxCC	Logistic	DeepHit	RSF
			5

Mean

Logistic

DeepSurv

DeepHit

CoxCC

RSF

SHAP

−2

−4

−6

(m)

(m/s)

TBF=Cox

SHAP

10 ,, SHAP . ,SHAP, ,. ,.

,,. , .

7 -Cox , DeepSurv(RSF) SHAP. ,, ., DeepSurv 0–2 ,SHAP 0 ; 5–10 ,SHAP 0 . 0–2, ,;5–10 ,. [21]“ ” ,, , ;, , ;, ,. RSF , 10, U ,5–12 ,.

**** ** *

http://www.c-s-a.org.cn

	4				DEEPSURV
					DEEPSURV
					RSF
	2
SHAP
	0
	0
	−2
	0	5	10	15	20

7 DeepSurv RSF SHAP

8 -Cox. ,, SHAP. RSF ,, . , 20 m, ,. , 20 m, RSF ,. DeepSurv ,,.

				DEEPSURV
	2			RSF
	2
SHAP	0
	0

	−2
	0	50	100	150
			(M)

8 DeepSurv RSF SHAP

,, ,. ,,, SHAP

. ,,;, ; , SHAP, . ,,; Cox; , DeepSurv . SHAP , ,. ,., .

1Niu DP, Guo L, Bi XL, et al. Preventive maintenance period decision for elevator parts based on multi-objective optimization method. Journal of Building Engineering, 2021, 44: 102984. [doi: 10.1016/j.jobe.2021.102984]

2Mishra KM, Huhtala KJ. Fault detection of elevator systems using multilayer perceptron neural network. Proceedings of the 24th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA). Zaragoza: IEEE, 2019. 904–909.

3Yao W, Jagota V, Kumar R, et al. Study and application of an elevator failure monitoring system based on the Internet of Things technology. Scientific Programming, 2022, 2022:

2517077.

4Theissler A, Pérez-Velázquez J, Kettelgerdes M, et al. Predictive maintenance enabled by machine learning: Use cases and challenges in the automotive industry. Reliability Engineering & System Safety, 2021, 215: 107864.

5Moncada-Torres A, van Maaren MC, Hendriks MP, et al. Explainable machine learning can outperform Cox regression predictions and provide insights in breast cancer survival. Scientific Reports, 2021, 11(1): 6968. [doi: 10.1038/s41598- 021-86327-7]

http://www.c-s-a.org.cn

**** ** *

6 Cheng B, Potter M. Bayesian weapon system reliability modeling with Cox-Weibull neural network. Proceedings of the 2023 Annual Reliability and Maintainability Symposium (RAMS). Orlando: IEEE, 2023. 1–6.

7 Yu H, Yang W, Wu S, et al. Deep-learning-based survival prediction of patients with cutaneous malignant melanoma. Frontiers in Medicine, 2023, 10: 1165865. [doi: 10.3389/ fmed.2023.1165865]

8Wiegrebe S, Kopper P, Sonabend R, et al. Deep learning for survival analysis: A review. Artificial Intelligence Review, 2024, 57(3): 65. [doi: 10.1007/s10462-023-10681-3]

9Civerchia F, Bocchino S, Salvadori C, et al. Industrial Internet of Things monitoring solution for advanced predictive maintenance applications. Journal of Industrial Information Integration, 2017, 7: 4–12. [doi: 10.1016/j.jii.

2017.02.003]

10Yang Z, Kanniainen J, Krogerus T, et al. Prognostic modeling of predictive maintenance with survival analysis for mobile work equipment. Scientific Reports, 2022, 12(1): 8529. [doi: 10.1038/s41598-022-12572-z]

11Xiao R, Zayed T, Meguid MA, et al. Improving failure modeling for gas transmission pipelines: A survival analysis and machine learning integrated approach. Reliability Engineering & System Safety, 2024, 241: 109672.

12Chen C, Liu Y, Sun XF, et al. Automobile maintenance prediction using deep learning with GIS data. Procedia CIRP, 2019, 81: 447–452. [doi: 10.1016/j.procir.2019.03.077]

13Snider B, McBean EA. Improving urban water security through pipe-break prediction models: Machine learning or survival analysis. Journal of Environmental Engineering, 2020, 146(3): 04019129. [doi: 10.1061/(ASCE)EE.1943-7870. 0001657]

14, . Cox .

, 2017, 39(1): 94–96. [doi: 10.3969/j.issn.10059679.2017.01.019]

15Katzman JL, Shaham U, Cloninger A, et al. DeepSurv: Personalized treatment recommender system using a Cox proportional hazards deep neural network. BMC Medical Research Methodology, 2018, 18(1): 24. [doi: 10.1186/ s12874-018-0482-1]

16Kvamme H, Borgan Ø, Scheel I. Time-to-event prediction with neural networks and Cox regression. Journal of Machine Learning Research, 2019, 20(129): 1–30.

17Lee C, Zame W, Yoon J, et al. DeepHit: A deep learning approach to survival analysis with competing risks. Proceedings of the 32nd AAAI Conference on Artificial Intelligence. New Orleans: AAAI, 2018. 2314–2321.

18Ishwaran H, Kogalur UB, Blackstone EH, et al. Random survival forests. The Annals of Applied Statistics, 2008, 2(3): 841–860.

19Turkson AJ, Ayiah-Mensah F, Nimoh V. Handling censoring and censored data in survival analysis: A standalone systematic literature review. International Journal of Mathematics and Mathematical Sciences, 2021, 2021:

9307475.

20McGough SF, Incerti D, Lyalina S, et al. Penalized regression for left-truncated and right-censored survival data. Statistics in Medicine, 2021, 40(25): 5487–5500. [doi: 10. 1002/sim.9136]

21Zhang XQ, Zubair MU. Extending the useful life of elevators through appropriate maintenance strategies. Journal of Building Engineering, 2022, 51: 104347. [doi: 10.1016/j.jobe. 2022.104347]

( : )

Соседние файлы в папке патенты

#
22.04.2024442.44 Кб03762.txt
#
22.04.2024435.37 Кб03763-1.txt
#
22.04.2024435.37 Кб03763.txt
#
22.04.2024496.68 Кб03764-1.txt
#
22.04.2024496.68 Кб03764.txt
#
29.09.20241.24 Mб0Elevator Risk...ysis and SHAP_ZENG Qian-Xin...HuanYANG Yong.pdf
#
22.04.202431.09 Кб0GB821790A.TXT
#
25.08.202368.23 Кб85Широкополосный переход от микрополосковой линии к волноводу.docx