Adaptive subregion-based active Kriging for collaborative multi-failure reliability assessment

Surrogate articular contact models for computationally efficient multibody dynamic simulations

Local Interpretable Model Agnostic Explanation (LIME) is a technique to explain a black box machine learning model using a surrogate model approach. While this technique is very popular, inherent to its approach, explanations are generated from the surrogate model and not directly from the black box model. In sensitive domains like healthcare, this need not be acceptable as trustworthy. These techniques also assume that features are independent and provide feature weights of the surrogate linear model as feature importance. In real life datasets, features may be dependent and a combination of a set of features with their specific values can be the deciding factor rather than individual feature importance. They also generate random instances around the point of interest to fit the surrogate model. These random instances need not be part of the original source or may even turn out to be meaningless. In this work, we compare LIME to explanations from the formal concept lattice. This does not use a surrogate model but a deterministic approach by generating synthetic data that respects implications in the original dataset and not randomly generating it. It obtains crucial feature combinations with their values as decision factors without presuming dependence or independence of features. Its explanations not only cover the point of interest but also global explanation of the model, similar and contrastive examples around the point of interest. The explanations are textual and hence easier to comprehend than comprehending weights of a surrogate linear model to understand the black box model.

Kriging surrogate model has extracted extensive attention in reliability evaluation, owing to its excellent applicability and operability nowadays, which confronts with difficulties in balancing the efficiency and accuracy for complicated mechanical assets with multiple failure modes. Consequently, this paper devises a multi-performance reliability analysis approach within the surrogate model framework, particularly innovative in its use of cluster mixing weight. Specifically, high-value test points are selected to fit the surrogate model after sorting the samples referring to the corresponding values; then, a cluster-based active learning strategy is employed to accomplish rapid convergence, and the particle swarm algorithm is utilized to optimize relevant parameters. Afterwards, the mixing weight for every performance referring to the contributions to the final reliability is determined, and the failure probability is subsequently predicted. Furthermore, the superiority of the proposed approach with the clustering surrogate model and mixing weight, compared with traditional sampling as well as other surrogate models, has been verified via case studies, contributing to overcoming the multi-performance reliability analysis oriented to complicated mechanical assets.

Current limit state surrogate modeling methods for system reliability analysis usually build surrogate models for failure modes individually or build composite limit states. In practical engineering applications, multiple system responses may be obtained from a single setting of inputs. In such cases, building surrogate models individually will ignore the correlation between different system responses and building composite limit states may be computationally expensive because the nonlinearity of composite limit state is usually higher than individual limit states. This paper proposes a new efficient Kriging surrogate modeling approach for system reliability analysis by constructing composite Kriging surrogates through selection of Kriging surrogates constructed individually and Kriging surrogates built based on singular value decomposition. The resulting composite surrogate model will combine the advantages of both types of Kriging surrogate models and thus reduce the number of required training points. A new stopping criterion and a new surrogate model refinement strategy are proposed to further improve the efficiency of this approach. The surrogate models are refined adaptively with high accuracy near the active failure boundary until the proposed new stopping criterion is satisfied. Three numerical examples including a series, a parallel, and a combined system are used to demonstrate the effectiveness of the proposed method.

To predict the quality of a process outcome with given process parameters in real-time, surrogate models are often adopted. A surrogate model is a statistical model that interpolates between data points obtained either by process measurements or deterministic models of the process. However, in manufacturing processes the amount of useful data is often limited, and therefore setting up a sufficiently accurate surrogate model is challenging. The present contribution shows how to handle limited data in a surrogate modeling approach using the example of a cup drawing process. The purpose of the surrogate model is to classify the quality of the drawn cup and to predict its final geometry. These classification and regression tasks are solved via machine learning methods. The training data is sampled on a relatively wide range varying three parameters of a finite element simulation, namely sheet metal thickness, blank holder force, and friction. The geometrical features of the cup are extracted using domain knowledge. Besides this knowledge-based approach, an outlook is given for a data-driven surrogate modeling approach.

Temperature prediction in complex systems like gas turbines provides insights to temperature dependent damage accumulation but usually involves a huge computational cost. For simulation-based prognostics, the computational cost is a major hindrance to a real time implementation. In this work an ensemble learning based multistage surrogate modeling approach is investigated as a possible solution for reducing the computational cost. First the nodal temperature of a turbine blisk is predicted using computational fluid dynamic (CFD) simulations for a limited number of engine operating points. Next the proposed ensemble learning based surrogate modeling approach is implemented to train surrogate models for every node defining the blisk. To achieve computational efficiency, the proposed surrogate modeling framework implements in sequence, clustering techniques for data analysis, multistage polynomial regression modeling, and ensemble learning based model combination. Finally the prediction errors are quantified using the leave-one-out cross-validation method. The result suggests that the computational time could be significantly reduced using the proposed ensemble learning based multistage surrogate modeling technique. The threshold value used to tune the polynomial regression model complexity is also shown to influence the time for surrogate model training.

The objective of this study is to develop an integrated computer-aided engineering (CAE) optimization system that can quickly and intelligently determine the optimal process conditions for injection molding. This study employs support vector regression (SVR) to establish the surrogate model based on executions of three-dimensional (3D) simulation for a selected dataset using the latin hypercube sampling (LHS) technique. Once the surrogate model can satisfactorily capture the characteristics of simulations with much less computing resources, a hybrid optimization genetic algorithm (GA) or a multi-objective optimization GA is then used to evaluate the surrogate model to search the global optimal solutions for the single or multiple objectives, respectively. The performance and capabilities of other surrogate modeling approaches, such as polynomial regression (PR) and artificial neural network (ANN), are also investigated in terms of accuracy, robustness, efficiency, and requirements for training samples. Experimental validations and applications of this work for process optimization of a special box mold and a precision optical lens are presented.

Computer models and simulations are essential system design tools that allow for improved decision making and cost reductions during all phases of the design process. However, the most accurate models tend to be computationally expensive and can therefore only be used sporadically. Consequently, designers are often forced to choose between exploring many design alternatives with less accurate, inexpensive models and evaluating fewer alternatives with the most accurate models. To achieve both broad exploration of the design space and accurate determination of the best alternatives, surrogate modeling and variable accuracy modeling are gaining in popularity. A surrogate model is a mathematically tractable approximation of a more expensive model based on a limited sampling of that model. Variable accuracy modeling involves a collection of different models of the same system with different accuracies and computational costs. We hypothesize that designers can determine the best solutions more efficiently using surrogate and variable accuracy models. This hypothesis is based on the observation that very poor solutions can be eliminated inexpensively by using only less accurate models. The most accurate models are then reserved for discerning the best solution from the set of good solutions. In this paper, a new approach for global optimization is introduced, which uses variable accuracy models in conjuction with a kriging surrogate model and a sequential sampling strategy based on a Value of Information (VOI) metric. There are two main contributions. The first is a novel surrogate modeling method that accommodates data from any number of different models of varying accuracy and cost. The proposed surrogate model is Gaussian process-based, much like classic kriging modeling approaches. However, in this new approach, the error between the model output and the unknown truth (the real world process) is explicitly accounted for. When variable accuracy data is used, the resulting response surface does not interpolate the data points but provides an approximate fit giving the most weight to the most accurate data. The second contribution is a new method for sequential sampling. Information from the current surrogate model is combined with the underlying variable accuracy models’ cost and accuracy to determine where best to sample next using the VOI metric. This metric is used to mathematically determine where next to sample and with which model. In this manner, the cost of further analysis is explicitly taken into account during the optimization process.

In this paper, we present an optimal-control-based method for ageing-aware charging. A surrogate modeling approach is used to approximate ageing-related Doyle-Fuller-Newman (DFN) model states, where the surrogate model is a combination of a black-box finite-dimensional linear-time-invariant model and a static nonlinear model that is a function of state-of-charge. We formulate the optimal-control problem as minimizing the side reactions for a given charging time and subject to several ageing-related constraints that are commonly used in literature. We will show that the ageing-related DFN model states can be well approximated by the proposed surrogate model. Furthermore, we will show that with the surrogate modeling approach, even in an open-loop execution of the optimal-control-based method, the considered constraints are only marginally violated when applied to the DFN model. Finally, we will compare the Pareto front achieved with the proposed optimal-control-based method with the Pareto fronts achieved with various multi-stage charging protocols. Here, we will show that the proposed optimal-control-based method achieves a significantly improved Pareto front over the multistage charging protocols.

This paper presents an engine sizing and cycle selection study of ultra high bypass ratio engines applied to a subsonic commercial aircraft in the N+2 (2025) timeframe. NASA has created the Environmentally Responsible Aviation (ERA) project to serve as a technology transition bridge between fundamental research (TRL 1–4) and potential commercial application (TRL 7). Specifically, ERA is focused on subsonic transport technologies that could reach TRL 6 by 2020 and can be integrated into an advanced vehicle concept to simultaneously meet the ERA project metrics for noise, emissions, and fuel burn. An important variable in exploring the technology trade space is the selection of the optimal engine cycle for use on the advanced aircraft. Previous literature demonstrated the cycle optimization using a design of experiments (DOE) to explore the engine cycle design space for a pre-defined technology package. However, since the optimal engine cycle is dependent upon the specific technology package, this process would have to be repeated to ensure optimal performance for each technology package. With more than 80 technologies to be analyzed, the combinatorial space of technology packages is enormous. As a result, executing a DOE to find the optimum engine cycle for each technology package is infeasible. To address this issue, it is proposed to use surrogate models that encompass the engine cycle and technology design space to enable fast and accurate optimization of the engine cycle for any given technology package. This paper describes the generation and analysis of surrogate models used for technology assessment and cycle optimization of an ultra high bypass geared turbofan engine architecture. The first study in the paper shows that a single surrogate model can be used to accurately simulate both a technology and cycle design space. To demonstrate the proposed surrogate modeling approach, the cycle design space for three different technology packages was analyzed. This study demonstrated that when an optimal cycle is found within the constrained interior of a design space, the surrogate modeling approach is quite accurate. The study also established that the surrogate models can also be used to assess potential cycles at the boundaries or even outside of the region for which they were trained.

View Video Presentation: https://doi.org/10.2514/6.2023-0233.vid Despite advancements made in computational and experimental analysis approaches, achieving successful store separation from rotorcraft remains a highly rigorous task. Wind tunnel-based approaches for store separation modeling largely remain infeasible due to the challenges associated with scaling store wake interactions. Furthermore, while high-fidelity computational fluid dynamics (CFD) approaches are capable of closely matching rotorcraft flight test, the high computational expense of these approaches greatly limits the total number of CFD simulations which can be run. However, even with this sparse sampling obtained through CFD potentially terabytes of high-fidelity data is still generated for the flow field. The objective of this study is to determine the feasibility of leveraging this data for the derivation of meaningful surrogate models to the topic of rotorcraft store separation. In this study, two surrogate modeling approaches will be investigated for their ability to predict store surface pressure distributions as a store is launched from a rotorcraft in hover. To generate these surrogate models, three CFD simulations are completed while varying the propulsive force assigned to the store. The CFD simulations were completed using the High-Performance Computing Modernization Program Computational Research and Engineering Acquisition Tools and Environments Air Vehicles Helios code. Once the surrogate models had been generated, an additional CFD simulation with a new propulsive force was assigned to the store. The results of this validation indicated that while the POD surrogate model struggled to provide detailed predictions of store-distributed loads, mean load variations could be modeled well. Results further indicated that through leveraging the CNN-based surrogate model, a significant improvement in distributed load modeling could be obtained. It was further identified that once distributed loads were integrated both POD and CNN-based surrogate models provided a viable path for the generation of a trajectory prediction-based surrogate model for rotorcraft applications. Both POD and CNN-based surrogate models provided significantly reduced computational costs. For each rotorcraft store separation CFD simulation, the computational cost required 10 days of simulation time across 880 CPU cores. While using the surrogate model, comparable predictions could be produced in under three minutes on a single core.

In this study, two surrogate modeling approaches will be investigated for their ability to predict store surface pressure distributions as a store is launched from a rotorcraft in hover. To generate these surrogate models, three computational fluid dynamics (CFD) simulations were completed while varying the propulsive force assigned to the store. Once the surrogate models were generated, an additional CFD simulation with a new propulsive force was assigned to the store. The results of this validation indicate that although the proper orthogonal decomposition (POD) surrogate model struggled to provide detailed predictions of store-distributed loads, mean load variations could be modeled well. The results further indicated that by leveraging the convolutional neural network (CNN)-based surrogate model, a significant improvement in distributed load modeling could be obtained. It was further identified that once distributed loads were integrated, both POD- and CNN-based surrogate models provided a viable path for the generation of a trajectory prediction-based surrogate model for rotorcraft applications. Both POD- and CNN-based surrogate models provided significantly reduced computational costs. For each rotorcraft store separation CFD simulation, the computational cost required 10 days of simulation time across 880 central processing unit cores. While using the surrogate model, comparable predictions could be produced in under 3 min on a single core.

Generic fully coupled framework for reliability assessment of offshore wind turbines under typical limit states

This work investigates surrogate modeling techniques for learning to approximate a computationally expensive function evaluation of 3D models. While in the past, 3D point clouds have been a data format that is too high dimensional for surrogate modeling, by leveraging advances in 3D object autoencoding neural networks, these point clouds can be mapped to a one-dimensional latent space. This leads to the fundamental research question: what surrogate modeling technique is most suitable for learning relationships between the 3D geometric features of the objects captured in the encoded latent vector and the physical phenomena captured in the evaluation software? Radial basis functions (RBFs), Kriging, and shallow 1D analogs of popular deep 2D image classification neural networks are investigated in this work. We find the nonintuitive result that departing from neural networks to decode latent representations of 3D objects into performance predictions is far more efficient than using a neural network decoder. In test cases using datasets of aircraft and watercraft 3D models, the non-neural network surrogate models achieve comparable accuracy to the neural network models. We find that an RBF surrogate model is able to approximate the lift and drag coefficients of 234 aircraft models with a mean absolute error of 1.97 × 10−3 and trains in only 3 seconds. Furthermore, the RBF surrogate model is able to rank a set of designs with an average percentile error of less than 8%. In comparison, a 1D ResNet achieves an average absolute error of 1.35 × 103 in 38 min for the same test case. We validate the comparable accuracy of the four techniques through a test case involving 214 3D watercraft models, but we also find that the distribution of the performance values of the data, in particular the presence of many outliers, has a significant negative impact on accuracy. These results contradict a common perception of neural networks as an efficient “one-size-fits-all” solution for learning black-box functions and suggests that even within systems that utilize multiple neural networks, potentially more efficient alternatives should be considered for each network in the system. Depending on the required accuracy of the application, this surrogate modeling approach could be used to approximate an expensive simulation software, or if the tolerance for error is low, it serves as a first pass which can narrow down the number of candidate designs to be analyzed more thoroughly.

A novel generalizable surrogate modeling approach is specifically developed for frequency reconfigurable antennas. The generalizable modeling processes is based on the rigorous mathematical derivation, including the solution of a non-linear overdetermined system, the optimization in the complex field, and the interpolation in multi-dimensional continuous space. As a post-processing method, the approach can convert the discrete data of CAD simulation to a surrogate model. Subsequently, a reconfigurable UWB antenna with a tunable notch-band is taken as an example to demonstrate that the surrogate modeling approach is feasible, effective, and precise. It also has the flexible ability to adapt to strict requirements and complicated scenarios. The proposed surrogate model is a good candidate for the interface standard between a reconfigurable antenna and signal processing part in a Cognitive Radio system.