TY  - EJOU
AU  - Jiang, Weixin 
AU  - Yuan, Qing 
AU  - Li, Zongze 
AU  - Gong, Junhua 
AU  - Bo	Yu, 

TI  - Efficient	and	Robust	Temperature	Field	Simulation	of	Long-Distance	 Crude	Oil	Pipeline	Based	on	Bayesian	Neural	Network	and	PDE
T2  - The International Conference on Computational \& Experimental Engineering and Sciences

PY  - 2023
VL  - 25
IS  - 4
SN  - 1933-2815

AB  - The	hydraulic	and	thermal	simulation	of	crude	oil	pipeline	transportation	is	greatly	significant	for	the	safe	
transportation	and	accurate	regulation	of	pipelines.	With	reasonable	basic	parameters,	the	solution	of	the	
traditional	partial	differential	equation	(PDE)	for	the	axial	soil	temperature	field	on	the	pipeline	can	obtain	
accurate	simulation	results,	yet	it	brings	about	a	low	calculation	efficiency	problem.	In	order	to	overcome	
the	low-efficiency	problem,	an	efficient	and	robust	hybrid	solution	model	for	soil	temperature	field	coupling	
with	Bayesian	 neural	 network	and	 PDE	is	 proposed,	which	 considers	 the	 dynamic	 changes	 of	 boundary	
conditions.	 Four	 models,	 including	 the	 proposed	 hybrid	 model,	 PDE,	 and	 two	 kinds	 of	neural	 networks	
(NNs),	are	adopted	to	predict	the	single	soil	temperature	field	and	transient	oil	temperature	at	the	outlet	of	
the	pipeline.	It	is	found	that	the	prediction	results	of	the	hybrid	model	are	closer	to	those	of	PDE	than	those	
of	 NNs,	 and	 the	 hybrid	 model	 can	 obtain	 more	 stable	 and	 reliable	 prediction	 results	 of	 transient	 oil	
temperature	than	NNs	on	the	basic	training	data	sets	with	different	quantities	and	boundary	conditions.	On	
the	other	hand,	the	calculation	time	of	the	hybrid	method	is	far	less	than	that	of	PDE,	and	the	hybrid	model	
is	applicable	 for	 the	situation	of	small	 training	data	 to	improve	 the	simulation	efficiency.	This	study	can	
provide	a	scientific	reference	for	the	fast	real-time	simulation	of	crude	oil	pipelines.
KW  - Crude oil	pipeline	transportation; numerical	simulation; soil	temperature	field; Bayesian	neural	network; fast	prediction

DO  - 10.32604/icces.2023.08861