Ivan Voutchkov

Multi-Objective Optimization Using Surrogates

Until recently, optimization was regarded as a discipline of rather theoretical interest, with limited real-life applicability due to the computational or experimental expense involved. Practical multiobjective optimization was considered almost as a utopia even in academic studies due to the multiplication of this expense. This paper discusses the idea of using surrogate models for multiobjective optimization. With recent advances in grid and parallel computing more companies are buying inexpensive computing clusters that work in parallel. This allows, for example, efficient fusion of surrogates and finite element models into a multiobjective optimization cycle. The research presented here demonstrates this idea using several response surface methods on a pre-selected set of test functions. We aim to show that there are number of techniques which can be used to tackle difficult problems and we also demonstrate that a careful choice of response surface methods is important when carrying out surrogate assisted multiobjective search.

On the thermo-mechanical events during friction surfacing of high speed steels

This paper is concerned with the friction surfacing of high-speed steels, BM2, BT15 and ASP30 onto plain carbon steel plate. The events that the matrix and carbides experience as the coating material pass from the coating rod to the substrate, in forming the coating, is described. The coating is observed to harden automatically within a few seconds of being deposited onto the cold substrate. This autohardening is observed to be an inherent feature of the friction surfacing process and the only post-coating heat treatment required is tempering, as with traditionally hardened high-speed steels. The mechanism of autohardening is discussed in terms of the mechtrode/coating/substrate thermal system.

An integrated approach to friction surfacing process optimisation

This paper discusses the procedures for data collection, management and optimisation of the friction surfacing process. Experimental set-up and characteristics of measuring equipment are found to match the requirements for accurate and unbiased data signals. The main friction surfacing parameters are identified and the first stage of the optimisation process is achieved by visually assessing the coatings and introducing the substrate speed vs. force map. The optimum values from this first stage forms a region around the middle of a trapezium-shaped area whose borders are found experimentally. Data collected for the second stage were analysed using the least squares method which were applied to find the coefficients of a second order regression model. Advantages of applying artificial intelligence methods to friction surfacing modelling are also described and the higher accuracy achieved using neural networks demonstrated.

Neurofuzzy approach to process parameter selection for friction surfacing applications

Friction surfacing is an advanced manufacturing process, which has been successfully developed and commercialised over the past decade. The process is used for corrosion and wear resistant coatings and for reclamation of worn engineering components. At present, the selection of process parameters for new coating materials or substrate geometries experimentally requires lengthy development work. The major requirement is for the flexibility to enable rapid changes of process parameters in order to develop new applications, with variations of materials and geometries in a cost effective and reliable manner. Further improvement requires development of appropriate mathematical models of the process, which will facilitate the introduction of optimisation techniques for efficient experimental work as well as the introduction of real time feedback adaptive control. This paper considers the use of combined artificial intelligence and modelling techniques. It includes a new frame of a Neurofuzzy-model based Decision Support System — FricExpert, which is aimed at speeding up the parameter selection process and to assist in obtaining values for cost effective development. Derived models can then be readily used for optimisation techniques, discussed in our earlier work.

Multiobjective Tuning of Grid-Enabled Earth System Models Using a Non-dominated Sorting Genetic Algorithm (NSGA-II)

The tuning of parameters in climate models is essential to provide reliable long-term forecasts of Earth system behaviour. In this paper we present the first application of the multiobjective non-dominated sorting genetic algorithm (NSGA-II) to the GENIE-1 Earth System Model (ESM). Twelve model parameters are tuned to improve four objective measures of fitness to observational data. Grid computing and data handling technology is harnessed to perform the concurrent simulations that comprise the generations of the genetic algorithm. Recent advances in the method exploit Response Surface Modelling to provide surrogate models of each objective. This enables more extensive and efficient searching of the design space. We assess the performance of the NSGA-II using surrogates by repeating a tuning exercise that has been performed using a proximal analytical centre plane cutting method and the Ensemble Kalman Filter on the GENIE-1 model. We find that the multiobjective algorithm locates Pareto-optimal solutions which are of comparable quality to those obtained using the single objective optimisation methods.

Multi-objective optimization of GENIE Earth system models

The tuning of parameters in climate models is essential to provide reliable long-term forecasts of Earth system behaviour. We apply a multi-objective optimization algorithm to the problem of parameter estimation in climate models. This optimization process involves the iterative evaluation of response surface models (RSMs), followed by the execution of multiple Earth system simulations. These computations require an infrastructure that provides high-performance computing for building and searching the RSMs and high-throughput computing for the concurrent evaluation of a large number of models. Grid computing technology is therefore essential to make this algorithm practical for members of the GENIE project.

Robust structural design of a simplified jet engine model, using multiobjective optimization

This study demonstrates advances in multiobjective optimization, supporting a robustness study of a simplified jet engine structural model. The ultimate goal is to find the best structural configuration of shell thicknesses along the engine that will result in reduced reaction force variation under a range of external loads, will be as light as possible and where fuel consumption will be minimal. These are competitive objectives, some of which are of stochastic rather than deterministic in nature. The paper demonstrates that a deep level multiobjective search pays off many times the investment in time and money by providing significant design improvement.

(Re-) Meshing using interpolative mapping and control point optimization

Abstract This work proposes a simple and fast approach for re-meshing the surfaces of smooth-featured geometries prior to CFD analysis. The aim is to improve mesh quality and thus the convergence and accuracy of the CFD analysis. The method is based on constructing an interpolant based on the geometry shape and then mapping a regular rectangular grid to the shape of the original geometry using that interpolant. Depending on the selected interpolation algorithm the process takes from less than a second to several minutes. The main interpolant discussed in this article is a Radial Basis Function with cubic spline basis, however other algorithms are also compared. The mesh can be optimized further using active (flexible) control points and optimization algorithms. A range of objective functions are discussed and demonstrated. The difference between re-interpolated and original meshes produces a metric function which is indicative of the mesh quality. It is shown that the method works for flat 2D surfaces, 3D surfaces and volumes.

Model-Based Robust Parametric Design of Automatic Cleaning Process

The goal of pharmaceutical industry is to manufacture products that meet patients’ needs and expectations, while satisfying the regulatory requirements. The products need to meet the required quality and purity characteristics that are represented to possess. Therefore, in a multi-product manufacturing facility, appropriately designed cleaning processes are essential to avoid cross-contamination between products and ensure patients’ health and safety. The latest trend in the development of cleaning validation is using quality by design methodology (QbD) to determine the most appropriate parameters of the cleaning processes that will reduce the risks of cross-contamination. The present study highlights the model-based approach for robust engineering design in order to achieve an efficient, reliable, and cost-effective cleaning process simultaneously.

Fast design optimization of jet engine structural mass and specific fuel consumption

With the increasing demands placed on modern jet engine performance and reliability, there comes the necessity for efficient optimization tools that engineers can readily utilize on their desks. This article discusses the application of a multi-objective optimization tool to the design process for a key structural element in the highly respected Rolls–Royce Trent jet engine series. The ability of this structure to maintain tightly controlled blade to casing tip clearances plays a significant role in maintaining the fuel efficiency of the engine. The design tool described here has been implemented on a conventional desktop computer and allows efficient trade-off analysis of structural mass versus specific fuel consumption for four-load cases.