Numerical Design Of Experiments Research Articles

A knowledge-based system for numerical design of experiments processes in mechanical engineering

This paper describes a specific knowledge-based system (KBS) to assist designers in configuring numerical design of experiments (NDoE) processes efficiently. NDoE processes are applied in product design to improve the quality of product, by taking into account variabilities and uncertainties. NDoE processes are defined by various and complex methodologies to achieve several objectives, as optimization, surrogate modeling or sensitivity analysis. On the other hand, NDoE processes may demand huge computing resources to execute hundreds simulations, and also advanced expert knowledge to set the best configuration amongst numerous possibilities. Designers aim to obtain most useful results with a minimal computational cost as soon as possible. Thus, the configuration step must be as fast as possible, and it must lead to an efficient combination of complex methods, algorithms and hyper-parameters, to obtain valuable information on the product. The proposed KBS and its inference engine, a bayesian network, is detailed and applied to a product developed by automotive industry. The KBS propose new efficient configurations to achieve designers' goal. This application shorten the configuration step of the NDoE process, and enables designers to use more complex methods. It also allows designers to capitalize knowledge and learn from each past NDoE process.

The ENBIS‐17 Quality and Reliability Engineering International Special Issue

The ENBIS‐17 Quality and Reliability Engineering International Special Issue

Spray impingement wall film breakup by wave entrainment

Fuel spray impingement on engine wall and piston in the spark-ignition direct-injection (SIDI) setting has been considered a major concern in the aspect of engine emission and combustion efficiency. Excess wall film will result in deterioration of engine friction, incomplete combustion, and substantial cycle-to-cycle variations. These effects are more pronounced during engine cold-start process. Therefore, the formation of wall film on engine wall/piston and the dynamic process of the wall film interacting with impinging spray and spray-induced gas flow are of great significance for reducing wall film mass. However, the dynamic process of wall film was not investigated thoroughly in existing literatures. This work will present a high-speed, simultaneous measurement of a single-hole spray structure, as well as wall film geometry and thickness, via Mie scattering and volumetric laser-induced fluorescence, respectively. Quantitative film thickness measurement was achieved via fluorescence intensity signal calibration with a known, wedge-shape liquid film apparatus. Remarkable wall film droplet entrainment at the leading edge of the liquid film waves was revealed in the measurement, which has not been adequately depicted or analyzed in existing spray impingement studies. A considerable amount of liquid droplets detaches from the liquid film via liquid film fingering, during which process the quantity of liquid mass on the wall is decreased. Quantitative analysis of such phenomenon is performed and we estimated that a liquid mass equivalent to 30–40% of the residual liquid film mass is detached from the liquid film via wave entrainment. Furthermore, through the comparative study of the side view of the spray and the liquid film caused by spray impingement, it is shown that non-uniform spray structure is likely the cause of liquid film wavy motions. These observations suggest that wave entrainment should be considered by numerical models and experimental designs to accurately predict spray impingement phenomenon.

A parametric and non-intrusive reduced order model of car crash simulation

Industrials have an intensive use of numerical simulations in order to avoid physical testing and to speed up the design stages of their products. The numerical testing is indeed quicker to set-up, less expensive, and supplies a lot of information about the system under study. Moreover, it can be much closer to the physical tests as the computation power increases. Despite the rise of this power, time consuming simulations remain challenging to be used in design process, especially in an optimization study. Crash simulations belong to this category. These rapid dynamic computations are used by RENAULT during the sizing of the vehicle structure in order to ensure that it meets specifications set up to reach safety criteria in case of accidents. They are completed using finite element software such as VPS (Virtual Performance Solver) developed by ESI group that will be used in this study. For car manufacturers, the goal of the optimization study is to minimize the mass of the vehicle (and thus its consumption) by modifying the thicknesses of some parts (from 20 to 100 variables). Industrials such as RENAULT currently perform optimization studies based on numerical design of experiments. The number of computations required by this technique is from 3 to 10 times the number of variables. This is too much in order to be intensively used in a design process.In order to decrease the time-to-market and to explore alternative technical solutions, we explore the potential of using a parametrized reduced order model in the optimization studies. The parametrized reduced order model gives an estimation of the high-fidelity result for a new set of parameters without using the solver, by analysing the existing results of previous computations with various sets of parameters. The developed reduced order model is called ReCUR. It is partly based on a CUR approach embedded in a regression analysis. The regression statistical model uses the data of a few calculations made with the solver. Other tools such as clustering and linear programming are used to get the regression analysis more efficient.It is hoped to drastically reduce the number of required simulations of a standard optimization study. In this paper, the construction of the reduced order model will be presented. Then, the relevancy of using the reduced order model into a design process will be exhibited through the treatment of two industrial test-cases. Some improvements of the method as well as several potential uses will then be outlined. The applications will highlight the promising power of the method to shorten design process using optimization and long-run simulations.

A numerical model-assisted experimental design study of inertia-based particle focusing in stepped microchannels

Focusing methods based on fluid inertia have been demonstrated to be able to concentrate particles with a high precision and predictability offering great potential for practical application as manipulation of bodily fluids as blood for clinical diagnostics. However, to perform focusing to one single position at the center location of a channel with high focusing quality on the one hand and low pressure losses and shear stresses on the other hand still is a challenge in microchannel design. Three different stepped microchannels are analyzed via bright-field microscopy to investigate the impact of various geometric parameters on the focusing behavior of spherical particles. Reynolds numbers within a range of \( 8 \le Re \le 75\) are measured to evaluate the impact of flow conditions on the focusing characteristics. The microchannels show the ability to focus particles to a single stream with a maximum focusing efficiency of 97.1\(\%\) and purity of 99.1\(\%\). The focusing strongly depends on the channel geometry, that is, step length, step height, and settling length. A semiempirical model is developed to predict force-induced focusing ability at maximum shear stresses that are reduced significantly compared to previous systems. This model is demonstrated to be in very good agreement with experimental results and therefore can be utilized for future device design. Finally, guidelines for the design of stepped single-stream focusing devices at low pressure losses and low maximum shear stresses are derived based on experimental, numerical, and semiempirical data.

Adaptive Numerical Designs for the Calibration of Computer Codes

Making good predictions of a physical system using a computer code requires the inputs to be carefully specified. Some of these inputs, called control variables, reproduce physical conditions, whereas other inputs, called parameters, are specific to the computer code and most often uncertain. The goal of statistical calibration consists in reducing their uncertainty with the help of a statistical model which links the code outputs with the field measurements. In a Bayesian setting, the posterior distribution of these parameters is typically sampled using Markov Chain Monte Carlo methods. However, they are impractical when the code runs are highly time-consuming. A way to circumvent this issue consists of replacing the computer code with a Gaussian process emulator, then sampling a surrogate posterior distribution based on it. Doing so, calibration is subject to an error which strongly depends on the numerical design of experiments used to fit the emulator. Under the assumption that there is no code discrepancy, we aim to reduce this error by constructing a sequential design by means of the expected improvement criterion. Numerical illustrations in several dimensions assess the efficiency of such sequential strategies.

Orthogonal optimization of a water hydraulic pilot-operated pressure-reducing valve

In order to optimize the comprehensive characteristics of a water hydraulic pilot-operated pressure-reducing valve, numerical orthogonal experimental design was adopted. Six parameters of the valve, containing diameters of damping plugs, volume of spring chamber, half cone angle of main spool, half cone angle of pilot spool, mass of main spool and diameter of main spool, were selected as the orthogonal factors, and each factor has five different levels. An index of flowrate stability, pressure stability and pressure overstrike stability (iFPOS) was used to judge the merit of each orthogonal attempt. Embedded orthogonal process turned up and a final optimal combination of these parameters was obtained after totally 50 numerical orthogonal experiments. iFPOS could be low to a fairly low value which meant that the valve could have much better stabilities. During the optimization, it was also found the diameters of damping plugs and main spool played important roles in stability characteristics of the valve.

Behavior of a masonry wall subjected to mining subsidence, as analyzed by experimental designs and response surfaces

In this paper we present the principles of a methodology for studying the impact of mining subsidence on masonry buildings, using numerical experimental designs, statistical analysis and surface responses, where the subsidence is the input variable, joint tensile strength and cohesion are the control variables and the total length of cracks is the output variable. Numerical experiments are carried out using a distinct element model of a masonry wall where masonry blocks are explicitly taken into account. Calculations are also carried out for two types of subsidence according to the position of the building in the convex curvature or the concave curvature zone of the ground surface subsidence. Statistical analysis of the results shows the dominant influence of the joint tensile strength over the joint cohesion and the absence of interaction between these two parameters.

Controlling bovine paratuberculosis at a regional scale: Towards a decision modelling tool

Johne’s disease (paratuberculosis), a worldwide enzootic disease of cattle caused by Mycobacterium avium subsp. paratuberculosis (Map), mainly introduced into farms by purchasing infected animals, has a large economic impact for dairy producers. Since diagnostic tests used in routine are poorly sensitive, observing Map spread in the field is hardly possible, whereas there is a need for evaluating control strategies. Our objective was to provide a modelling framework to compare the efficacy of regional control strategies combining internal biosecurity measures and testing of traded animals, against Map spread in a metapopulation of dairy cattle herds. We represented 12,857 dairy herds located in Brittany (France), based on data from 2005 to 2013, used to calibrate herd sizes and demographic rates and to define trade events in a multiscale model of Map infection dynamics. By clustering and categorical descriptive analysis of intensive simulations of this model, based on a numerical experimental design, a large panel of control measures was explored. Their efficacy was assessed on model outputs such as the prevalence and probability of extinction at the metapopulation level. In addition, we proposed a scoring for the effort required to implement control measures and prioritized control strategies based on their theoretical epidemiological efficacy. Our results clearly indicate that eradication cannot be achieved on the mid term using available control measures. However, we identified relevant combinations of measures that lead to the control of Map spread with realistic level of implementation and coverage. The study highlights the challenge of controlling paratuberculosis in an endemically infected region as related to the poor test characteristics and frequent trade movements. Our model lays the foundations for a flexible and efficient tool to help collective animal health managers in defining relevant control strategies at a regional scale, accounting for local specificities in terms of contact network and farms’ characteristics.

PulseR: Versatile computational analysis of RNA turnover from metabolic labeling experiments.

Metabolic labelling of RNA is a well-established and powerful method to estimate RNA synthesis and decay rates. The pulseR R package simplifies the analysis of RNA-seq count data that emerge from corresponding pulse-chase experiments. The pulseR package provides a flexible interface and readily accommodates numerous different experimental designs. To our knowledge, it is the first publicly available software solution that models count data with the more appropriate negative-binomial model. Moreover, pulseR handles labelled and unlabelled spike-in sets in its workflow and accounts for potential labeling biases (e.g. number of uridine residues). The pulseR package is freely available at https://github.com/dieterich-lab/pulseR under the GPLv3.0 licence. a.uvarovskii@uni-heidelberg.de or christoph.dieterich@uni-heidelberg.de. Supplementary data are available at Bioinformatics online.

Modeling and optimization of open crack detection by flying spot thermography

The aim of this work is the detection of open defects in metallic materials using laser-material interaction coupled with infrared thermography. This process is a possible alternative for magnetic particles testing and dye penetrant testing in the field of non-destructive testing. In this context, a numerical simulation model is implemented with COMSOL Multiphysics®. This model allows us to look for the process optimal parameters through numerical experimental designs. It's an original work for modeling a back and forth scanning. Experimental works are also done in order to validate our numerical model and also to have reliable input data.

Performance enhancement of a pump impeller using optimal design method

This paper presents the performance evaluation of a regenerative pump to increase its efficiency using optimal design method. Two design parameters which define the shape of the pump impeller, are introduced and analyzed. Pump performance is evaluated by numerical simulation and design of experiments(DOE). To analyze three-dimensional flow field in the pump, general analysis code, CFX, is used in the present work. Shear stress turbulence model is employed to estimate the eddy viscosity. Experimental apparatus with an open-loop facility is set up for measuring the pump performance. Pump performance, efficiency and pressure, obtained from numerical simulation are validated by comparison with the results of experiments. Throughout the shape optimization of the pump impeller at the operating flow condition, the pump efficiency is successfully increased by 3 percent compared to the reference pump. It is noted that the pressure increase of the optimum pump is mainly caused by higher momentum force generated inside blade passage due to the optimal blade shape. Comparisons of pump internal flow on the reference and optimum pump are also investigated and discussed in detail.

POD surrogate models using adaptive sampling space parameters for springback optimization in sheet metal forming

Abstract The use of high strength steels for the stamping process in automotive body parts requires master spring back effects. Several parameters of the stamping process have an influence on springback effects. It is possible to optimize these parameters, but only way to achieve this optimization procedure is to use numerical simulation of the stamping process. Nevertheless, the optimization process requires an expensive evaluation of a high fidelity model over the whole design space. To reduce the overall computational cost, a surrogate model for the optimization process replaces the high fidelity model. Extensive design of numerical experiments on the overall design space of the high fidelity model is needed to build this surrogate model. To improve the efficiency of the overall optimization process, this paper presents Proper Orthogonal Decomposition (POD) surrogate models using adaptive sampling design space. Here the POD surrogate model aims to represent the final displacement field from the initial high-fidelity simulation and use the reduced basis and the radial basis functions (RBF) interpolation of the POD coefficient to describe and predict the final shape. During the optimization process, the new samples of the high-fidelity model are added using the minimization of the predicted objective function criterion. The proposed methodology is illustrated with the “U-bend” from the Numisheet2011 benchmark. Two parameters, the blank holder force and die radius are chosen to optimize the spring back effect.

Measurements and Stochastic FEA with Application in Thermomechanical Characterization of Electronic Packages

The aim of this study was to validate/calibrate two tools to be able to reliably measure/predict warpage at ambient temperature, especially for ball grid array (BGA) electronic packages. The tools used in this study were a high-precision microscope and a finite-element model. First, the authors calibrated the microscope by comparing the obtained results with the results obtained with a shadow moire machine (considered as reference) and then the finite-element model was calibrated to the microscope. The numerical study was not restricted to a deterministic approach; a stochastic study was also performed for taking into account parameter uncertainties. The results demonstrated that both tools are reliable alternatives for thermomechanical characterization of BGA packages at room temperature. The results obtained in the finite-element model calibration phase showed the importance of adopting a probabilistic approach, at the same time proving that the elaborated warpage numerical model is a good basis for future advanced analysis such as optimization or numerical design of experiments while having reasonable investment and time-consumption costs.

A framework for the probabilistic analysis of PEMFC performance based on multi-physical modelling, stochastic method, and design of numerical experiments

In this article, a probabilistic approach is applied to evaluate the impact of the GDL porosity uncertainty on the electrical performances of a PEMFC. The study is based on the use of a dynamical, symbolic, and acausal knowledge model. Some statistical distributions are introduced on the input model parameter (porosity) and the statistical distributions induced on the output parameters (cell voltage, resistance) are analyzed. The difference observed between the shapes of the input and output distributions (respectively some Gaussian and inverse Gamma distributions with a threshold phenomenon) is the result of strong nonlinearities linked with the integration of multiphase flow phenomena in the modelling (e.g. diffusion limit of the humidified air in the GDL). The study is also conducted for different conditions of temperature and pressure through a design of numerical experiments. One of the results obtained is that the variation coefficient related to the GDL porosity has, compared to the other parameters with their intervals of variation considered, little effect on the average output distributions. However, the dispersion introduced on the porosity impacts their shapes (e.g. significant effect on the standard deviation).

Design of a single-phase PTS numerical experiment for a reference Direct Numerical Simulation

The integrity assessment of the Reactor Pressure Vessel (RPV) is considered to be an important issue for lifetime extension of nuclear reactors. A severe transient that can threaten the integrity of the RPV is the existence of a Pressurized Thermal Shock (PTS) during a Loss-of-Coolant Accident (LOCA). A PTS consists of a rapid cooling of the RPV wall under pressurized conditions that may induce the criticality of existing or postulated defects inside the vessel wall. The most severe PTS event has been identified by Emergency Core Cooling (ECC) injection during a LOCA. The traditional one-dimensional system codes fail to reliably predict the complex three-dimensional thermal mixing phenomena in the downcomer occurring during the ECC injection. Hence, CFD can bring real benefits in terms of more realistic and more predictive capabilities. However, to gain trust in the application of CFD modelling for PTS, a comprehensive validation programme is necessary. In the absence of detailed experimental data for the RPV cooling during ECC injection, high fidelity Direct Numerical Simulation (DNS) databases constitute a valid alternative and can serve as a reference.The aim of this work is to design a numerical experiment aimed to generate a high quality reference DNS database for a simplified PTS scenario. This takes into account the turbulent mixing in the downcomer and the evolution of the temperature distribution for both structures and fluid during a single-phase flow PTS scenario. In spite of simplifications, such a DNS analysis represents a very demanding application. A priori, it should be demonstrated that all the relevant turbulent scales will be fully resolved, which requires a huge computational power. A wide range of Reynolds Averaged Navier–Stokes (RANS) calculations are performed in order to determine a realistic PTS computational domain. Furthermore, the resulting numerical experiment is optimized in terms of geometry, boundary conditions, and physical properties.

Robustness Analysis of a Geant4-GATE Simulator for Nanoradiosensitizers Characterization.

The increase of computational environments dedicated to the simulation of nanoparticles (NP)-X-Rays interactions has opened new perspectives in computer-aided-design of nanostructured materials for biomedical applications. Several published studies have shown a crucial need of standardization of these numerical simulations. That is why, a robustness multivariate analysis was performed in this paper. A gold nanoparticle (GNP) of 100 nm diameter was selected as a standard nanosystem activated by a X-ray source placed just below the NP. Two response variables were examined: the dose enhancement in seven different spatial regions of interest around the NP and the duration of the experiments. Nine factors were pre-identified as potentially critical. A Plackett-Burman design of numerical experiments was applied to estimate and test the effects of each simulation factors on the examined responses. Four factors-the working volume, the spatial resolution, the spatial cutoff, and the computational mode (parallelization)-do not significantly affect the dose deposition results and none except the last one may reduce the computational duration. The energy cutoff may cause significant variations of the dose enhancement in some specific regions of interest: the higher the cutoff, the closer the secondary particles will stop from the GNP. By contrast, the Auger effect as well as the choice of the physical medium and the fluence level clearly appear as critical simulation parameters. Consequently, these four factors may be compulsory examined before comparing and interpreting any simulation results coming from different simulation sessions.

A Numerical Design and Feasibility Study of Self-Wastage Experiment Using Simulant Material in a Sodium Fast Reactor

A sodiumewater reaction takes place when high-pressured water vapor leaks into sodium through a tiny defect on the surface of the heat transfer tube in a steam generator of the sodium-cooled fast reactor. The sodiumewater reaction brings deterioration of the mechanical strength of the heat transfer tube at the initial leakage site. As a result, it damages the crack itself, which may eventually enlarge into a larger opening. This self-enlargement is called “self-wastage phenomenon.” In this study, a simulant experiment was proposed to reproduce the self-enlargement of a crack and to evaluate the mechanism of the selfwastage. The damage on the surface of the crack was simulated by making the neutralization reaction with hydrochloric acid solution and sodium hydroxide solution. A numerical investigation was carried out to validate the feasibility of the approach and to determine experimental conditions. From the computation results, it is observed that when 5M HCl is injected into 5M of NaOH with 0.05 ㎧ inlet velocity, the temperature at the surface near the crack increased over 319.26 K. The computational results show that the self-wastage phenomenon is capable of being reproduced by the simulant experiment.

ODE: An Ontology for Numerical Design of Experiments

Numerical Designs of Experiments (DoE) are used in a product development process for several purposes as optimization, surrogate modelling or sensitivity analysis. Methods used to shorten the execution of DoE demands advanced knowledge and expertise to be selected and configured. A knowledge management approach can be applied to capitalize and reuse knowledge. There is a real lack in term of models for this specific application. Thus, an ontology for the numerical DoE process is proposed. It is linked to existing ontologies and data models developed for numerical simulations. This ontology will be used as a support for a decision-aid system.

Investigation on the fuel injection quantity fluctuation of a common-rail fuel injection system using response surface methodology

Variations in the parameters affect the injection characteristics of a common-rail injection system and leads to fluctuation in the fuel injection quantity. This paper is aimed at combining the numerical modelling and design of experiments to investigate the significant effects of the interactions between the common-rail injector parameters on the fluctuation in the fuel injection quantity using response surface methodology. A numerical model of a common-rail injector was developed. The model was validated by comparing simulation results with experimental measurements, which showed that it can accurately predict the fuel injection quantity of the system. The design of experiments was performed using a two-level five-factor D-optimal design. The factors studied were the pre-tension of the control valve spring, the diameter of the outlet orifice, the diameter of the inlet orifice, the needle lift and the diameter of the nozzle holes. A quadratic response surface model was suggested by means of partial least-squares regression analysis. The distribution of the standardized residuals, the relation between the predicted fluctuation and the observed fluctuation in the fuel injection quantity, the coefficient R2 of determination and the adjusted coefficient [Formula: see text] of determination for the response surface model were analysed, which demonstrated the significance of the model. Analysis of the F value and the p value for the model terms at the 95% confidence level revealed that the interactions between the control valve spring pre-tension and the nozzle hole diameter, between the outlet orifice diameter and the nozzle hole diameter, between the inlet orifice diameter and the nozzle hole diameter and between the needle lift and the nozzle hole diameter had significant effects on the fluctuation in the fuel injection quantity of the system. Furthermore, the significant effects of the interactions between these parameters were analysed in detail.