Iterative Experimental Design Research Articles

Iterative experimental design and identifiability analysis of composite material failure models

The parameter identification of failure models for composite plies can be cumbersome, due to multiple effects as the consequence of brittle fracture. Our work proposes an iterative, nonlinear design of experiments (DoE) approach that finds the most informative experimental data to identify the parameters of the Tsai-Wu, Tsai-Hill, Hoffman, Hashin, max stress and Puck failure models. Depending on the data, the models perform differently, therefore, the parameter identification is validated by the Euclidean distance of the measured points to the closest ones on the nominal surface. The resulting errors provide a base for the ranking of the models, which helps to select the best fitting. Following the validation, the sensitivity of the best model is calculated by partial differentiation, and a theoretical surface is generated. Lastly, an iterative design of the experiments is implemented to select the optimal set of experiments from which the parameters can be identified from the least data by minimizing the fitting error. In this way, the number of experiments required for the identification of a model of a composite material can be significantly reduced. We demonstrate how the proposed method selected the most optimal experiments out of generated data. The results indicate that if the dataset contains enough information, the method is robust and accurate. If the data set lacks the necessary information, novel material tests can be proposed based on the optimal points of the parameters' sensitivity of the generated failure model surface.

Immersive visualisation for a large metrological data set

A purpose of scientific visualisation is to facilitate the extraction of information within underlying data. Visualisation of a large data set requires an appropriate visualisation method where, instead of focusing on expected features, all acquired data are treated on an equal footing. Here, we identify critical differences between raw and processed data acquired from displacement measurements at the nanoscale. We illustrate the strength of immersive data visualisation in rapidly and intuitively identifying non-obvious data trends, supported by correlation with the associated measurement uncertainties. These insights can be directly leveraged for optimisation of instrumentation and measurement protocols, application of appropriate measurement corrections, and iterative experimental design changes.

Bayesian cell therapy process optimization.

Optimizing complex bioprocesses poses a significant challenge in several fields, particularly in cell therapy manufacturing. The development of customized, closed, and automated processes is crucial for their industrial translation and for addressing large patient populations at a sustainable price. Limited understanding of the underlying biological mechanisms, coupled with highly resource-intensive experimentation, are two contributing factors that make the development of these next-generation processes challenging. Bayesian optimization (BO) is an iterative experimental design methodology that addresses these challenges, but has not been extensively tested in situations that require parallel experimentation with significant experimental variability. In this study, we present an evaluation of noisy, parallel BO for increasing noise levels and parallel batch sizes on two in silico bioprocesses, and compare it to the industry state-of-the-art. As an in vitro showcase, we apply the method to the optimization of a monocyte purification unit operation. The in silico results show that BO significantly outperforms the state-of-the-art, requiring approximately 50% fewer experiments on average. This study highlights the potential of noisy, parallel BO as valuable tool for cell therapy process development and optimization.

Active learning enabled reactor characterization for mass transfer in aerobic oxidation reactions

A generalizable active learning framework enables accurate prediction of mass transfer coefficients (kLa), and iterative design of experiments to efficiently characterize new reactor configurations with minimal experimental trials.

Design of a low Young’s modulus Ti-Zr-Nb-Sn biocompatible alloy by in situ laser powder bed fusion additive manufacturing process

A new metastable β Ti-22Zr-11 Nb-2Sn (at%) biomedical titanium alloy was elaborated from elemental powders by in situ laser powder bed fusion technique (L-PBF). Iterative design of experiments was used to identify optimized manufacturing parameters to obtain dense and homogeneous parts (>99.3%) with low unmolten niobium fraction (<0.05%). The microstructural and mechanical properties of this alloy were investigated in its as-fabricated state and after two heat-treatments at 400 °C and 700 °C, respectively. The alloy, showing a β-type microstructure, possesses a very low Young’s modulus (∼50 GPa) combined with a high tensile strength reaching about 1100 MPa after heat-treatment. These properties are very suitable for medical devices in osseous site such as hip prostheses or dental implants and the results were compared to the Ti-6Al-4 V ELI (wt%) medical grade fabricated from pre-alloyed powders. Thus, the present work demonstrates the significant potential of the in situ L-PBF technique to elaborate highly biocompatible titanium alloys with tailored chemical compositions.

Influence of the Dross Formation of the Laser-Cut Edge on the Fatigue Strength of AISI 304

Laser cutting is a thermal cutting process based on material melting that results in characteristic features of the cut edge. The dross in particular is a crucial quality-determining feature which occurs especially when processing higher sheet thicknesses. The influence of the dross geometry on the fatigue behavior of AISI 304 was investigated in this work. Using iterative experimental design, samples with different dross geometries were produced by varying laser cutting parameters. Four characteristic dross geometries were identified and used to classify manufacturing parameters: dross-free, small droplets, large droplets and very coarse dross. Fatigue tests were performed up to 107 load cycles and revealed a dependence of the fatigue behavior on the dross geometries due to their different notch effects. It was found that the dross dominated the fatigue strength only above a certain dross height. At low dross heights, the surface relief of the cut edge dominated fatigue strength. The different cut edge properties (surface relief and dross) depend on the process parameters during laser cutting. Gas pressure and feed rate in particular showed a significant influence. The findings of this work provide information about the fatigue behavior’s dependence on dross geometry, which can be transferred to higher sheet thicknesses or complex sample geometries.

Multivariate prediction intervals for bagged models

Accurate uncertainty estimates can significantly improve the performance of iterative design of experiments, as in sequential and reinforcement learning. For many such problems in engineering and the physical sciences, the design task depends on multiple correlated model outputs as objectives and/or constraints. To better solve these problems, we propose a recalibrated bootstrap method to generate multivariate prediction intervals for bagged models such as random forest and show that it is well-calibrated. We apply the recalibrated bootstrap to a simulated sequential learning problem with multiple objectives and show that it leads to a marked decrease in the number of iterations required to find a satisfactory candidate. This indicates that the recalibrated bootstrap could be a valuable tool for practitioners using machine learning to optimize systems with multiple competing targets.

Rational Design of a Polyurethane Foam.

Polyurethane (PU) foams are exceptionally versatile due to the nature of PU bond formation and the large variety of polymeric backbones and formulation components such as catalysts and surfactants. This versatility introduces a challenge, namely a near unlimited number of variables for formulating foams. In addition to this, PU foam development requires expert knowledge, not only in polyurethane chemistry but also in the art of evaluating the resulting foams. In this work, we demonstrate that a rational experimental design framework in conjunction with a design of experiments (DoE) approach reduces both the number of experiments required to understand the formulation space and reduces the need for tacit knowledge from a PU expert. We focus on an in-depth example where a catalyst and two surfactants of a known formulation are set as factors and foam physical properties are set as responses. An iterative DoE approach is used to generate a set of foams with substantially different cell morphology and hydrodynamic behaviour. We demonstrate that with 23 screening formulations and 16 final formulations, foam physical properties can be modelled from catalyst and surfactant loadings. This approach also allows for the exploration of relationships between the cell morphology of PU foam and its hydrodynamic behaviour.

Intelligent Iterative Experimental Design to Achieve Maximum Model Quality for Phase Change of 22MnB5

Conducting experiments for material modeling is very costly and time-consuming when many parameters are involved, resulting in a large number of test conditions. Therefore, it is expedient to develop algorithms for the iterative identification of optimal test conditions. This method should allow the model to learn automatically so that only a small number of test conditions are selected at the beginning of the model calibration. In order to decide whether further experiments should be carried out and which test conditions need to be investigated, meta-models are generated, and the expected gain score is calculated. The next sample is selected based on the highest score, and this procedure continues until the material models meet a termination criteria. The result from the study shows that the implemented method uses 12 test conditions to generate a phase transformation model for 22MnB5 steel. The material models fitted with the proposed method provide acceptable predictions when compared with experimental data.

Using a Spreadsheet-Based Simulation to Practice and Evaluate Iterative Column Selection and Experimental Design in Chemical Fingerprinting with GC–MS

Optimization of experimental conditions in chromatography is recognized as a wear-and-tear process involving hundreds of experiments in order to maximize the number of separated compounds for each sample. The critical thinking skills and practices entailed cannot be achieved with students in a conventional lecture class. This report provides a Microsoft Excel-based spreadsheet allowing the fast simulation of fingerprinting results and the number of separated peaks for a given sample analyzed using gas chromatography–mass spectrometry. The applications were demonstrated for the simulation of extracted ion chromatograms in order to find a peak for the target compound in a model kerosene sample containing 254 volatile compounds with the optimization of the separation of fatty acid methyl esters. Using the simulation applications, students are able to perform experimental design in a lecture class. Test score results reveal the selectivity of the approach toward students with laboratory practical experience and experimental skills. This approach is expected to be useful to all higher education students in chemistry.

Which image-based phenotypes are most promising for using AI to understand cellular functions and why?

Which image-based phenotypes are most promising for using AI to understand cellular functions and why?

Iterative model-based experimental design for efficient uncertainty minimization of chemical mechanisms

The uncertainties of chemical kinetic model parameters induce uncertainties in model predictions. Automatic optimization and uncertainty minimization techniques have been developed to constrain these uncertainties based on sets of experimental target data for quantities of interest. While such methods were frequently used to optimize models for relatively well-studied systems with large numbers of available targets, only few of these experimental data points may be of crucial importance. In addition, for novel fuel candidates such as biofuels and synthetic fuels, the number of available measurements is generally limited. Thus, an important aspect to be explored in this context is the number of experimental data points required to achieve a certain degree of uncertainty reduction, and the determination of optimal experimental conditions for these. To target this question, a model-based experimental design framework based on the criterion of D-optimality is used in the present work to automatically identify these optimal conditions. As an example, the auto-ignition of dimethyl ether is investigated. The majority of experiments with high priority cover the intermediate- and low-temperature regimes, where the employed prior model exhibits the largest prediction uncertainties. It is also found that 90 % of the maximum observed reduction of average prediction uncertainty in ignition delay times can be achieved based on only the ten most informative experiments alone. The results thus demonstrate that few well-selected measurements allow for efficient model uncertainty reduction, and the employed approach provides an effective means of identifying the optimal conditions, which is useful for further experimental investigation. On the other hand, the inclusion of more experiments into the calibration process still provides additional benefit in terms of the posterior uncertainties of a number of important model parameters, which points to the importance of taking such data into account in case of their availability.

Traces matter: Targeted optimization of monoclonal antibody N-glycosylation based on/by implementing automated high-throughput trace element screening.

The use of high-throughput systems in cell culture process optimization offers various opportunities in biopharmaceutical process development. Here we describe the potential for acceleration and enhancement of product quality optimization and de novo bioprocess design regarding monoclonal antibody N-glycosylation by using an iterative statistical Design of Experiments (DoE) strategy based on our automated microtiter plate-based system for suspension cell culture. In our example, the combination of an initial screening of trace metal building blocks with a comprehensive DoE-based screening of 13 different trace elemental ions at three concentration levels in one run revealed most effective levers for N-glycan processing and biomass formation. Obtained results served to evaluate optimal concentration ranges and the right supplementation timing of relevant trace elements at shake flask and 2 L bioreactor scale. This setup identified manganese, copper, zinc, and iron as major factors. Manganese and copper acted as inverse key players in N-glycosylation, showing a positive effect of manganese and a negative effect of copper on glycan maturation in a zinc-dependent manner. Zinc and iron similarly improved cell growth and biomass formation. These findings allowed determining optimal concentration ranges for all four trace elements to establish control on desired product quality attributes regarding premature afucosylated and mature galactosylated glycan species. Our results demonstrates the power of combining robotics with DoE screening to enhance product quality optimization and to improve process understanding, thus, enabling targeted product quality control.

Iterative experimental design based on active machine learning reduces the experimental burden associated with reaction screening

Through iterative selection of maximally informative experiments, active learning renders exhaustive screening obsolete. Chosen experiments are used to train models that are accurate over the entire domain, thus reducing the experiment burden.

Versatile sequential sampling algorithm using Kernel Density Estimation

Understanding the physical mechanisms governing scientific and engineering systems requires performing experiments. Therefore, the construction of the Design of Experiments (DoE) is paramount for the successful inference of the intrinsic behavior of such systems. There is a vast literature on one-shot designs such as low discrepancy sequences and Latin Hypercube Sampling (LHS). However, in a sensitivity analysis context, an important property is the stochasticity of the DoE which is partially addressed by these methods. This work proposes a new stochastic, iterative DoE – named KDOE – based on a modified Kernel Density Estimation (KDE). It is a two-step process: (i) candidate samples are generated using Markov Chain Monte Carlo (MCMC) based on KDE, and (ii) one of them is selected based on some metric. The performance of the method is assessed by means of the C2-discrepancy space-filling criterion. KDOE appears to be as performant as classical one-shot methods in low dimensions, while it presents increased performance for high-dimensional parameter spaces. It is a versatile method which offers an alternative to classical methods and, at the same time, is easy to implement and offers customization based on the objective of the DoE.

Automated Design Optimization of a Small-Scale High-Swirl Cavity-Stabilized Combustor

A numerical optimization study is performed on a small-scale high-swirl cavity-stabilized combustor. A parametric geometry is created in cad software that is coupled with meshing software. The latter automatically transfers meshes and boundary conditions to the solver, which is coupled with a postprocessing tool. Steady, incompressible three-dimensional simulations are performed using a multiphase Realizable k-ε Reynolds-averaged Navier-Stokes (RANS) approach with a nonadiabatic flamelet progress variable (FPV) model. There are nine geometrical input parameters. There are five output parameters, viz., pattern factor (PF), RMS of the profile factor deviation, averaged exit temperature, averaged exit swirl angle, and total pressure loss. An iterative design of experiments (DOE) with a recursive Latin hypercube sampling (LHS) is performed to filter the most important input parameters. The five major input parameters are found with Spearman's order-rank correlation and R2 coefficient of determination. The five input parameters are used for the adaptive multiple objective (AMO) optimization. This provided a candidate design point with the lowest weighted objective function, which was verified through computational fluid dynamic (CFD) simulation. The combined filtering and optimization procedures improve the baseline design point in terms of pattern and profile factor. The former halved from that of the baseline design point, whereas the latter turned from an outer peak to a center peak profile, closely mimicking an ideal profile. The exit swirl angle favorably increased 25%. The averaged exit temperature and the total pressure losses remained nearly unchanged from the baseline design point.

Adaptive optimal operation of a parallel robotic liquid handling station

Results are presented from the optimal operation of a fully automated robotic liquid handling station where parallel experiments are performed for calibrating a kinetic fermentation model. To increase the robustness against uncertainties and/or wrong assumptions about the parameter values, an iterative calibration and experiment design approach is adopted. Its implementation yields a stepwise reduction of parameter uncertainties together with an adaptive redesign of reactor feeding strategies whenever new measurement information is available. The case study considers the adaptive optimal design of 4 parallel fed-batch strategies implemented in 8 mini-bioreactors. Details are given on the size and complexity of the problem and the challenges related to calibration of over-parameterized models and scarce and non-informative measurement data. It is shown how methods for parameter identifiability analysis and numerical regularization can be used for monitoring the progress of the experimental campaigns in terms of generated information regarding parameters and selection of the best fitting parameter subset.

Experimentally Infused Plant and Controller Optimization Using Iterative Design of Experiments—Theoretical Framework and Airborne Wind Energy Case Study

This research presents an iterative framework for optimizing the plant and controller for complex systems by fusing expensive but valuable experiments with cheap yet less accurate simulations. At each iteration, G-optimal design is used to generate experiments and simulations within a prescribed design space that is shrunken in size after each successful iteration. The shrinking of the design space is determined through statistical characterization of a response surface model, and further shrinking is achieved at successive iterations through a numerical model correction factor that is driven by the results of experiments. An initial validation of this iterative design optimization framework was performed on an airborne wind energy (AWE) system, where tethers and an aerostat are used in place of a tower to elevate the turbine to high altitudes. Using a unique lab-scale setup for the experiments, the aforementioned iterative methodology was used to optimize the center of mass location and pitch angle set point for the airborne wind energy system. The optimum configuration yielded a substantial improvement in system responses as compared to a numerically optimized configuration. The framework was recently extended to include four variables (horizontal and vertical stabilizer areas, center of mass location, and pitch angle set point).

Application of iterative robust model-based optimal experimental design for the calibration of biocatalytic models.

The aim of model calibration is to estimate unique parameter values from available experimental data, here applied to a biocatalytic process. The traditional approach of first gathering data followed by performing a model calibration is inefficient, since the information gathered during experimentation is not actively used to optimize the experimental design. By applying an iterative robust model-based optimal experimental design, the limited amount of data collected is used to design additional informative experiments. The algorithm is used here to calibrate the initial reaction rate of an ω-transaminase catalyzed reaction in a more accurate way. The parameter confidence region estimated from the Fisher Information Matrix is compared with the likelihood confidence region, which is not only more accurate but also a computationally more expensive method. As a result, an important deviation between both approaches is found, confirming that linearization methods should be applied with care for nonlinear models. © 2017 American Institute of Chemical Engineers Biotechnol. Prog., 33:1278-1293, 2017.

Iterative DoE Improved emission models and better optimisation results within a shortened measurement time

Design of experiments (DoE) is a widely used approach for powertrain calibration tasks to reduce testing time. The generation of DoE points and the test execution are normally performed in two separate and sequential steps. However, while building the models for certain highly nonlinear engine responses (e.g., particulate mass), state-of-the-art designs cannot distribute sufficient points in areas that are particularly relevant for the later optimisation. To improve the conventional DoE approach, this paper suggests a methodology that leaves out candidates which appear to have compromised performance in one or more output channels. The behaviour of defined measurement channels is modelled during the test execution and additional candidates are positioned iteratively. After each measurement, the models are updated. Practical examples show that this proposed and unique method known as 'COR DoE' leads to a reduction of testing effort, as well as a significantly improved modelling, optimisation and verification accuracy.