A marine offshore HVAC system EPC turnkey intelligent design system

Abstract

This invention relates to an intelligent design system for EPC general contracting of marine HVAC systems, specifically in the field of marine technology. It effectively solves the technical challenges of unpredictable ship-engine coupling resonance and local structural fatigue in marine HVAC system design. By constructing a parametric coupled dynamic model and a physical information neural network, it achieves real-time, high-precision prediction of structural response under complex operating conditions, breaking through the computational bottleneck of traditional finite element calculations. The topology optimization strategy based on structural acoustic intensity flow can actively avoid energy convergence areas, significantly reducing the risk of vibration transmission. Furthermore, the closed-loop calibration mechanism using measured data endows the system with self-evolution capabilities, ensuring the true consistency between the design model and the physical entity, and greatly improving the design quality, construction efficiency, and operational safety of EPC general contracting projects.

Description

Technical Field

[0001] This invention relates to the field of marine technology, and more specifically, to an intelligent design system for EPC general contracting of marine HVAC systems. Background Technology

[0002] In the field of modern shipbuilding and marine engineering, floating facilities such as floating production storage and offloading (FPSO) vessels, drilling ships, and large ocean-going transport vessels operate in extremely harsh deep-sea environments for extended periods. These vessels not only need to withstand extremely high sea state loads from strong winds and waves, but also need to meet the comfort requirements of the crew's long-term living conditions and the environmental control requirements of precision instruments and equipment. Therefore, as a core component of ship auxiliary systems, the heating, ventilation, and air conditioning (HVAC) system has become increasingly large in scale, typically including high-power chillers, large air handling units, and a complex network of pipes throughout the ship. In actual operating scenarios, the hull structure is not an absolutely rigid body, but continuously undergoes elastic deformation and longitudinal bending under alternating wave loads. At the same time, the ship's main propulsion system, auxiliary generator sets, and the high-speed rotating components of the HVAC system itself together constitute a multi-source excitation environment, causing the hull deck, bulkheads, and superstructure to be in a state of continuous broadband vibration. Especially under high sea states, the overall dynamic response of the hull beams and the vibration characteristics of local plates and frames are superimposed, constructing an extremely complex dynamic physical field environment.

[0003] However, in existing ship EPC general contracting design processes, there is a common technical problem of disconnect between hull structural strength analysis and outfitting equipment dynamics design in the layout and installation design of HVAC systems. Traditional existing technical solutions typically employ the static load method, where designers treat the HVAC equipment as a stationary mass block, checking the support strength of its mounting base under hydrostatic pressure, or focusing solely on isolated vibration isolation design for the equipment itself, such as adding damping springs or rubber pads. A significant drawback of this design approach is that it ignores the fact that large HVAC equipment is a significant source of mass and vibration, and changes in its installation position directly affect the modal parameters of local hull plates, and fails to adequately consider the reverse dynamic input of hull structural elastic deformation to the equipment base under harsh sea conditions. Due to the lack of interdisciplinary fluid, structural, and equipment coupling analysis models, it is difficult to predict resonance phenomena caused by the coincidence of equipment operating frequencies with the natural frequencies of local hull structures during the design phase. Such design flaws often lead to fatigue cracks in the deck structure at the HVAC equipment base under specific sea or operating conditions after the ship is delivered and put into operation. This is caused by long-term exposure to alternating stresses that exceed expectations, which seriously damages the watertightness and structural integrity of the cabins. At the same time, the structural noise generated not only severely reduces the comfort of living, but may also cause the ship to fail the vibration and noise specifications of the relevant classification society, resulting in huge maintenance and modification and economic losses. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing an intelligent design system for EPC general contracting of marine HVAC systems, thereby resolving the issues raised in the background section.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: an intelligent design system for EPC general contracting of marine HVAC systems, comprising: a coupled dynamics modeling module, a response prediction model generation module, a layout optimization decision module, and a parameter closed-loop calibration module, wherein; The coupled dynamics modeling module, in response to the input of hull structure design parameters and HVAC equipment physical parameters, constructs parameterized ship-machine coupled dynamics equations that characterize the interaction between the hull structure and HVAC equipment; The response prediction model generation module obtains the parameterized ship-engine coupled dynamics equations, uses the parameterized ship-engine coupled dynamics equations as physical constraints to train the neural network, and generates a physical information vibration prediction proxy model that can predict structural vibration response data based on equipment position variables. The layout optimization decision module calls the physical information vibration prediction proxy model to calculate the structural vibration response value corresponding to different equipment installation locations, and selects the HVAC equipment optimization layout scheme that meets the preset vibration fatigue requirements based on the structural vibration response value. The parameter closed-loop calibration module corrects the model parameters in the parameterized ship-machine coupled dynamics equations based on the deviation between the measured data of the same type of ship and the output data of the physical information vibration prediction proxy model, and feeds the corrected model parameters back to the coupled dynamics modeling module.

[0006] In a preferred embodiment, the specific steps for the coupled dynamics modeling module to construct parameterized ship-engine coupled dynamics equations include: First, modal synthesis and dimensionality reduction characterization of the hull structural domain are performed. The system identifies interface nodes and internal nodes in the physical model of the hull structure based on the input hull structure design parameters. By solving the constraint principal modes and fixed interface principal modes, a modal transformation matrix is ​​constructed. The modal transformation matrix is ​​then used to perform orthogonal projection transformation on the hull structure mass matrix and hull structure stiffness matrix in the physical coordinate system to generate the hull generalized modal mass matrix and hull generalized modal stiffness matrix. Secondly, the system constructs a parameterized position-dependent coupling operator. It defines the HVAC equipment installation position vector as a variable design parameter, introduces a finite element shape function matrix and a numerical approximation distribution function to describe the spatial interpolation relationship between the HVAC equipment base and the continuous structural domain of the hull, calculates the physical coupling effect generated by the physical stiffness tensor of the HVAC equipment base at the current installation position, and uses a modal transformation matrix to project the physical coupling effect onto the generalized modal space to generate a parameterized generalized coupling stiffness matrix that changes in real time with the HVAC equipment installation position vector.

[0007] In a preferred embodiment, the specific steps of the coupled dynamics modeling module in constructing parameterized ship-engine coupled dynamics equations further include: Finally, the global dynamic equations in the hybrid coordinate system are assembled. The system establishes a hybrid coordinate system containing the generalized modal coordinate vector of the hull and the physical displacement vector of the HVAC equipment. An antisymmetric gyroscope matrix characterizing the Coriolis force effect of the high-speed rotating components of the HVAC equipment is constructed. The dimensional mapping matrix is ​​used to align and assemble the generalized modal mass matrix of the hull, the generalized modal stiffness matrix of the hull, the inertial matrix of the HVAC equipment extracted based on the physical parameters of the HVAC equipment, and the parameterized generalized coupling stiffness matrix. A set of differential-algebraic equations that can describe the dynamic response of the system under the action of the generalized external load vector is constructed, which is the parameterized ship-machine coupled dynamic equation.

[0008] In a preferred embodiment, the specific steps of the response prediction model generation module in generating the physical information vibration prediction proxy model include: First, the parameterized spatiotemporal manifold neural mapping architecture is constructed. The system establishes a deep neural network architecture, using the time independent variable and the HVAC equipment installation position vector as the input layer data of the deep neural network. A nonlinear mapping function from the input layer to the output layer is constructed, and the output layer of the deep neural network is configured to directly predict the system's generalized state prediction vector. The system's generalized state prediction vector includes the ship's generalized modal coordinate vector and the HVAC equipment's physical displacement vector. Secondly, the system performs automatic differential of the physical residuals by embedding parameterized coupling operators. The system uses automatic differential techniques to calculate the first and second derivatives of the generalized state prediction vector of the system with respect to the time independent variable. It extracts the system global mass matrix, system structural Rayleigh damping matrix, antisymmetric gyroscope matrix, and system basic stiffness matrix that constitute the parameterized ship-engine coupled dynamic equations. The parameterized generalized coupling stiffness matrix is ​​explicitly embedded as an interface to characterize the influence of position variables on the physical field. Through linear combination, a dynamic physical residual operator that can quantify the degree to which the current prediction value violates the dynamic equilibrium law is constructed.

[0009] In a preferred embodiment, the specific steps of the response prediction model generation module in generating the physical information vibration prediction proxy model further include: Finally, the loss function of physical constraint and data-driven fusion is minimized. The system randomly samples physical configuration points in the spatiotemporal domain and calculates the norm of the dynamic physical residual operator to construct the physical residual loss term. At the same time, a small number of simulation ground truth samples calculated based on finite element analysis are obtained, and the deviation between the generalized state prediction vector of the system and the simulation ground truth samples is calculated to construct the data observation loss term. The global training objective function is constructed by weighted summation of the physical residual loss term and the data observation loss term. The global training objective function is minimized by gradient-based optimization algorithm to iteratively update the weights and bias parameters of the deep neural network until the global training objective function converges, thus generating the physical information vibration prediction surrogate model.

[0010] In a preferred embodiment, the specific steps for the layout optimization decision module to generate an optimized layout scheme for HVAC equipment include: First, the system performs a global structural acoustic intensity vector field reconstruction. Within the preset feasible design domain, the system performs a discretized grid scan on the HVAC equipment installation location vector. For each candidate location generated by the scan point, the system calls the physical information vibration prediction proxy model to obtain the generalized state prediction vector of the system in the time domain. The system generalized state prediction vector is restored to the physical stress field and velocity field using the finite element shape function derivative operator. The instantaneous structural acoustic intensity vector is calculated based on the physical stress field and velocity field and then time-averaged to construct the structural acoustic intensity vector field. Finally, the divergence distribution of the structural acoustic intensity vector field is calculated to identify the energy source region and energy sink region in the hull structure. Secondly, the system performs a non-stationary cumulative fatigue damage assessment. Based on the stress history data derived from the generalized state prediction vector of the system, the system uses the continuous domain approximation of the rainflow counting method and combines the material SN curve parameters to calculate the non-stationary fatigue damage index for predefined key fatigue hotspots in the hull structure. By operating on the stress concentration correction factor and the VonMises equivalent stress mapping operator, a global cumulative fatigue damage index is generated that can quantify the degree of irreversible damage to the key structure of the hull if the HVAC equipment is installed in the current location.

[0011] In a preferred embodiment, the specific steps of the layout optimization decision module in generating an optimized layout scheme for HVAC equipment further include: Finally, a multi-objective potential field topology optimization decision is performed, and a multi-physics comprehensive evaluation potential function is constructed. This function is formed by weighted fusion of the divergence distribution of the structural acoustic intensity vector field, the global cumulative fatigue damage index, and the geometric constraint penalty term characterizing the geometric dimensions and maintenance space limitations of the HVAC equipment. The global minimum point of the multi-physics comprehensive evaluation potential function is searched within the feasible design domain using a hybrid simulated annealing strategy, and the HVAC equipment installation location vector that minimizes the multi-physics comprehensive evaluation potential function is determined as the optimal layout scheme for the HVAC equipment.

[0012] In a preferred embodiment, the specific steps for the parameter closed-loop calibration module to correct the model parameters in the parameterized ship-engine coupled dynamics equations include: Heterogeneous data mapping and random likelihood function construction are performed. The system acquires measured vibration response time-series data of the target type of ship under typical working conditions, defines a sparse observation mapping matrix, and uses the sparse observation mapping matrix to project the system generalized state prediction vector output by the physical information vibration prediction proxy model onto the physical coordinate space corresponding to the measured vibration response time-series data. The uncertainty parameter set of the model to be calibrated, which constitutes the parameterized ship-machine coupled dynamic equation, is identified. The deviation between the measured vibration response time-series data and the projected prediction data is calculated. Combined with the prediction error covariance matrix, which characterizes the statistical characteristics of measurement noise and model structural error, a log-likelihood function is constructed to evaluate the probability ability of the current parameter combination to interpret the measured data.

[0013] In a preferred embodiment, the specific steps of the parameter closed-loop calibration module in correcting the model parameters in the parameterized ship-engine coupled dynamics equations further include: To perform posterior parameter inference based on Markov chain Monte Carlo, the system sets the prior probability distribution of the parameter set of the uncertainty parameter set of the model to be calibrated, combines the log-likelihood function, and uses the Markov chain Monte Carlo sampling algorithm to extract samples from the joint probability distribution to approximate the posterior probability density function of the parameters. By statistically analyzing a large number of posterior samples, the maximum a posteriori estimated parameter set is extracted, and the maximum a posteriori estimated parameter set is defined as the corrected model physical parameters.

[0014] In a preferred embodiment, the specific steps of the parameter closed-loop calibration module in correcting the model parameters in the parameterized ship-engine coupled dynamics equations further include: After reconstructing the dynamic baseline equations and backfeeding knowledge, the system feeds back the corrected model physical parameters to the coupled dynamic modeling module. This triggers the coupled dynamic modeling module to use the corrected model physical parameters to numerically update and reconstruct the system structure Rayleigh damping matrix and the parameterized generalized coupled stiffness matrix that constitute the parameterized ship-engine coupled dynamic equations, thereby completing the closed-loop calibration of the physical knowledge of the intelligent design system.

[0015] The beneficial effects of this invention are as follows: It effectively solves the technical challenge of unpredictable ship-engine coupled resonance and local structural fatigue in marine HVAC system design. By constructing a parametric coupled dynamic model and a physical information neural network, it achieves real-time, high-precision prediction of structural response under complex operating conditions, breaking through the computational bottleneck of traditional finite element calculations. The topology optimization strategy based on structural acoustic intensity flow can actively avoid energy convergence areas, significantly reducing the risk of vibration transmission. Furthermore, the closed-loop calibration mechanism using measured data endows the system with self-evolution capabilities, ensuring the true consistency between the design model and the physical entity, and greatly improving the design quality, construction efficiency, and operational safety of EPC general contracting projects. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system block diagram of the method of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0019] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0020] This embodiment provides, for example Figure 1-2The intelligent design system for EPC general contracting of marine HVAC systems, as shown, includes: a coupled dynamics modeling module, a response prediction model generation module, a layout optimization decision module, and a parameter closed-loop calibration module, wherein; The coupled dynamics modeling module, in response to the input of hull structure design parameters and HVAC equipment physical parameters, constructs parameterized ship-machine coupled dynamics equations that characterize the interaction between the hull structure and HVAC equipment; The response prediction model generation module obtains the parameterized ship-engine coupled dynamics equations, uses the parameterized ship-engine coupled dynamics equations as physical constraints to train the neural network, and generates a physical information vibration prediction proxy model that can predict structural vibration response data based on equipment position variables. The layout optimization decision module calls the physical information vibration prediction proxy model to calculate the structural vibration response value corresponding to different equipment installation locations, and selects the HVAC equipment optimization layout scheme that meets the preset vibration fatigue requirements based on the structural vibration response value. The parameter closed-loop calibration module corrects the model parameters in the parameterized ship-machine coupled dynamics equations based on the deviation between the measured data of the same type of ship and the output data of the physical information vibration prediction proxy model, and feeds the corrected model parameters back to the coupled dynamics modeling module.

[0021] In a preferred embodiment of the present invention, the core function of the coupled dynamics modeling module is to construct a physical and mathematical model that can respond to design changes in real time.

[0022] The specific steps for constructing parameterized ship-engine coupled dynamics equations using the coupled dynamics modeling module include: First, modal synthesis and dimensionality reduction characterization of the hull structural domain are performed. Based on the input hull structural design parameters, the system identifies interface nodes and interior nodes in the hull structural physical model. In this step, given that the degrees of freedom of a hull structure are typically in the millions, direct dynamic calculations are extremely inefficient. The system employs the Craig-Bampton substructure synthesis method, defining the hull structure as a parent-child structure, and identifies the interface nodes of the area to be installed, as well as the remaining interior nodes.

[0023] The mode transformation matrix is ​​constructed by solving for the constrained principal modes and the fixed interface principal modes. This matrix is ​​a mapping bridge connecting the high-dimensional physical space and the low-dimensional modal space, and it consists of the constrained principal modal set and the truncated fixed interface principal modal set.

[0024] Using the mode transformation matrix For the hull structure mass matrix in the physical coordinate system With the hull structure stiffness matrix Perform orthogonal projection transformation to generate the generalized modal mass matrix of the hull. With the generalized modal stiffness matrix of the hull ; The specific transformation calculation is performed using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: The mass matrix of the hull structure in the physical coordinate system is used to characterize the mass distribution characteristics of the hull structure in high-fidelity physical space. The stiffness matrix of the hull structure in the physical coordinate system is used to characterize the elastic deformation characteristics of the hull structure in high-fidelity physical space. Let represent the Craig-Bampton mode transformation matrix, as a dimension reduction projection operator; This represents the generalized modal mass matrix of the ship's hull, which is calculated using this formula. Its purpose is to compress a physical mass matrix with millions of degrees of freedom into a low-dimensional matrix, significantly reducing computational and storage requirements. This represents the generalized modal stiffness matrix of the hull, which is calculated using this formula. Its purpose is to preserve the key natural frequencies and mode shapes of the hull structure while reducing the amount of computation.

[0025] In this embodiment, the hull structure material is set to marine high-strength steel, and its density reference value is set to [value missing]. The Young's modulus reference value is set as follows: With a Poisson's ratio of 0.3, these parameters constitute the physical matrix. and The foundation.

[0026] In the simulation scenario of this embodiment, the original ship hull physical model has a degree of freedom on the order of 10^5. (Millions of levels). By setting the cutoff frequency to 0~200Hz (covering the main excitation frequency band of HVAC equipment), the transformation matrix is ​​constructed by selecting the first 150 fixed interface principal modes and 60 interface degrees of freedom. The generalized mode matrix generated after calculation using this formula is... and The dimension is only This significant reduction in dimensionality reduces the computation time for subsequent coupled iterations from hours to seconds.

[0027] This formula effectively solves the problem that existing finite element models (FEMs) of ship hulls typically contain millions of degrees of freedom (DOF), and directly calculating a vibration response using physical matrices can take hours. This formula compresses a matrix of millions of degrees into a generalized matrix of only a few hundred orders. , While preserving the key dynamic characteristics of the hull (such as the main resonance frequency and mode shape), the computational load is reduced by several orders of magnitude, thereby supporting real-time iterative optimization of HVAC equipment positions and laying the computational foundation for subsequent rapid coupling calculations.

[0028] Secondly, by executing the parameterized position-dependent coupling operator construction, the system will convert the HVAC equipment installation location vector Defined as variable design parameters, this step innovatively introduces a moving shape function interpolation technique to avoid mesh remeshing caused by equipment movement, and introduces a finite element shape function matrix. With numerical approximation distribution function To describe the spatial interpolation relationship between the HVAC equipment base and the continuous structural domain of the hull, the system uses a numerical approximation of the Dirac distribution function to mathematically locate the specific point of action of the equipment base within the continuous hull structural domain. The physical stiffness tensor of the HVAC equipment base is then calculated. The physical coupling effect generated at the current installation location, and the use of the mode transformation matrix. Projecting the physical coupling effect into the generalized modal space generates a vector that varies with the HVAC equipment installation location. Real-time changing parameterized generalized coupled stiffness matrix .

[0029] The specific operator construction is processed using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: This represents the HVAC equipment installation location vector, which serves as the core variable design parameter of this system. Its value range is the set of feasible regions for the cabin deck. Represents the spatial coordinate variables of the hull structure, used to traverse the continuous integral domain of the hull structure surface; Represents the shape function matrix of the finite element, used to describe the displacement interpolation relationship in continuous physical space; This represents the approximate distribution function of the Dirac numerical value, used for precise positioning of the device's point of action; Represents the physical stiffness tensor of the HVAC equipment base, used to characterize the local physical connection properties between the equipment and the ship's deck; This represents the parameterized generalized coupling stiffness matrix, which is calculated using this formula.

[0030] In this embodiment, the selected HVAC equipment is a large marine chiller unit, and its base uses a spring-damped composite shock absorber. The physical stiffness tensor of the base is set according to the parameters provided by the equipment supplier. The reference range for the main diagonal elements is: This parameter value not only determines the vibration isolation efficiency of the equipment itself, but also directly affects the energy transfer characteristics of the coupled system.

[0031] In this embodiment, the ship deck model is discretized using four-node shell elements, and the corresponding shape function matrix is... A bilinear interpolation polynomial was constructed. This was done to accurately capture the device position during numerical integration. The aforementioned Dirac function An approximation is performed using a smooth kernel function, with its effective support region defined as the physical contact area of ​​the device base (e.g., ...). (to prevent numerical singularity).

[0032] The core purpose of this formula in this implementation is to establish a stiffness operator that updates in real time as the design parameters change. This means that when the optimization algorithm adjusts the device position, the system only needs to recalculate the low-dimensional matrix without rebuilding the entire geometric model, thus achieving true "parametric" driving.

[0033] The specific steps for constructing parameterized ship-engine coupled dynamics equations using the coupled dynamics modeling module also include: Finally, the global dynamic equations are assembled in a hybrid coordinate system, and the system establishes a coordinate vector containing the generalized modal coordinates of the hull. Physical displacement vector of HVAC equipment To accurately describe the complex dynamic behavior inside the equipment, the system employs a hybrid coordinate system modeling strategy, constructing an antisymmetric gyroscope matrix that characterizes the Coriolis force effect of high-speed rotating components in HVAC equipment. This matrix is ​​used to reflect the gyroscopic torque generated by the rotor of a high-power fan or compressor during the rolling motion of a ship. It utilizes a dimension mapping matrix. The generalized modal mass matrix of the ship hull hull generalized modal stiffness matrix HVAC equipment inertia matrix extracted based on HVAC equipment physical parameters and parameterized generalized coupling stiffness matrix Dimension alignment and assembly are performed to construct a vector that can describe the generalized external load vector. The set of differential-algebraic equations for the dynamic response of the system under action is the parameterized ship-engine coupled dynamic equation.

[0034] The specific equation assembly is handled using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: Represents the generalized modal coordinate vector of the hull, used to describe the vibration response of the hull structure in modal space; This represents the physical displacement vector of the HVAC equipment, used to describe the translational and rotational responses of the HVAC equipment in physical space. This represents the inertia matrix of the HVAC equipment, which includes information on the equipment's mass and moment of inertia. This represents the stiffness matrix of the HVAC equipment itself, characterizing the structural elasticity of the equipment body; The Rayleigh damping matrix represents the system structure and is used to characterize the energy dissipation characteristics of the system. The antisymmetric gyroscope matrix is ​​used to characterize the Coriolis force coupling effect generated by high-speed rotating components inside the device. This represents the dimension mapping matrix, used to handle the alignment and transformation of different degrees of freedom in a mixed coordinate system; It represents the generalized external load vector, which includes the projected external wave load and the unbalanced excitation force inside the equipment.

[0035] In this embodiment, the system damping matrix A Rayleigh Damping model was used to construct the first and second modal damping ratios of the hull structure, which were then defined. Reference value This is to simulate the energy dissipation characteristics of steel structures within their elastic range.

[0036] Considering the high-speed rotation characteristics of the centrifugal compressor inside the HVAC unit, this embodiment constructs an antisymmetric gyroscope matrix. The rated operating speed of the compressor rotor was set to 3600 RPM, and the moment of inertia was extracted based on the parameters of the equipment rotor drawing. This was used to quantify the Coriolis force coupling effect generated by the equipment under the swaying condition of the ship.

[0037] Generalized external load vector Both external sea state and internal excitation were taken into account. The sea state load was generated based on the JONSWAP wave spectrum, with the corresponding characteristic wave height set at 5.5m (sea state level 5). The internal excitation force of the equipment was calculated based on the residual unbalanced force according to the G2.5 equilibrium level under the ISO1940-1 standard.

[0038] This formula is a parametric global dynamic differential equation for ship-machine coupling in a mixed coordinate system. It integrates the hull structure, equipment body, connecting base, rotational effects, and external loads. By solving this equation (or having a neural network learn this equation), we can predict the magnitude of hull and equipment vibrations under any sea state and any installation location.

[0039] In a preferred embodiment of the present invention, the response prediction model generation module is mainly responsible for using artificial intelligence technology to replace the traditional numerical solver.

[0040] The specific steps of the response prediction model generation module in generating the physical information vibration prediction surrogate model include: First, the parameterized spatiotemporal manifold neural mapping architecture is constructed, and a deep neural network architecture is established. In this step, to adapt to the hybrid coordinate system established by the preceding modules, a multi-layer fully connected deep neural network (DNN) is constructed. The time independent variable... Vector of HVAC equipment installation location As input layer data for a deep neural network, here, the time independent variable... Covers the preset simulation duration (e.g.) HVAC equipment installation location vector It directly inherits from the preceding module and serves as the variable-condition input for the network. A nonlinear mapping function is constructed from the input layer to the output layer, and the output layer of the deep neural network is configured to directly predict the generalized state prediction vector of the system. To ensure the existence of the second derivative in subsequent calculations, the network employs a smooth nonlinear activation function (such as the hyperbolic tangent function tanh). The system's generalized state prediction vector. It includes the generalized modal coordinate vector of the hull and the physical displacement vector of the HVAC equipment, which is the network's estimate of the system's vibration response at a specific time and location.

[0041] The specific neural mapping architecture is processed using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: It represents the independent variable of time, serving as the evolutionary dimension of the dynamic system; This represents the HVAC equipment installation location vector, which serves as the variable operating condition input for the network; This represents the set of parameters to be trained in the neural network, including the weight matrices of each layer. and bias vector Its function is to store the physical laws learned by the proxy model; This represents a smooth nonlinear activation function, which enables the network to fit nonlinear dynamic behavior and ensures second-order continuous differentiability. This represents the system's generalized state prediction vector, which is the output of the physical information vibration prediction surrogate model.

[0042] In this embodiment, the deep neural network adopts a fully connected architecture.

[0043] Network depth: Set the number of hidden layers to 4 to 8 (e.g., 6 layers) to ensure sufficient feature extraction capability; Network width: Set each layer to contain 40 to 100 neurons (e.g., 50) to balance fitting accuracy and computational efficiency; Activation function The hyperbolic tangent function (Tanh) is chosen because it has infinite differentiability, which satisfies the mathematical requirement for the existence of the second derivative in physical residual calculation.

[0044] Network weight matrix Random values ​​are assigned using the Xavier initialization method, and the bias vector is... Initialize to a zero vector to accelerate training convergence.

[0045] In this implementation, the formula establishes a direct mapping from "design variable (position) + time" to "system response (vibration)". Once trained, given position and time, it can calculate the result in milliseconds using simple matrix multiplication, without needing to solve complex differential equations. Secondly, the system performs automatic differential of the physical residuals with embedded parameterized coupling operators, and uses automatic differential techniques to calculate the system's generalized state prediction vector. Regarding the independent variable of time The first and second derivatives. This step is crucial for achieving "physics-driven" operation. The system does not rely on difference approximation but instead utilizes the automatic differentiation (AD) mechanism of deep learning frameworks for precise symbolic differentiation.

[0046] Extracting the system global mass matrix that constitutes the parameterized ship-engine coupled dynamics equations System structure Rayleigh damping matrix Antisymmetric gyroscope matrix and the system's fundamental stiffness matrix And parameterize the generalized coupling stiffness matrix As an explicit embedding of the interface representing the influence of location variables on the physical field, this means that the neural network must not only learn the laws of temporal evolution, but also parameterize the generalized coupling stiffness matrix. The study investigates the nonlinear modulation effect of equipment position changes on the system stiffness matrix. A kinetic residual operator is constructed through linear combination to quantify the degree to which the current predicted value violates the laws of dynamic equilibrium. .

[0047] The specific physical residual calculation is performed using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: This represents the system's global mass matrix, used to characterize the system's overall inertial characteristics. The Rayleigh damping matrix represents the system structure and is used to characterize the energy dissipation characteristics of the system. The antisymmetric gyroscope matrix is ​​used to characterize the Coriolis force coupling effect generated by high-speed rotating components inside HVAC equipment. This represents the system's fundamental stiffness matrix, used to characterize the elastic properties of the system without coupling terms; This represents the parameterized generalized coupling stiffness matrix, which serves as a mathematical interface for how positional variables influence physical fields. This represents the nth derivative of the system's generalized state prediction vector with respect to time. This represents the generalized external load vector, which includes external wave loads and unbalanced excitation forces within the equipment. Represents the dynamic physical residual operator (or physical residual vector).

[0048] In calculating the derivative term In this embodiment, instead of using the finite difference approximation, the automatic differentiation (AD) engine of the deep learning framework is used for accurate solution to avoid the destruction of physical conservation by discretization error.

[0049] In construction At that time, set the rotor pole moment of inertia of the centrifugal compressor in the HVAC equipment. The rated operating speed is 3600 RPM, thereby introducing the dynamic coupling effect of rotating machinery on the local modes of the hull.

[0050] The wave excitation part in the model is generated based on the JONSWAP spectrum, with a defined wave height. Crossing the zero cycle Simulates random wave loads under severe sea conditions in the North Atlantic.

[0051] In this implementation, the formula transforms complex differential equations into computable algebraic error terms, serving as "physical constraints" to guide neural network training. The specific steps of the response prediction model generation module in generating the physical information vibration prediction surrogate model also include: Finally, the loss function of the physical constraint and data-driven fusion is minimized. The system randomly samples physical placement points in the spatiotemporal domain and calculates the dynamic physical residual operator. The norm is used to construct the physical residual loss term.

[0052] These physical placement points are randomly sampled in the spatiotemporal domain without real labels to verify whether the model conforms to physical laws. Simultaneously, a small number of simulation ground truth samples based on finite element analysis are obtained. Calculate the generalized state prediction vector of the system With simulation true samples The deviation between the physical residual and the data observation loss term is used to construct the data observation loss term. Introducing a small number of ground truth samples is to compensate for potential convergence difficulties in pure physical training. A global training objective function is constructed by weighted summation of the physical residual loss term and the data observation loss term. And it uses a gradient-based optimization algorithm to minimize the global training objective function. The weights and bias parameters of the deep neural network are iteratively updated until the global training objective function is reached. Convergence, i.e., generating a physical information vibration prediction proxy model.

[0053] The specific loss function is constructed using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: The number of physical configuration points represents the number of unlabeled points randomly sampled in the spatiotemporal domain. This represents the number of finite element observation data points, which is the number of small sample points used for supervised learning; This represents the true value sample from the finite element simulation, serving as a monitoring signal for the known solution; , This represents the dynamic weighting coefficient, which balances the contributions of physical constraints and data fitting to gradient descent. This represents the square of the Euclidean norm, used to quantify the magnitude of the error. This represents the global training objective function.

[0054] In this embodiment, the formula serves as the minimization objective of the optimization algorithm, forcing the neural network to simultaneously satisfy both data approximation and physical equation constraints, thereby training a high-precision physical information vibration prediction surrogate model.

[0055] In a preferred embodiment of the present invention, the layout optimization decision module no longer relies on experience-based judgment, but instead utilizes the rapid prediction capabilities provided by the preceding modules to perform a panoramic scan of the design space from two deep physical dimensions: "energy transfer" and "damage accumulation." The specific steps by which the layout optimization decision module generates an optimized layout scheme for HVAC equipment include: First, a global structural acoustic intensity vector field reconstruction is performed. The system reconstructs the HVAC equipment installation location vector within a preset feasible design domain. Discretized mesh scanning is performed, in which the system first performs a mesh scan of the HVAC equipment installation location vector within the feasible design domain. The system is divided into grids, and for each candidate location generated at a scan point, the physical information vibration prediction proxy model is invoked to obtain the generalized state prediction vector of the system in the time domain. This process leverages the millisecond-level inference capabilities of the surrogate model. Subsequently, the generalized state prediction vector of the system is calculated using the finite element shape function derivative operator. Reducing to physical stress field and velocity field This is to transform abstract data from modal space into energy characteristics of physical space. Based on the physical stress field. and velocity field Calculate the instantaneous structural acoustic intensity vector and perform time averaging to construct the structural acoustic intensity vector field. Then, the structural acoustic intensity vector field is calculated. The divergence distribution is used to identify energy source and sink regions in the hull structure; The specific calculation of the structural acoustic intensity divergence field is handled using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: This represents the system's generalized state prediction vector, which originates from the output of the preceding physical information vibration prediction surrogate model. The prediction and assessment period is typically taken as a complete statistical period of wave load. This represents the stress recovery operator, used to map generalized displacements to stress tensors in physical space; This represents the velocity recovery operator, used to map generalized velocities to velocity vectors in physical space; It represents the time-averaged structural acoustic intensity vector, which characterizes the vibrational energy passing through a unit cross-sectional area per unit time. This represents the divergence operator.

[0056] In this embodiment, the stress recovery operator and velocity recovery operator It is constructed based on the shape function derivative matrix of a four-node isoparametric shell element. Through these two operators, the system can map the generalized modal displacements output by the surrogate model back to the stress tensor field and velocity vector field in physical space, thereby realizing the energy characteristic restoration from modal space to physical space.

[0057] Integral Time Window The value is set to 12.0 seconds. This value is selected based on the wave average zero-cycle of typical sea conditions in the target sea area to ensure that the time-domain average calculation of the structural acoustic intensity can fully smooth out transient fluctuations and reflect the steady-state energy transfer characteristics of the system.

[0058] In this embodiment, the formula serves as a "compass" for the optimization algorithm. By minimizing this function, the system can automatically find an optimal solution that avoids the resonance point, cuts off energy transmission, and is geometrically feasible.

[0059] Secondly, the system performs a non-stationary cumulative fatigue damage assessment. Based on the stress history data derived from the generalized state prediction vector of the system, the system uses the continuous domain approximation of the rainflow counting method and combines the material SN curve parameters to calculate the non-stationary fatigue damage index for predefined key fatigue hotspots in the hull structure. By operating on the stress concentration correction factor and the VonMises equivalent stress mapping operator, a global cumulative fatigue damage index is generated that can quantify the degree of irreversible damage to the key structure of the hull if the HVAC equipment is installed in the current location.

[0060] The specific calculation of the non-stationary fatigue cumulative damage index is handled using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: This represents the HVAC equipment installation location vector, which is the decision variable to be optimized. Indicates the number of critical fatigue hotspots, referring to predefined high-risk areas of stress concentration in the hull structure; This represents the Von Mises equivalent stress mapping operator, which transforms the state vector into a scalar equivalent stress. These represent the parameters of the material's SN curve, which respectively represent the fatigue strength coefficient and fatigue strength index; This represents the stress concentration correction factor, which is determined for structural forms at different node locations. This represents the global cumulative fatigue damage index.

[0061] In this embodiment, fatigue assessment follows the DNVGL-RP-C203 fatigue design specification. For welded joints on the hull deck, a Class D SN curve is selected, and the fatigue strength index is set. fatigue strength coefficient This was used to simulate the fatigue failure behavior of high-strength steel in a typical marine environment.

[0062] Considering the local geometric abrupt changes at the weld between the HVAC equipment base and the deck, a stress concentration correction factor is set. The default value is 1.5, which is dynamically adjusted to 2.5 for high stress areas at corners to compensate for the limitations of the global finite element model in terms of mesh size.

[0063] In this embodiment, the formula transforms the complex random stress history into a scalar damage value, which serves as a direct criterion for determining whether the installation location is safe.

[0064] The specific steps involved in generating an optimized HVAC equipment layout plan by the layout optimization decision module also include: Finally, a multi-objective potential field topology optimization decision is performed, and the system constructs a multi-physics comprehensive evaluation potential function. This step is the final decision-making stage. The function is derived from the structural acoustic intensity vector field. divergence distribution, global cumulative fatigue damage index And geometric constraint penalty terms characterizing the geometric dimensions and maintenance space limitations of HVAC equipment. This potential function, weighted and fused, transforms complex engineering constraints into mathematical "terrain," where "valleys" represent the optimal solution. A multiphysics comprehensive evaluation potential function is searched within the feasible design domain using a hybrid simulated annealing strategy. The global minimum point will be found, and the multiphysics comprehensive evaluation potential function will be determined. Minimize the HVAC equipment installation location vector The optimal layout scheme for HVAC equipment has been determined.

[0065] The specific multiphysics comprehensive evaluation potential function is constructed using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: , This represents the weighting coefficient, which is adjusted according to the shipowner's emphasis on structural safety and vibration comfort. This represents a reference fatigue damage value, used for dimensional normalization. It represents the total norm of global energy dissipation and is used to characterize the activity of vibrational energy in a structure; This represents a geometric constraint penalty term that tends to infinity when the location causes interference between the equipment and the ship's structure. This represents the multiphysics field comprehensive evaluation potential function (or comprehensive optimization objective function).

[0066] In this embodiment, the optimization of the FPSO superstructure takes into account that the safety of the hull structure is more important than living comfort, and fatigue weights are set accordingly. Energy weight It should be noted that this weight allocation can be adaptively adjusted according to the mission profile of the ship type (such as passenger ro-ro ship, research vessel).

[0067] To eliminate the difference in numerical magnitude between the fatigue damage index (dimensionless) and energy dissipation (dimension of power), a reference fatigue damage value is introduced. Normalization is performed. In this embodiment, The value is 1.0, which corresponds to the theoretical allowable limit of cumulative damage during the design life.

[0068] Geometric constraint penalty term The system employs an exterior penalty function method. It reads the coordinates of obstacles from the ship's general arrangement drawing; once the equipment outline interferes with an obstacle, the corresponding value is output. This forces the optimization algorithm to immediately abandon the infeasible solution.

[0069] In this embodiment, the formula serves as a "compass" for the optimization algorithm. By minimizing this function, the system can automatically find an optimal solution that avoids the resonance point, cuts off energy transmission, and is geometrically feasible.

[0070] In a preferred embodiment of the present invention, the parameter closed-loop calibration module serves as the self-evolution engine of the intelligent design system, aiming to address the objective deviation between the theoretical model during the design phase and the actual physical entity after construction and delivery. The specific steps of the parameter closed-loop calibration module in correcting the model parameters in the parameterized ship-engine coupled dynamics equations include: By performing heterogeneous data mapping and constructing random likelihood functions, the system obtains measured vibration response time-series data of similar target ships under typical operating conditions. These data typically originate from the B63B 79 / 00 monitoring system. Due to the limited number of sensor deployment points, the system defines a sparse observation mapping matrix. This matrix is ​​used to solve the spatial matching problem between the degrees of freedom of the finite element model and the finite number of sensor measurement points. It utilizes a sparse observation mapping matrix. The system generalized state prediction vector output by the physical information vibration prediction surrogate model Projected onto measured vibration response time series data Based on the corresponding physical coordinate space, the system identifies the set of uncertainty parameters of the model to be calibrated that constitute the parameterized ship-engine coupled dynamic equations. This mainly includes physical quantities that are difficult to measure directly, such as structural modal damping ratio and contact stiffness coefficient at connection interfaces. Calculation of measured vibration response time-series data. Compared with the projected prediction data The deviation between them, combined with the prediction error covariance matrix which characterizes the statistical properties of measurement noise and model structure error. Construct a log-likelihood function to evaluate the ability of the current parameter combination to interpret the probability of the measured data. .

[0071] The specific construction of the log-likelihood function is carried out using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: The measured vibration response time series data at the k-th time step serves as the benchmark evidence for calibration. This represents the system generalized state prediction vector at time step k, which is calculated by the physical information vibration prediction surrogate model. This represents a sparse observation mapping matrix, which is used to project a high-dimensional global prediction vector onto the physical coordinate space of a finite number of sensors. This represents the prediction error covariance matrix, used to characterize the statistical properties of measurement noise and model structure errors; This represents the set of uncertainty parameters of the model to be calibrated and is the core operation object of this module; This represents the log-likelihood function.

[0072] In this embodiment, the formula rigorously quantifies the deviation between "model prediction" and "physical measurement" using weighted Mahalanobis distance, providing a statistical gradient direction for subsequent probabilistic inference.

[0073] The specific steps for the parameter closed-loop calibration module to correct the model parameters in the parameterized ship-engine coupled dynamics equations also include: To perform posterior parameter inference based on Markov chain Monte Carlo methods, the system sets the uncertainty parameter set of the model to be calibrated. Prior probability distribution of parameters This step introduces Bayesian statistical inference theory, fusing subjective design experience (prior) with objective measured data. It combines the log-likelihood function... The Markov chain Monte Carlo sampling algorithm is used to extract samples from the joint probability distribution to approximate the parametric posterior probability density function. .

[0074] The specific posterior probability update of the parameters is processed using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: ) represents the prior probability distribution of the parameters, which is set based on design specifications or historical ship type data; It represents the posterior probability density function of the parameters, which is the latest cognitive distribution of the model parameters after incorporating evidence from measured data.

[0075] In this embodiment, the formula realizes the true distribution characteristics of physical parameters such as structural damping and connection stiffness from noise data.

[0076] The maximum a posteriori (MAP) parameter set was extracted by statistically analyzing a large number of posterior samples. And the maximum a posteriori estimated parameter set Defined as the corrected physical parameters of the model.

[0077] The specific steps for the parameter closed-loop calibration module to correct the model parameters in the parameterized ship-engine coupled dynamics equations also include: By performing dynamic baseline equation reconstruction and knowledge re-injection, the system will correct the physical parameters of the model. Feedback is sent to the coupled dynamics modeling module, the endpoint of the closed loop, which aims to translate the results of statistical inference into concrete physical model updates. This triggers the coupled dynamics modeling module to utilize the corrected model physical parameters. Numerical updates and reconstructions are performed on the system structure Rayleigh damping matrix and parameterized generalized coupling stiffness matrix that constitute the parameterized ship-engine coupled dynamics equations, in order to complete the closed-loop calibration of the physical knowledge of the intelligent design system.

[0078] The specific parameterized matrix closed-loop correction is handled using the following formula: ; The specific meanings and functions of each parameter in the above formula are as follows: This represents the maximum a posteriori estimated parameter set, i.e., the true values ​​of the parameters after calibration with measured data; , Represents the system's global mass matrix and fundamental stiffness matrix, referenced from the preceding module; This represents the modified Rayleigh damping coefficient function; This represents the stiffness correction factor function, used to compensate for stiffness deviations caused by the base installation process; This represents the Rayleigh damping matrix of the reconstructed system structure. This represents the reconstructed physical stiffness tensor of the vibration isolation base, used to update the parameterized generalized coupled stiffness matrix. ; This indicates an assignment / update operation.

[0079] In this implementation, the formula transforms the abstract statistical inference results into specific physical matrix correction instructions, ensuring that the mathematical model within the coupled dynamics modeling module remains synchronized with the real state of the physical world, so that the next design iteration will be launched based on a physical model that is closer to reality.

[0080] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0081] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0082] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0083] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0084] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An intelligent design system for EPC general contracting of marine HVAC systems, characterized in that, include: The system comprises a coupled dynamics modeling module, a response prediction model generation module, a layout optimization decision-making module, and a parameter closed-loop calibration module. The coupled dynamics modeling module, in response to the input of hull structure design parameters and HVAC equipment physical parameters, constructs parameterized ship-machine coupled dynamics equations that characterize the interaction between the hull structure and HVAC equipment; The response prediction model generation module obtains the parameterized ship-engine coupled dynamics equations, uses the parameterized ship-engine coupled dynamics equations as physical constraints to train the neural network, and generates a physical information vibration prediction proxy model that can predict structural vibration response data based on equipment position variables. The layout optimization decision module calls the physical information vibration prediction proxy model to calculate the structural vibration response value corresponding to different equipment installation positions, and selects the HVAC equipment optimization layout scheme that meets the preset vibration fatigue requirements based on the structural vibration response value. The parameter closed-loop calibration module corrects the model parameters in the parameterized ship-machine coupled dynamics equations based on the deviation between the measured data of the same type of ship and the output data of the physical information vibration prediction proxy model, and feeds the corrected model parameters back to the coupled dynamics modeling module.

2. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 1, characterized in that: The specific steps for constructing parameterized ship-engine coupled dynamics equations using the coupled dynamics modeling module include: First, modal synthesis and dimensionality reduction characterization of the hull structural domain are performed. The system identifies interface nodes and internal nodes in the physical model of the hull structure based on the input hull structure design parameters. By solving the constraint principal modes and fixed interface principal modes, a modal transformation matrix is ​​constructed. The modal transformation matrix is ​​then used to perform orthogonal projection transformation on the hull structure mass matrix and hull structure stiffness matrix in the physical coordinate system to generate the hull generalized modal mass matrix and hull generalized modal stiffness matrix. Secondly, the system constructs a parameterized position-dependent coupling operator. It defines the HVAC equipment installation position vector as a variable design parameter, introduces a finite element shape function matrix and a numerical approximation distribution function to describe the spatial interpolation relationship between the HVAC equipment base and the continuous structural domain of the hull, calculates the physical coupling effect generated by the physical stiffness tensor of the HVAC equipment base at the current installation position, and uses a modal transformation matrix to project the physical coupling effect onto the generalized modal space to generate a parameterized generalized coupling stiffness matrix that changes in real time with the HVAC equipment installation position vector.

3. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 2, characterized in that: The specific steps for constructing parameterized ship-engine coupled dynamics equations using the coupled dynamics modeling module also include: Finally, the global dynamic equations in the hybrid coordinate system are assembled. The system establishes a hybrid coordinate system containing the generalized modal coordinate vector of the hull and the physical displacement vector of the HVAC equipment. An antisymmetric gyroscope matrix characterizing the Coriolis force effect of the high-speed rotating components of the HVAC equipment is constructed. The dimensional mapping matrix is ​​used to align and assemble the generalized modal mass matrix of the hull, the generalized modal stiffness matrix of the hull, the inertial matrix of the HVAC equipment extracted based on the physical parameters of the HVAC equipment, and the parameterized generalized coupling stiffness matrix. A set of differential-algebraic equations that can describe the dynamic response of the system under the action of the generalized external load vector is constructed, which is the parameterized ship-machine coupled dynamic equation.

4. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 3, characterized in that: The specific steps of the response prediction model generation module in generating the physical information vibration prediction surrogate model include: First, the parameterized spatiotemporal manifold neural mapping architecture is constructed. The system establishes a deep neural network architecture, using the time independent variable and the HVAC equipment installation position vector as the input layer data of the deep neural network. A nonlinear mapping function from the input layer to the output layer is constructed, and the output layer of the deep neural network is configured to directly predict the system's generalized state prediction vector. The system's generalized state prediction vector includes the ship's generalized modal coordinate vector and the HVAC equipment's physical displacement vector. Secondly, the system performs automatic differential of the physical residuals by embedding parameterized coupling operators. The system uses automatic differential techniques to calculate the first and second derivatives of the generalized state prediction vector of the system with respect to the time independent variable. It extracts the system global mass matrix, system structural Rayleigh damping matrix, antisymmetric gyroscope matrix, and system basic stiffness matrix that constitute the parameterized ship-engine coupled dynamic equations. The parameterized generalized coupling stiffness matrix is ​​explicitly embedded as an interface to characterize the influence of position variables on the physical field. Through linear combination, a dynamic physical residual operator that can quantify the degree to which the current prediction value violates the dynamic equilibrium law is constructed.

5. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 4, characterized in that: The specific steps of the response prediction model generation module in generating the physical information vibration prediction surrogate model also include: Finally, the loss function of physical constraint and data-driven fusion is minimized. The system randomly samples physical configuration points in the spatiotemporal domain and calculates the norm of the dynamic physical residual operator to construct the physical residual loss term. At the same time, a small number of simulation ground truth samples calculated based on finite element analysis are obtained, and the deviation between the generalized state prediction vector of the system and the simulation ground truth samples is calculated to construct the data observation loss term. The global training objective function is constructed by weighted summation of the physical residual loss term and the data observation loss term. The global training objective function is minimized by gradient-based optimization algorithm to iteratively update the weights and bias parameters of the deep neural network until the global training objective function converges, thus generating the physical information vibration prediction surrogate model.

6. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 5, characterized in that: The specific steps by which the layout optimization decision module generates an optimized layout plan for HVAC equipment include: First, the system performs a global structural acoustic intensity vector field reconstruction. Within the preset feasible design domain, the system performs a discretized grid scan on the HVAC equipment installation location vector. For each candidate location generated by the scan point, the system calls the physical information vibration prediction proxy model to obtain the generalized state prediction vector of the system in the time domain. The system generalized state prediction vector is restored to the physical stress field and velocity field using the finite element shape function derivative operator. The instantaneous structural acoustic intensity vector is calculated based on the physical stress field and velocity field and then time-averaged to construct the structural acoustic intensity vector field. Finally, the divergence distribution of the structural acoustic intensity vector field is calculated to identify the energy source region and energy sink region in the hull structure. Secondly, the system performs a non-stationary cumulative fatigue damage assessment. Based on the stress history data derived from the system's generalized state prediction vector, the system uses the continuous domain approximation of the rainflow counting method and combines it with the material's SN curve parameters to calculate the non-stationary fatigue damage index for predefined key fatigue hotspots in the hull structure. By operating on the stress concentration correction factor and the Von Mises equivalent stress mapping operator, the system generates a global cumulative fatigue damage index that can quantify the degree of irreversible damage to the key hull structure if the HVAC equipment were installed in the current location.

7. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 6, characterized in that: The specific steps involved in generating an optimized HVAC equipment layout plan by the layout optimization decision module also include: Finally, a multi-objective potential field topology optimization decision is performed, and a multi-physics comprehensive evaluation potential function is constructed. This function is formed by weighted fusion of the divergence distribution of the structural acoustic intensity vector field, the global cumulative fatigue damage index, and the geometric constraint penalty term characterizing the geometric dimensions and maintenance space limitations of the HVAC equipment. The global minimum point of the multi-physics comprehensive evaluation potential function is searched within the feasible design domain using a hybrid simulated annealing strategy, and the HVAC equipment installation location vector that minimizes the multi-physics comprehensive evaluation potential function is determined as the optimal layout scheme for the HVAC equipment.

8. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 7, characterized in that: The specific steps for the parameter closed-loop calibration module to correct the model parameters in the parameterized ship-engine coupled dynamics equations include: Heterogeneous data mapping and random likelihood function construction are performed. The system acquires measured vibration response time-series data of the target type of ship under typical working conditions, defines a sparse observation mapping matrix, and uses the sparse observation mapping matrix to project the system generalized state prediction vector output by the physical information vibration prediction proxy model onto the physical coordinate space corresponding to the measured vibration response time-series data. The uncertainty parameter set of the model to be calibrated, which constitutes the parameterized ship-machine coupled dynamic equation, is identified. The deviation between the measured vibration response time-series data and the projected prediction data is calculated. Combined with the prediction error covariance matrix, which characterizes the statistical characteristics of measurement noise and model structural error, a log-likelihood function is constructed to evaluate the probability ability of the current parameter combination to interpret the measured data.

9. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 8, characterized in that: The specific steps for the parameter closed-loop calibration module to correct the model parameters in the parameterized ship-engine coupled dynamics equations also include: To perform posterior parameter inference based on Markov chain Monte Carlo, the system sets the prior probability distribution of the parameter set of the uncertainty parameter set of the model to be calibrated, combines the log-likelihood function, and uses the Markov chain Monte Carlo sampling algorithm to extract samples from the joint probability distribution to approximate the posterior probability density function of the parameters. By statistically analyzing a large number of posterior samples, the maximum a posteriori estimated parameter set is extracted, and the maximum a posteriori estimated parameter set is defined as the corrected model physical parameters.

10. The intelligent design system for EPC general contracting of marine HVAC systems according to claim 9, characterized in that: The specific steps for the parameter closed-loop calibration module to correct the model parameters in the parameterized ship-engine coupled dynamics equations also include: After reconstructing the dynamic baseline equations and backfeeding knowledge, the system feeds back the corrected model physical parameters to the coupled dynamic modeling module. This triggers the coupled dynamic modeling module to use the corrected model physical parameters to numerically update and reconstruct the system structure Rayleigh damping matrix and the parameterized generalized coupled stiffness matrix that constitute the parameterized ship-engine coupled dynamic equations, thereby completing the closed-loop calibration of the physical knowledge of the intelligent design system.

Images