Thermal-fluid subsystem digital twin

Abstract

A system and method for generating and integrating surrogate neural network models developed from physical models of several subsystems on a vehicle, such as an aircraft, and simulating the integration of the models in a virtual flight deck (VFD), the method including the steps of modeling each of the subsystems as a physical model, developing a surrogate neural network model from each of the physical models, placing the surrogate neural network models on a common bus, and integrating some or all of the surrogate neural network models provided on the bus in the VFD.

Description

[Technical Field]

[0001]

[0001] The present disclosure relates generally to systems and methods for generating and integrating surrogate neural network models developed from physical models of several vehicle subsystems, and more particularly to systems and methods for generating and integrating surrogate neural network models developed from physical models of several subsystems on an aircraft and simulating the integration of the models in a virtual flight deck (VFD). [Background technology]

[0002]

[0002] Modern aircraft design and testing employ model-based engineering and system simulation. Model-based design captures and tracks aircraft requirements early in the design process, then documents the requirements and continues to check for compliance. The modeling process separately models various subsystems on the aircraft, such as the fuel subsystem, avionics subsystem, hydraulic subsystem, and environmental control subsystem (ECS).

[0003] Known model-based design simulators have several drawbacks. For example, known model-based design simulators separate subsystems, and data from subsystems is typically shared manually and rarely integrated, resulting in a weak architecture with fixed connections between components. Furthermore, known simulators are computationally intensive, limited, and sometimes cumbersome. Simulation and analysis are based on discrete point sampling rather than continuous curves, and are therefore susceptible to large interpolation errors or rely on experimental data from test flight phases. Furthermore, current model-based design simulators do not allow for practical co-simulation of more than a limited set of subsystems within a physics-based platform. Currently, software exists for modeling physics-based subsystems, and these models can be scaled up to include multiple domains using, for example, Matlab, AMEsim, and Flowmaster. However, when simulating high-frequency models, such as hydraulic drive subsystems with low-frequency ECSs, execution times become unacceptable. Currently, there is no way to overcome the scaling issue of integrating all thermal-fluid subsystems into a single model in model-based aircraft design. Summary of the Invention [Means for solving the problem]

[0004] The following description discloses and describes a system and method for generating and integrating surrogate neural network models developed from physical models of several subsystems on a vehicle, such as an aircraft, and simulating the integration of the models in a virtual flight deck (VFD). The method includes the steps of modeling each of the subsystems as a physical model, developing a surrogate neural network model from each of the physical models, placing the surrogate neural network models on a common bus, and integrating some or all of the surrogate neural network models provided on the bus in the VFD.

[0005]

[0005] Further features of the present disclosure will become apparent from the following description and appended claims, taken in conjunction with the accompanying drawings. [Brief explanation of the drawings]

[0006] [Figure 1]

[0006] FIG. 1 is a block diagram of a system that illustrates a platform for designing, simulating, and testing a vehicle, including a digital twin that generates and integrates surrogate neural network models of vehicle subsystems. [Figure 2]

[0007] FIG. 1 is a block diagram of a digital twin showing the integration of a surrogate neural network model with a virtual flight deck (VFD). [Figure 3]

[0008] A diagram of a neural network model generated by a digital twin. DETAILED DESCRIPTION OF THE INVENTION

[0007]

[0009] The following description of embodiments of the present disclosure directed to systems and methods for generating and integrating surrogate neural network models developed from physical models of several subsystems on an aircraft and simulating the integration of the models in a virtual flight deck (VFD) is merely exemplary and is not intended to limit the disclosure or its application or uses. For example, the vehicle modeled and simulated described herein is an aircraft. However, the systems and methods are applicable to other vehicles.

[0008]

[0010] As described in detail below, the present disclosure proposes a system and method for virtual simulation of vehicles, such as aircraft, that employs a common data bus architecture to share information between models of on-vehicle subsystems. Each on-vehicle subsystem is developed and modeled using classical physics modeling techniques, including network 1D bulk-averaged physics-based approximations. Once the subsystem models are created, each subsystem is analyzed in a standalone model or "trainer." The standalone model is simulated using random inputs through machine learning until the neural network model achieves the same output response as the physics-based model. The neural network model is then used in place of the physics-based model in vehicle-level simulations. This allows for simultaneous co-simulation of all thermal-fluid subsystems within the vehicle model and observation of interactions between the subsystems.

[0009]

[0011] A physical model is a software structure that represents the movement of an object. A neural network is a software structure that can learn to perform a task by processing examples, without being programmed with task-specific rules. Neural networks generally contain neurons or nodes, each with a "weight" that is multiplied by the input to the node to obtain a probability of whether something is correct. More specifically, each node has a weight, which is a floating-point number that is multiplied by the input to the node to generate an output for that node that corresponds to a certain percentage of the input. The weights are "trained," or set, by first having the neural network analyze a known data set under supervision and minimizing a cost function to ensure that the network obtains the highest probability of a correct output. Neural networks often contain several layers of nodes that perform nonlinear processing, with each successive layer receiving the output from the previous layer. Typically, the layers include an input layer that receives raw data from sensors, multiple hidden layers that extract abstract features from the data, and an output layer that identifies specific objects based on feature extraction from the hidden layers. One common type of neural network is known as a convolutional neural network (CNN), which is a type of feedforward neural network that can be used to model data associated with input data that has a grid-like topology.

[0010]

[0012] As used herein, a machine learning program, machine learning algorithm, or machine learning module generally refers to a type of artificial intelligence that includes one or more algorithms capable of learning and / or adjusting parameters based on input data provided to the algorithm. In some cases, machine learning programs, machine learning algorithms, and machine learning modules are used at least in part in implementing artificial intelligence (AI) functions, systems, and methods. Machine learning programs may be configured to perform stored operations such as decision tree learning, association rule learning, artificial neural networks, recurrent artificial neural networks, long short-term memory networks, inductive logic programming, support vector machines, clustering, Bayesian networks, reinforcement learning, representation learning, similarity and metric learning, sparse dictionary learning, genetic algorithms, k-nearest neighbors (KNN), and the like. One type of algorithm suitable for use in the machine learning modules described herein is an artificial neural network or neural network inspired by biological neural networks. An artificial neural network can learn to perform a task by processing examples without being programmed with task-specific rules. The artificial intelligence systems and structures described herein may employ deep learning. Deep learning is a specific type of machine learning that achieves higher learning performance by representing a specific real-world environment as a hierarchy of concepts of increasing complexity. Deep learning typically employs a software structure with multiple layers of neural networks that perform nonlinear processing, with each successive layer receiving the output from the previous layer.

[0011]

[0013] FIG. 1 is a block diagram of system 10 illustrating a platform for designing, simulating, and testing a vehicle. In this description, the vehicle is an aircraft. However, system 10 is also applicable to designing, simulating, and testing other types of vehicles, such as land and marine vehicles. A simulated integration of various subsystems on the vehicle is developed in digital twin 12. In box 14, various vehicle parameters, requirements, capabilities, etc. are identified, for example, by a customer, and, based on those parameters, a process is performed to generate a conceptual design of the vehicle in box 16. In box 18, various CAD designs for the configuration and layout of the vehicle and its subsystems are generated, and in box 20, the vehicle's aerodynamic and propulsion configuration and characteristics are modeled. The vehicle's conceptual design, CAD design, and aerodynamic and propulsion model are provided to digital twin 12 for integration and simulation, and the simulation provides feedback to the CAD design and aerodynamic and propulsion model via a suitable protocol, such as model-based engineering (MBE). MBE refers to the use of analysis and simulation tools throughout the product lifecycle to reduce the use of physical prototypes. As will be explained in more detail below, through the use of neural networks and physics models, different models of different subsystems are integrated in the digital twin 12 in a less comprehensive manner.

[0012]

[0014] FIG. 2 is a block diagram of digital twin 12 showing various virtual subsystems that are part of a digital aircraft 40 being designed and simulated on a virtual flight deck (VFD) 42. Each subsystem is first modeled as a physics-based model, which is then used to develop and train a surrogate neural network model through machine learning. The physics model and / or surrogate neural network model are then integrated by VFD 42. In this non-limiting example, the subsystems include hydraulic subsystem 44 modeled by physics model 46, heat exchanger subsystem 48 modeled by surrogate neural network model 50, fuel subsystem 52 modeled by physics model 54, RAM air subsystem 56 modeled by surrogate neural network model 58, ECS 60 modeled by physics model 62 or surrogate neural network model 64, environmental subsystem 66 modeled by surrogate neural network model 68, propulsion subsystem (engine) 70 modeled by physics model 72, and aerodynamic subsystem 74 modeled by surrogate neural network model 76. The arrows between subsystems 44 , 48 , 52 , 56 , 60 , 66 , 70 , and 74 represent a common bus architecture that provides interconnection and integration of the various models simulated in VFD 42 .

[0013]

[0015] 3 is a neural network 80 illustrating an example of a surrogate neural network model generated by VFD 42 that is an integration of some or all of the models of subsystems 44, 48, 52, 56, 60, 66, 70, and 74 described above. Neural network 80 includes an input layer 82, an output layer 84, and a hidden layer 86 therebetween, with layers 82, 84, and 86 including nodes 88. Integrating the various vehicle subsystem models 46, 50, 54, 58, 62, 64, 68, 72, and 76 in the VFD 42 as described provides scalable interconnection between the vehicle subsystems 44, 48, 52, 56, 60, 66, 70, and 74, enables simulation of the subsystems 44, 48, 52, 56, 60, 66, 70, and 74 using machine learning to enable scalable integration, enables integration of all thermal-fluid vehicle subsystems, aerodynamic subsystems, and propulsion subsystems while maintaining physics-based fidelity, and integrates mission-level analysis with detailed subsystem analysis.

[0014]

[0016] Returning to FIG. 1 , all of the various surrogate neural network models of vehicle subsystems 44, 48, 52, 56, 60, 66, 70, and 74 provided in digital twin 12 are stored in repository 22 for use in other vehicles and other design platforms as needed. For example, the surrogate neural network models can be provided to flight simulator 24, which may interact with a mission analysis process in box 26, which may be controlled by various vehicle parameters, requirements, and capabilities in box 14. Furthermore, the surrogate neural network models may be employed in an experimental environment in box 28, where tests may be performed on various vehicle subsystems and software. The surrogate neural network models may also be deployed in the vehicle itself at the time the vehicle is manufactured in box 30, so that simulation data can be compared to the vehicle's actual performance to compare what should happen with what is actually happening. The surrogate neural network models may also be deployed in future vehicles in box 32.

[0015]

[0017] The foregoing description discloses and describes merely exemplary embodiments of the present disclosure. Those skilled in the art will readily appreciate from such description, and from the accompanying drawings and claims, that various changes, modifications, and variations can be made without departing from the spirit and scope of the present disclosure, as defined in the following claims.

Claims

1. 1. A method for virtually designing and testing a vehicle, the vehicle including a plurality of vehicle subsystems, the method comprising: modeling each of the subsystems as a physical model; developing a surrogate neural network model from each of said physical models; placing the surrogate neural network models on a common bus; integrating some or all of the surrogate neural network models provided on the bus; A method comprising:

2. The method of claim 1 , wherein integrating some or all of the surrogate neural network model comprises simulating the subsystem in a virtual simulator.

3. The method of claim 1 , wherein the vehicle is an aircraft.

4. The method of claim 3 , wherein the plurality of vehicle subsystems includes a hydraulic subsystem, a heat exchanger subsystem, a fuel subsystem, a RAM air subsystem, an environmental control subsystem (ECS), an environmental subsystem, a propulsion subsystem, and an aerodynamic subsystem.

5. 4. The method of claim 3, wherein integrating some or all of the surrogate neural network model comprises integrating the surrogate neural network in a virtual flight deck (VFD) and simulating the subsystem in the VFD.

6. The method of claim 1 , further comprising storing the surrogate neural network model in a repository for use by other systems.

7. The method of claim 6 , wherein the other systems include one or more of a flight simulator, a mission analysis processor, a laboratory environment, and an actual vehicle.

8. 1. A method for virtually designing and testing an aircraft, the aircraft including a plurality of subsystems, the method comprising: modeling each of the subsystems as a physical model; developing a surrogate neural network model from each of said physical models; placing the surrogate neural network models on a common bus; integrating some or all of the surrogate neural network model provided on the bus in a virtual flight deck (VFD); simulating the subsystem in the VFD; A method comprising:

9. The method of claim 8 , wherein the plurality of subsystems comprises a hydraulic subsystem, a heat exchanger subsystem, a fuel subsystem, a RAM air subsystem, an environmental control subsystem (ECS), an environmental subsystem, a propulsion subsystem, and an aerodynamic subsystem.

10. The method of claim 8 , further comprising storing the surrogate neural network model in a repository for use by other systems.

11. The method of claim 10 , wherein the other systems include one or more of a flight simulator, a mission analysis processor, a laboratory environment, and an actual aircraft.

12. 1. A system for virtually designing and testing a vehicle, the vehicle including a plurality of vehicle subsystems, the system comprising: means for modeling each of said subsystems as a physical model; means for developing a surrogate neural network model from each of said physical models; means for placing the surrogate neural network models on a common bus; means for integrating some or all of the surrogate neural network models provided on the bus; Including, the system.

13. The system of claim 12 , wherein the means for integrating some or all of the surrogate neural network model simulates the subsystem in a virtual simulator.

14. The system of claim 12 , wherein the vehicle is an aircraft.

15. 15. The system of claim 14, wherein the plurality of vehicle subsystems comprises a hydraulic subsystem, a heat exchanger subsystem, a fuel subsystem, a RAM air subsystem, an environmental control subsystem (ECS), an environmental subsystem, a propulsion subsystem, and an aerodynamic subsystem.

16. 15. The system of claim 14, wherein the means for integrating some or all of the surrogate neural network model integrates the surrogate neural network in a virtual flight deck (VFD) and simulates the subsystem in the VFD.

17. The system of claim 12 , further comprising means for storing the surrogate neural network model in a repository for use by other systems.

18. The system of claim 17 , wherein the other systems include one or more of a flight simulator, a mission analysis processor, a laboratory environment, and an actual vehicle.

Images