Speakers - Workshop

Engineering design often involves qualitative and quantitative design variables, which requires systematic methods for the exploration of these mixed-variable design spaces. Existing machine learning (ML) models that can handle mixed variables as inputs require a large amount of data but do not provide uncertainty quantification that is crucial for sequential (adaptive) design of experiments. We have developed a novel Latent Variable Gaussian Process (LVGP) based ML approach that involves a latent variable (LV) representation of qualitative inputs, and automatically discovers a categorical-to-numerical nonlinear map that transforms the underlying high dimensional physical attributes into the LV space. The nonlinear mapping also provides an inherent ordering and structure for the levels of the qualitative factor(s), which leads to substantial insight and interpretable ML. In addition, LVGP provides uncertainty quantification of prediction which is critical for adaptive sampling to sequentially choose samples based on current observations and the method also offers easy integration with Bayesian optimization (BO) or other reinforcement learning strategies for the purpose of design optimization. Although recent developments in mixed-variable Bayesian optimization have shown promise, high dimensional qualitative variables, i.e., with many levels, impose a large design cost as they typically require a larger dataset to quantify the effect of each level on the optimization objective. We develop a descriptor aided Bayesian optimization approach to address this challenge by leveraging domain knowledge about underlying physical descriptors. We show that physical descriptors can be intuitively embedded into the LVGP approach and used to selectively explore levels of qualitative variables in the Bayesian optimization framework. Using material design examples, we will show that our LVGP approach is robust to certain types of incomplete domain knowledge and significantly reduces the design cost for problems with high-dimensional qualitative variables.

For some systems, degrading components within the system can be grouped into clusters, which are subsets of components exhibiting similar dependency patterns. A complex system could have one or several of these clusters, and components can be partially a member of a cluster or even multiple clusters. Models for system reliability with dependent components are reviewed and new design and maintenance optimization problems are introduced. System reliability optimization analyses involving multiple failure processes are already challenging research topics, but particularly when the failure processes, such as degradation failures and random shocks, are competing and dependent. When component degradation models are extended to complex systems with multiple components, different perspectives and models are needed considering this dependency. Previous models for dependent failure and/or degradation processes can be categorized as Markov chains, joint multivariate distributions copula functions or shared shock exposure models. These different approaches for modeling the reliability for systems with dependent failure processes or degradation processes are reviewed, and new models are presented based on extended random effects gamma processes or a superposition of independent gamma processes. Each approach has specific advantages considering model fidelity, precision, ability to generalize or extend and practicality. Based on these new extended gamma process models, reliability, maintenance, and combined reliability/maintenance optimization models are presented to provide cost effective maintenance plans, and some preliminary results presented.

The pace of the developments of new materials, advances in manufacturing processes and sensors technology and changes in environments have resulted in the introduction of new products with unique designs and configurations. In this workshop we present approaches to reliability testing and estimations under these conditions. Reliability and resilience of systems under natural and man-made interruptions of the system’s functions will be addresses. Finally, the human longevity poses longevity new medical devices and human body parts replacements with “artificial” ones, their manufacturing, quality and reliability pose challenging research issues in the reliability engineering field. This presentation addresses key challenges in reliability modeling, estimation and testing approaches for such products and highlights key challenges that require investigation by the reliability engineering community.

In Fisher (1971), a lady was able to distinguish (by tasting) from whether the tea or the milk was first added to the cup. This is probably the first popular Order of Addition (OofA) experiment. In general, there are m required components and we hope to determine the optimal sequence for adding these m components one after another. It is often unaffordable to test all the m! treatments (for example, m!=10! is about 3.5 millions), and the design problem arises. We consider the model in which the response of a treatment depends on the pairwise orders of the components. The optimal design theory under this model is established, and the optimal values of the D-, A-, E-, and M/S-criteria are derived. For Model-Free approach, an efficient sequential methodology is proposed, building upon the basic concept of quick-sort algorithm, to explore the optimal order without any model specification. The proposed method is capable to obtain the optimal order for large m (≥ 20). This work can be regarded as an early work of OofA experiment for large number of components. Some theoretical supports are also discussed. One case study for job scheduling will be discussed in detail.

A multistage manufacturing process (MMP) refers to a system consisting of multiple units, stations, or operations to finish a final product. The quality of the final product is a result of complex interactions among multiple stages. In other words, the quality characteristics of one stage are not only influenced by local variations at that stage but also by variations propagated from upstream stages. With the advancement of sensing and computing technologies, the complexity of in-situ sensor outputs has been evolving, from univariate data to multivariate data, to functional curves, to images, to 3D point cloud data, and to high-resolution videos. At the same time, data science methods employed have also been evolving dramatically from the early time when PCA, clustering and classification, and state space models were developed, to the later time when Bayesian models and big data analysis were used, and to the most recent time when one sees the application of tensor-based models, Koopman operator theory, and numerous other machine learning methods. This talk will tell the story behind the research journey in the past 30 years for modeling and analysis of MMP. It will discuss the co-evolution of sensing technologies and data science, and how their evolution drives and propels the advancement in modeling and analysis of MMP, how those methodological developments models provide bases in addressing fundamental issues in variation modeling, tolerance synthesis, distributed sensing, root cause diagnosis, and error compensation for system variation reduction; how those R&D efforts motivated by real engineering problems. This presentation provides a summary of major milestones in modeling and analysis of MMP, how it evolved with sensing and data science, and where it was implemented. Some ideas about research directions will be discussed as well.

This presentation provides a relatively non-technical overview of the development of statistical process monitoring methods from their introduction by Walter Shewhart in the Industry 2.0 era to the challenges now faced in Industry 4.0. Monitoring methods have been adapted and extended over time to reflect increases in the amount of available data, changes in data structure and characteristics, and increases in computing power. It is argued that it is increasingly the case in practice that monitoring methods should be desensitized, counter to the prevailing research quest for steadily increasing sensitivity in detecting any process change however small it may be. Current research directions will be discussed along with ideas for future work. In line with the goals of the conference, some tips for doing research and publishing results will be shared.

Reference: Jones-Farmer, L. A., Paynabar, K., Ranjan, C., and Woodall, W. H. (2023). “Statistical Process Monitoring from Industry 2.0 to Industry 4.0: Insights into Research and Practice.” Submitted to Technometrics.

We present a new method, called analysis-of-marginal-tail-means (ATM), for effective robust optimization of discrete black-box problems. ATM has important applications in many real-world engineering problems (e.g., manufacturing optimization, product design, and molecular engineering), where the objective to optimize is black-box and expensive, and the design space is inherently discrete. One weakness of existing methods is that they are not robust: these methods perform well under certain assumptions, but yield poor results when such assumptions (which are difficult to verify in black-box problems) are violated. ATM addresses this by combining both rank- and model-based optimization, via the use of marginal tail means. The trade-off between rank- and model-based optimization is tuned by first identifying important main effects and interactions from data, then finding a good compromise which best exploits additive structure. ATM provides improved robust optimization over existing methods, particularly in problems with (i) a large number of factors, (ii) unordered factors, or (iii) experimental noise. We demonstrate the effectiveness of ATM in simulations and in two real-world engineering problems: the first on robust parameter design of a circular piston, and the second on product family design of a thermistor network.