Recent advances in engineering design optimisation: Challenges and future trends

Traditional engineering design optimisation which is the process of identifying the right combination of product parameters is often done manually, time consuming and involves a step by step approach. This paper identifies recent approaches to automating the manual optimisation process and the challenges that it presents to the engineering community. Engineering design optimisation is classified based on design evaluation effort and degrees of freedom viewpoints. An overview of different approaches for design optimisation is presented. The study identifies scalability as the major challenge for design optimisation techniques. Large-scale optimisation requires significant computing power and efficient algorithms such as swarm intelligence.     

A probabilistic approach for the optimisation of ultrasonic array inspection techniques

Ultrasonic arrays are now used routinely for the inspection of engineering structures in order to maintain their integrity and assess their performance. Such inspections are usually optimised manually using empirical measurements and parametric studies which are laborious, time-consuming, and may not result in an optimal approach. In this paper, a general framework for the optimisation of ultrasonic array inspection techniques in NDE is presented. Defect detection rate is set as the main inspection objective and used to assess the performance of the optimisation framework. Statistical modelling of the inspection is used to form the optimisation problem and incorporate inspection uncertainty such as crack type and location, material properties and geometry, etc. A genetic algorithm is used to solve the global optimisation problem. As a demonstration, the optimisation framework is used with two objective functions based on array signal amplitude and signal-to-noise ratio (SNR). The optimal use of plane B-scan and total focusing method imaging algorithms is also investigated. The performance of the optimisation scheme is explored in simulation and then validated experimentally. It has been found that, for the inspection scenarios considered, TFM provides better detectability in a statistical sense than plane B-scan imaging in scenarios where uncertainty in the inspection is expected.     

Robustification of CNC Controllers for Machine Tools Motor Drives

This paper presents a methodology for enhancing the robustness of a CNC controlled system by convex optimisation of the Youla parameter. The methodology requires as a first step the design of an initial polynomial controller. This controller is then robustifled considering temporal and frequency constraints, which are formulated by means of the Youla parameterisation within a convex optimisation framework. The optimal Youla parameter is finally obtained by solving this optimisation problem. In this way, a compromise between robustness and closed loop behaviour can be easily managed. An application to the position control of an induction motor drive is presented, where the robustness of different controllers (PID or GPC) regarding model uncertainties in high frequency is enhanced while respecting a temporal template for the disturbance rejection.     

The harmonic fitting method for the assessment of the substitute geometry estimate error. Part II: statistical approach, machining process analysis and inspection plan optimisation

The Harmonic Fitting Method (HFM), presented in the first part of this work, is here used to describe the estimate error in statistical terms. In fact, the estimate error can be considered as a random variable that depends on three different random processes: the machining, the part positioning, and the measurement processes. With the HFM it is possible to determine the moments of the estimate error (systematic value and covariance matrix) as a function of the inspection plan. It will be demonstrated that the systematic part of the estimate error derives from the systematic part of the machined surface deviations, and that the random part derives from the variability of the machining and measurement processes. In particular, the causes of the machining process variability will be analysed in terms of harmonics, thus establishing a direct relationship between the machining process and the substitute geometry estimate error. Moreover, the possibility of obtaining the probability density function of the estimate error from the HFM will be analysed, together with the problem of the inspection plan optimisation. The HFM will be used to design the optimal inspection plan for a circular geometric feature, and it will be demonstrated that, when the part positioning is subjected to a rotation uncertainty, the estimate errors of the diameter and of the eccentricity are likely to follow a Gaussian and a Rayleigh distribution, respectively. A real case of turned shafts will be considered, and the HFM optimisation of the inspection plan will be discussed.     

Design optimization using Statistical Confidence Boundaries of response surfaces: Application to robust design of a biomedical implant

This paper deals with the use of Statistical Confidence Boundaries (SCB) of response surfaces in robust design optimization. An empirical model is therefore selected to describe a real design constraint function. This constraint is thus approximated by a second order polynomial expansion which is fitted to numerical simulations that use a Finite Element Method (FEM). A technique is also proposed to analyze the effects of the uncertainties of the inputs of the simulations. This approach is employed to optimize the design of a biomedical wrist implant. A real optimized implant is then manufactured and tested to validate the numerical model.     

Adaptive Monte Carlo applied to uncertainty estimation in five axis machine tool link errors identification with thermal disturbance

Knowledge of a machine tool axis to axis geometric location errors allows compensation and corrective actions to be taken to enhance its volumetric accuracy. Several procedures exist, involving either lengthy individual test for each geometric error or faster single tests to identify all errors at once.This study focuses on the closed kinematic chain method which uses a single setup test to identify the eight link errors of a five axis machine tool. The identification is based on volumetric error measurements for different poses with a non-contact Cartesian measuring instrument called CapBall, developed in house.In order to evaluate the uncertainty on each identified error, a multi-output Monte Carlo approach is implemented. Uncertainty sources in the measurement and identification chain – such as sensors output, machine drift and frame transformation uncertainties – can be included in the model and propagated to the identified errors. The estimated uncertainties are finally compared to experimental results to assess the method. It also reveals that the effect of the drift, a disturbance, must be simulated as a function of time in the Monte Carlo approach.Results shows that the machine drift is an important uncertainty source for the machine tested.     

Design of forging process variables under uncertainties

Forging is a complex nonlinear process that is vulnerable to various manufacturing anomalies, such as variations in billet geometry, billet/die temperatures, material properties, and workpiece and forging equipment positional errors. A combination of these uncertainties could induce heavy manufacturing losses through premature die failure, final part geometric distortion, and reduced productivity. Identifying, quantifying, and controlling the uncertainties will reduce variability risk in a manufacturing environment, which will minimize the overall production cost. In this article, various uncertainties that affect the forging process are identified, and their cumulative effect on the forging tool life is evaluated. Because the forging process simulation is time-consuming, a response surface model is used to reduce computation time by establishing a relationship between the process performance and the critical process variables. A robust design methodology is developed by incorporating reliability-based optimization techniques to obtain sound forging components. A case study of an automotive-component forging-process design is presented to demonstrate the applicability of the method.     

A generic coordinate transformation uncertainty assessment approach and its application in machine vision metrology

A measurement result is complete only when accompanied by a quantitative statement of its uncertainty [1]. However, there is no unified approach to the expression and computation of uncertainty, especially when coordinate transformations are involved. A generic approach to the uncertainty analysis of coordinate transformations based on constrained optimization uncertainty analysis is presented in this paper. The nonlinear constrained optimization algorithm used in this paper is a more generic algorithm for coordinate referencing than conventional 3-2-1 nesting method. We performed analysis on two types of uncertainty computation that relate to the final measurement: uncertainty of a single point in a world frame and uncertainty of distance measurement. The proposed method provides a mechanism for on-line assessment of measurement uncertainties on any coordinate measuring machines, mechanical or optical. Results of Monte Carlo simulation from a structured-light optical coordinate measurement machine (CMM) show that under normal machine operation conditions the approximations are valid.     

An analytical assessment of measurement uncertainty in precision inspection and machine calibration

Dimensional inspection is commonly used to scrutinize the quality of manufactured products against the established standards and specifications. Meanwhile, dimensional measurement of an artifact is also commonly used as one of the methods for machine-performance calibration. However, the quality and reliability of many inspection and measurement processes are often contaminated by various uncertainties. Two prominent sources for measurement uncertainties are: (a) the imperfection of a measuring device, and (b) the dimensional deviation and geometric characteristics of a measured feature. Usually, the effects of both types of uncertainty are compounded by one another. To ensure the quality and reliability of any inspection process, measurement uncertainty needs to be addressed for all data acquisition activities. A method is also needed to identify and decouple the effect of compounded uncertainties. If this can be done, then the data collected can be properly adjusted and a more meaningful analysis result can be drawn. In this paper, the issues of uncertainty identification for machine calibration and dimension measurement using artifacts with various geometric features are discussed. Analytical models are derived first to identify and then decouple the compounded effect of both types of uncertainties. Finally, case studies are used to illustrate the procedures for both identifying and decoupling the compounded effect of the measurement uncertainties.     

Assembly and Disassembly Processes in Product Life Cycle Perspectives

Industrial companies change the paradigms of business operations from optimisation of manufacturing processes to optimisation of products life cycles in order to activate the value of products, taking into account the potentials of product services in all phases of each product's life. From design to the end of their life capital intensive products, like manufacturing or assembly systems, are linked to a manufacturer network by global communication systems. This network allows special services even in the phases of usage and recycling. For this new paradigm it is necessary to develop strategies, methods and technologies to manage the business processes and the information and knowledge required in all phases of a product's life and to industrialise the processes of design, assembly, usage, service and remanufacturing by disassembly and recycling. It is the objective of this paper to define the processes for management of life cycle with a focus on assembly, service and disassembly of capital intensive products.     

Comparison between X-ray micro-computed tomography and optical scanning tomography for full 3D strain measurement by digital volume correlation

In this study, we present three-dimensional (3D) measurements by digital volume correlation (DVC) with images from X-ray micro-computed tomography (XμCT) and optical scanning tomography (OST). In this article, we compare both techniques and we define their field of application by evaluating their measurement uncertainty in the case of rigid body translations, homogeneous strain tests and localised compression tests. For displacement, measurement uncertainty is around 0.037 voxel for OST and 0.049 voxel for XμCT which correspond to a measured strain about 0.1%. For larger strain, uncertainty increases with strain intensity but relative error remains constant about 10%. To resume, measurement uncertainties given by DVC are similar to the ones which we obtain generally for a two-dimensional (2D) study with DIC.     

Reduction of planning efforts for decision making under uncertainty in global production network design

Continuous design of production networks is an essential element to overcome historically grown, inefficient production networks, as they are common for manufacturing companies. In order to enable continuous network design, fast and low-effort methods for investment and allocation decision making are required without losing decision quality. This paper introduces a decision making approach that reduces planning efforts by systematically focusing on main influencing factors and reducing their uncertainty. The approach was applied to a real allocation decision of a machine tool manufacturer.     

Sensitivity of transfer functions of transport of hydrogen in elastic metals to the uncertainty of their parameters

A comparison of sensitivity of the estimates of bulk elastic modulus of the M–H system and diffusion coefficient of hydrogen, acquired from three alternative transfer functions (TFs) of transport of hydrogen through a thin membrane specimen of metal, was performed with respect to the uncertainty of equilibrium concentration of hydrogen, its partial molar volume and thickness of the specimen. In most cases, the estimate of elastic modulus is more sensitive to these uncertainties than the estimate of diffusion coefficient, and the effects of uncertainty of the partial molar volume are of utmost importance. However, in the case of relatively small concentrations of hydrogen, uncertainty of the specimen’s thickness is most important for the estimate of elastic modulus acquired from one-port flux TF, while the estimates of both parameters acquired from the concentration TF are immune to the uncertainties discussed here. Nevertheless, other factors determine that for practical use the one-port flux TF is the most adequate of the three.     

An evolutionary performance-enhanced PSO approach by using a BP neural-learning-based PID controller

In this paper, a novel evolutionary strategy-based particle swarm optimisation (PSO) approach is presented. The evolutionary strategy is dependent on a BP neural proportional integral derivative (PID) controller, which is intentionally placed between the position and the global best position in a closed loop of their relationship. The BP neural PID controller is designed to aid the position to well track the global best position. Furthermore, it can be utilised to improve the evolutionary dynamics of the PSO algorithm. The experiments for performance evaluation are tried on both analytical benchmark functions and an automatic control system of pulp consistency in pulping and paper-making engineering. The experimental results illustrate that the proposed approach enhances the diversity of swarms, considerably improves the global convergence efficiency and outperforms the PSO algorithm. In addition, it is also easily used in real industrial practice and offers a novel and convenient solution to engineering optimisation design of industrial systems.     

Uncertainty and modelling cost based methodology for modelling choices in multiscale structures

Choices made in the design phases of multiscale structures are guided by the integration of knowledge on parameters at different scales of observation of the structure. These choices are often based on many experimental and predictive campaigns which increase modelling costs. The work developed here integrates the consideration of uncertainties in order to rationalise the cost of modelling (predictive and experimental) while controlling uncertainty over those parameters that are of interest for the structure scale. The methodology is applied to the study of a thick composite pressure vessel to be used for hydrogen storage.     

Quantification of uncertainties in grain size predictions of a microstructure-based flow stress model and application to gear wheel forging

For reliable process design, full knowledge of the possible spread of the predicted target values such as grain size is desirable. In real production the spread of final product properties is caused by uncertainties in the processing conditions and the material behavior. This paper proposes a strategy which allows for incorporating the material behaviors uncertainties in a microstructure model. This model is applied to the design of a hot-forging process. It is shown that the probability distribution of the grain size value is asymmetric and predicts occurrences of grain sizes with a large deviation from the most probable grain size.     

Valuation of changeable production systems

The introduction of changeable production systems increases the adaptability of factories in an uncertain environment. This paper analyzes the valuation of the economic efficiency of such systems using the real options theory. The developed approach proposes a structured valuation process for selecting options related to production engineering and factory planning. Based on a classification of possible modifications of a production system, options profiles are derived. Considering the existing uncertainty of a company’s environment these profiles may be utilized to identify real options which ought to be included when valuating a technical production system design.     

An intelligent agent-based architecture for resilient digital twins in manufacturing

Digital twins (DTs) offer the potential for improved understanding of current and future manufacturing processes. This can only be achieved by DTs consistently and accurately representing the real processes. However, the robustness and resilience of the DT itself remain an issue. Accordingly, this paper offers an approach to deal with uncertainty and disruptions, as the DT detects these effectively and self-adapts as needed to maintain representativeness. The paper proposes an intelligent agent-based architecture to improve the robustness (including accuracy of representativeness) and resilience (including timely update) of the DT. The approach is demonstrated on a case of cryogenic secondary manufacturing.