Search

Application of Advance Surrogate Models For Prop[ulsion Plant Performance and Fuel Management

Machine Design Algroithm using multi point design methodology


Benefits of using advanced surrogate models

The propulsion plant system as a whole is multidisciplinary engineering in nature. It is concerned with the aero-thermodynamic, aeromechanical, systems, operational and control aspects of the installation, an important feature of which is the design and performance of the overall power plant or propulsion system. This comprises the engine, air intake, exhaust system, and the nacelle or housing surrounding the engine which is quite challenging and constrained effort by means of classical mechanical system.


Partial Differential Equations (PDE) are used to represent a wide variety of physical phenomena such as aerodynamic flows and machine dynamic, are numerically solved by approaches such as the Finite Difference Method (FDM), Finite Volume Method (FVM) or Finite Element Method (FEM). Any of them relies on discretizing the space domain in which the equation is solved by means of a mesh or grid. The solution is then approximated in a discrete set of time instants. Subsequently, the continuous mathematical operators are also approximated using the so-called numerical schemes. Such methods have been deeply studied and developed, but their accuracy and robustness strongly rely on the refinement level of the discretization.


New reliability based multi design point methodologies, adopted using unconventional approaches of engaging reliability prediction by Artificial Neural Network, surrogate models, can also make use of new computational hardware such as neural chips and quantum annealers

.

The key to improving computational efficiency is to avoid performing the time-consuming probabilistic analysis by thermodynamic-based aero engine simulation model. Artificial neural network (ANN) is utilized to establish surrogate models to predict the aero engine performance reliability under multiple flight conditions.

Challenges of using advanced surrogate models

A key feature of this algorithm is that no dataset neither mesh is required for training the neural network. Instead, only a subset of points randomly sampled from the time and space domains are needed to assess whether the neural network output fits the PDE, its initial condition and its boundary conditions.


Multidisciplinary design optimization is an ongoing challenge in the aerospace industry, resulting in long design lead times and untapped optimisation potential. Quantum computing may offer a viable path towards efficient multi-parameter optimization covering the entire design space. Here we ask for the application ofquantum computing solutions to a problem involving airframe loads, mass modelling and structural analysis.


The target is to preserve structural integrity while optimizing propulsion performance. Weight optimisation is key to low operating costs and reduced environmental impact. The challenge arises when computing a broad range of aircraft design configurations simultaneously which is currently not possible with classical computing.


These methods are usually implemented in standard machine learning libraries such as TensorFlow Quantum. In particular, this library also provides tools to automatically differentiate the neural network, making the computation of the network spatial and time derivatives straightforward.


Implementation of advanced surrogate models

This approach is based on classical systems extension to quantum states. The proposed theory can be applied to analysis of multiple (including non- Hamiltonian) dissipative dynamical systems. Developed methods and algorithms integrated in quantum simulators will allow us to solve a wide range of problems with scientific and practical significance The reliability-based multi-design point methodology is proposed to acquire the appropriate key design parameters by comprehensive reliability analysis for multiple operating conditions of interest.


This methodology establishes corresponding thermodynamics-based simulation model of aircraft engine with uncertainty component performance;

It divides the whole flight profile into multiple critical operating conditions and simultaneously regard them as the on-design points, and then respectively determine their overall performance requirements;


This methodology generates a certain number of training scenarios and train the neural network for establishing reliability prediction surrogate models of each concerned operating condition;


It constructs objective function by the attained surrogate models to facilitate the comprehensive performance reliability analysis in optimization design.Simultaneous consideration of mutually complementary coordinate and momentum portraits provides a deeper understanding of the propulsion plant dynamical system chaotic behavior.


Implementation of advanced surrogate models:

Direct applications of ANN surrogate modeling in aerodynamic design


ANN is used in many applications of surrogate models due to its huge convenience available for problems with large amountof data. The principle behind surrogate modeling is that data at input and output is related through the pattern of the trained neural network


This methodology establishes corresponding thermodynamics-based simulation model of aircraft engine with uncertainty component performance;


Turbomachinery uncertainty analysis requires performing a large number of simulations, the computational cost of which can be greatly alleviated with ANN surrogate modelling


Its predictive capability is shown insensitive to numerical instabilities and convergence difficulties typically associated with computational processes.


This methodology generates a certain number of training scenarios and train the neural network for establishing reliability prediction surrogate models of each concerned operating condition;


Surrogate models with ANN have been shown a good alternative toconventional solution with regard to the prediction of aerodynamic coefficients of airplanes of high accuracy.


Advantages of using advanced surrogate models


ANN is used in many applications of surrogate models due to its huge

convenience available for problems with large amountof data. It respectively reduced the average error and maximum error of Artificial Neural Network.


Utilization of the ANN surrogate models facilitate the reliability-based cycle design optimization, which replaces the time-consuming probabilistic analysis based on Monte Carlo simulation


Optimization design solution of presented methodology reasonably increases the aero engine performance redundancy to precisely reach the expected reliability of all concerned operating conditions.


This methodology is universal and can be easily applied to other types of propulsion plant.

It facilitates the application of the reliability-based multi-design point method for candidates of future aviation propulsion, which are noted for excellent multi-mission adaptability.


Surrogate models with ANN have been shown a good alternative to conventional solution with regard to the prediction of aerodynamic


Coefficients of airplanes of high accuracy. ANN has been used for space mapping for transonic air foil aerodynamic shape optimization


Further remarks using advanced surrogate models


ANN surrogate modelling offers a feasible approach to aerodynamic design in the field of rapid investigation of design space and optimal solution searching. The direct motive of a surrogate model is to establish a data-driven and bottom-up approach used when an outcome of interest cannot be easily obtained.

The principle behind surrogate modeling is that data at input and output are

related through the pattern of the trained neural network.

It is noted that ANN needs a certain data set to be effectively trained. In particular, new forms of geometric representation method are proposed specifically to the network training.

ANN not only performs well in surrogate modeling, but also produces potential advantages in optimization. The utilization of various levels of information, the multilevel surrogate modeling is proposed to obtain the optimal area in the design space quickly thus saving computational resource.

There are many types of combination of optimization algorithms with ANN. Lastly, deep learning is beneficial to the field of aerodynamic design surrogate model because it no longer needs the a priori treatment to the input data and thus human experience can be relied less.

Advanced deep learning techniques (deep belief network, deep reinforcement learning, etc.) shouldalso be considered to be further alternativesto accommodate development of high-precision surrogate model






.

8 views0 comments