CN-121296307-B - Digital twin-based marine diesel engine gas circuit control method

CN121296307BCN 121296307 BCN121296307 BCN 121296307BCN-121296307-B

Abstract

The invention relates to the technical field of gas circuit control of marine diesel engines, in particular to a digital twin-based gas circuit control method of marine diesel engines, which adopts the technical scheme that the problem that a traditional static model cannot adapt to the variable working conditions of ships is solved through a diesel engine multi-dimensional twin body, a gas flow dynamic sub-model based on an LSTM neural network and an in-cylinder combustion mechanism sub-model based on a thermodynamic law are fused to realize full-working-condition coverage, and a fault response closed loop is realized through a failure feature library based on historical fault data; aiming at the problem that the sensor is invalid due to the severe environment of the ship, the twin body is driven by the virtual-real interaction layer to carry out multi-parameter coupling simulation, the data reliability is improved, and the dynamic optimization module and the fault-tolerant control module are adopted to carry out double optimization based on the simulation result, so that the limitation of fixed weight coefficients is broken through, and the multi-objective dynamic optimization is realized.

Inventors

YU HAIYANG
Tang Jifei
TU SHANDONG

Assignees

华东理工大学
艾瑞斯计算机科技(苏州)有限公司

Dates

Publication Date: 20260512
Application Date: 20251029

Claims (8)

1. The marine diesel engine gas circuit control method based on digital twinning is characterized by comprising the following operation steps: S1, constructing a diesel engine multi-dimensional twin body, and fusing a gas flow dynamic sub-model based on an LSTM neural network, an in-cylinder combustion mechanism sub-model based on a thermodynamic law and a failure feature library based on historical failure data; s2, collecting sensor group data in real time, and driving a twin body to perform multi-parameter coupling simulation through a virtual-real interaction layer; S3, performing double optimization based on a simulation result, generating VGT/EGR reference control quantity by a dynamic optimization module with the aim of maximizing fuel efficiency, comparing output difference of a physical layer and a twin body by a fault-tolerant control module, and activating a fault compensation strategy when exhaust temperature deviation is more than or equal to 8%; S4, issuing the optimized control instruction to the air path executing mechanism; The system comprises a physical layer, a digital twin layer, an intelligent decision layer and a virtual-real interaction layer, wherein the physical layer comprises a ship diesel engine body, a gas circuit executing mechanism and a sensor group; The sensor group comprises a supercharging pressure sensor, an exhaust temperature sensor, an intake oxygen concentration sensor and a turbine speed sensor, wherein the gas path executing mechanism comprises a variable section turbine executor and an exhaust gas recirculation valve; In S3, the operation steps of dynamic optimization include: B1, constructing a reinforcement learning-physical guiding hybrid optimizer, wherein an objective function of the reinforcement learning-physical guiding hybrid optimizer is J=w 1 ·BSFC+w 2 ·NOx+w 3 ·|P boost -P target |+λ·|deltau|, wherein BSFC is a brake fuel consumption rate, NOx is a nitrogen oxide emission amount, P boost is an actual boost pressure, P target is a desired boost pressure target value, deltau is a variation of a control amount, λ is a control smoothing factor, w 1 、w 2 、w 3 is a weight coefficient, and the control amount u= [ VGT opening degree, EGR rate ]; B2, solving an optimal solution set of an objective function on line through a Pareto front tracking engine, generating a 10 5 -group working condition simulation data training DDPG network by utilizing a digital twin body, and establishing a Pareto front curved surface of fuel efficiency-emission by adopting a non-dominant ordering genetic algorithm; And B3, dynamically selecting an operating point according to the ship navigation state, wherein the operating point comprises a lowest point of BSFC (base station back flow controller) in a cruising stage, a lowest point of NOx in an arrival-departure stage and a fastest supercharging response point in a transient acceleration stage.
2. The marine diesel engine gas circuit control method based on digital twinning of claim 1, wherein the virtual-real interaction layer realizes bidirectional real-time data synchronization of the physical layer and the digital twinning layer through OPC UA protocol, and synchronization delay is less than or equal to 1ms.
3. The marine diesel engine gas circuit control method based on digital twin according to claim 1, wherein in S2, the operation step of the multi-parameter coupling simulation includes: A1, receiving an original data stream of a sensor group in real time, and giving confidence weights to each sensor data through a dynamic confidence distribution module; a2, constructing a mixed input vector based on the confidence weight, and replacing a physical sensor value with the twin predicted value when the confidence weight is smaller than 0.7; a3, starting a layered hybrid simulation engine, performing three layers, namely finishing gas path parameter initial value calculation within 1ms based on a simplified mechanism model in a first layer, calling an LSTM dynamic sub-model to dynamically correct the supercharging pressure and the exhaust temperature for 10ms in a second layer, and activating a high-precision CFD simulation module to perform millisecond transient flow field analysis when the opening mutation of the EGR valve is detected to be more than 15% in a third layer.
4. The digital twin based marine diesel gas circuit control method according to claim 1, wherein in S3, the fault tolerant control comprises the steps of: C1, calculating the relative deviation of key parameters based on multidimensional deviation analysis of digital twin body output and sensor measured data; C2, starting a dynamic threshold fault triggering mechanism; c3, when any parameter deviation exceeds an actual threshold, performing fault root cause positioning, calling a fault knowledge base to perform Bayesian network reasoning, and outputting a fault type confidence vector; And C4, activating an adaptive compensation strategy chain according to the fault type to generate compensation control quantity.
5. The marine diesel engine gas circuit control method based on digital twinning according to claim 1, wherein in S4, the operation step of issuing the control command includes: D1, constructing a dual-channel redundant execution architecture, wherein a main channel transmits a reference control quantity to an actuator driver through a CAN bus, and a backup channel transmits a compensation control quantity to an actuator intelligent terminal through an industrial WiFi 6; Executing instruction pre-verification, inputting the instruction to be issued into a digital twin body to conduct millisecond closed loop pre-modeling, and triggering instruction reconstruction when the deviation between the predicted boost pressure and the target boost pressure exceeds 10 kilopascals or the predicted exhaust temperature exceeds a safe temperature limit; and D3, generating a final execution instruction by adopting a dynamic softening control algorithm.
6. The marine diesel engine gas circuit control method based on digital twinning according to claim 1, wherein in the step S1, the operation steps of gas flow dynamic modeling comprise the steps of constructing a physical guiding type ConvLSTM network, receiving a multi-sensor space-time data cube in an input layer, extracting local flow field characteristics by adopting a 3×3 convolution kernel in a convolution LSTM layer, and embedding a Navier-Stokes equation simplified form as a regularization term in a physical constraint layer.
7. The marine diesel engine gas circuit control method based on digital twinning is characterized in that in S1, the operation steps of modeling an in-cylinder combustion mechanism comprise the steps of constructing a dynamic partition coupling calculation framework, dividing a combustion chamber into a core combustion area, a boundary layer area and a wall area, wherein the core area adopts a chemical dynamics model, the boundary layer area adopts a simplified transportation model, the wall area adopts a transient heat conduction model, implementing self-adaptive chemical mechanism dimension reduction, dynamically selecting a reaction mechanism dimension based on real-time working condition parameters, and embedding a soot-NOx collaborative prediction module to calculate in real time through a coupling equation.
8. The marine diesel engine gas circuit control method based on digital twinning according to claim 1 is characterized in that in S1, the operation steps of constructing and updating a failure feature library comprise constructing a dynamic knowledge graph framework, implementing a multi-mode feature extraction engine, deploying a federal incremental learning mechanism and realizing a failure evolution prediction model.

Description

Digital twin-based marine diesel engine gas circuit control method Technical Field The invention relates to the technical field of marine diesel engine gas circuit control, in particular to a marine diesel engine gas circuit control method based on digital twinning. Background The gas circuit control of the marine diesel engine does not need battery and circuit dependence, is suitable for severe environments such as high humidity, low temperature and the like, has higher safety especially in flammable and explosive places, has firm structure, low failure rate, low maintenance cost and long service life, can be independently operated by an electric control system, can still maintain basic functions when power supply is interrupted, and ensures the safety of the marine, so the marine diesel engine has the advantages of high reliability, convenience in maintenance and strong adaptability. At present, the traditional static model cannot adapt to the variable working condition of the ship, the sensor is invalid due to the severe environment of the ship, and the flexibility of control instruction optimization is limited by the fixed weight coefficient. In view of the above, we propose a marine diesel engine gas circuit control method based on digital twinning to solve the existing problems. Disclosure of Invention The invention aims to provide a marine diesel engine gas circuit control method based on digital twinning, which aims to solve the problems in the background technology. In order to achieve the purpose, the invention provides the technical scheme that the marine diesel engine gas circuit control method based on digital twin comprises the following operation steps: S1, constructing a diesel engine multi-dimensional twin body, and fusing a gas flow dynamic sub-model based on an LSTM neural network, an in-cylinder combustion mechanism sub-model based on a thermodynamic law and a failure feature library based on historical failure data; s2, collecting sensor group data in real time, and driving a twin body to perform multi-parameter coupling simulation through a virtual-real interaction layer; S3, performing double optimization based on a simulation result, generating VGT/EGR reference control quantity by a dynamic optimization module with the aim of maximizing fuel efficiency, comparing output difference of a physical layer and a twin body by a fault-tolerant control module, and activating a fault compensation strategy when exhaust temperature deviation is more than or equal to 8%; And S4, issuing the optimized control instruction to the air path executing mechanism. The system comprises a physical layer, a digital twin layer, an intelligent decision layer and a virtual-real interaction layer, wherein the physical layer comprises a marine diesel engine body, an air path executing mechanism and a sensor group, the digital twin layer is provided with a diesel engine multi-dimensional twin body which integrates a real-time data driving model, a mechanism model and a fault knowledge base, the intelligent decision layer comprises a dynamic optimizing module and a fault tolerance control module, the virtual-real interaction layer realizes bidirectional real-time data synchronization of the physical layer and the digital twin layer through an OPC UA protocol, and the synchronization delay is less than or equal to 1ms. Further, the sensor group comprises a boost pressure sensor, an exhaust temperature sensor, an intake oxygen concentration sensor and a turbine rotating speed sensor, and the gas circuit executing mechanism comprises a variable-section turbine actuator and an exhaust gas recirculation valve. Further, in S2, the operation steps of the multiparameter coupling simulation include: A1, receiving an original data stream of a sensor group in real time, and giving confidence weights to each sensor data through a dynamic confidence distribution module; a2, constructing a mixed input vector based on the confidence weight, and replacing a physical sensor value with the twin predicted value when the confidence weight is smaller than 0.7; a3, starting a layered hybrid simulation engine, performing three layers, namely finishing gas path parameter initial value calculation within 1ms based on a simplified mechanism model in a first layer, calling an LSTM dynamic sub-model to dynamically correct the supercharging pressure and the exhaust temperature for 10ms in a second layer, and activating a high-precision CFD simulation module to perform millisecond transient flow field analysis when the opening mutation of the EGR valve is detected to be more than 15% in a third layer. Further, in S3, the operation steps of dynamic optimization include: B1, constructing a reinforcement learning-physical guidance hybrid optimizer; B2, solving an optimal solution set of an objective function on line through a Pareto front tracking engine, generating a 10 5 -group working condition simulation data training DDPG network by utilizi