Narrow Results

Due to the rapid advancement of computational power and the Internet of Things (IoT) together with recent developments in digital-twin and cyber-physical systems, the integration of big data analytics techniques (BDAT) (e.g., data mining, machine learning (ML), deep learning, data farming, etc.) into traditional simulation methodologies, along with the enhancement of past simulation optimization techniques, has given rise to an emerging field known as simulation learning and optimization (SLO). Unlike simulation optimization, SLO is a data-driven fusion approach that integrates conventional simulation modeling methodologies, simulation optimization techniques, and learning-based big data analytics (BDA) to enhance the ability to address stochastic and dynamic decision-making problems in the real-world complex system and quickly find the best outcomes for the decision support in both real-time and nonreal-time analyses to achieve the data-driven digital twin capability. Although some literature in the past has mentioned similar applications and concepts, no paper provides a structural explanation and comprehensive review of relevant literature on SLO from both methodological and applied perspectives. Consequently, in the first part of this paper, we conduct a literature review on a novel SLO methodology emerging from the fusion of traditional simulation methodology, simulation optimization algorithms, and BDAT. In the second part of this paper, we review the applications of SLO in various contexts, with detailed discussions on manufacturing, maintenance, redundancy allocation, hybrid energy systems, humanitarian logistics, and healthcare systems. Lastly, we investigate potential research directions for SLO in this new era of big data and artificial intelligence (AI).

The redundancy allocation problem (RAP) is an important reliability optimization problem and has been an active area of research for the past decades. In the literature, most of these system reliability problems are considered in series–parallel where the reliability of each component is assumed a known constant. However, in reality, the component reliability may exhibit quality variations and the reliability of each component shall be better considered stochastic. This motivates this paper to extend the RAP to a more realistic situation in which the reliability of each component is stochastic in nature and the system can have a complex network structure, i.e., the system components are connected with each other neither in series nor in parallel but in some logical relationship. Due to the complex network structure and aggregated profound stochasticity, the overall system reliability is not analytically available but instead has to rely on simulation for estimation. We therefore develop a Particle-Swarm-based Simulation Optimization (PSSO) method to enable the optimal solution to be identified efficiently. An extensive numerical study verifies the efficacy and efficiency of our proposed method in realistic settings and is worth further investigation.

This paper proposes one new stochastic approximation algorithm for solving simulation-based optimization problems. It employs a weighted combination of two independent current noisy gradient measurements as the iterative direction. It can be regarded as a stochastic approximation algorithm with a special matrix step size. The almost sure convergence and the asymptotic rate of convergence of the new algorithm are established. Our numerical experiments show that it outperforms the classical Robbins–Monro (RM) algorithm and several other existing algorithms for one noisy nonlinear function minimization problem, several unconstrained optimization problems and one typical simulation-based optimization problem, i.e., (s, S)-inventory problem.

Recent advances in simulation optimization research and explosive growth in computing power have made it possible to optimize complex stochastic systems that are otherwise intractable. In the first part of this paper, we classify simulation optimization techniques into four categories based on how the search is conducted. We provide tutorial expositions on representative methods from each category, with a focus in recent developments, and compare the strengths and limitations of each category. In the second part of this paper, we review applications of simulation optimization in various contexts, with detailed discussions on health care, logistics, and manufacturing systems. Finally, we explore the potential of simulation optimization in the new era. Specifically, we discuss how simulation optimization can benefit from cloud computing and high-performance computing, its integration with big data analytics, and the value of simulation optimization to help address challenges in engineering design of complex systems.

Simulation optimization can be used to solve many complex optimization problems in automation applications such as job scheduling and inventory control. We propose a new framework to perform efficient simulation optimization when simulation models with different fidelity levels are available. The framework consists of two novel methodologies: ordinal transformation (OT) and optimal sampling (OS). The OT methodology uses the low-fidelity simulations to transform the original solution space into an ordinal space that encapsulates useful information from the low-fidelity model. The OS methodology efficiently uses high-fidelity simulations to sample the transformed space in search of the optimal solution. Through theoretical analysis and numerical experiments, we demonstrate the promising performance of the multi-fidelity optimization with ordinal transformation and optimal sampling (MO²TOS) framework.

Real-time decision making has acquired increasing interest as a means to efficiently operating complex systems. The main challenge in achieving real-time decision making is to understand how to develop next generation optimization procedures that can work efficiently using: (i) real data coming from a large complex dynamical system, (ii) simulation models available that reproduce the system dynamics. While this paper focuses on a different problem with respect to the literature in RL, the methods proposed in this paper can be used as a support in a sequential setting as well. The result of this work is the new Generalized Ordinal Learning Framework (GOLF) that utilizes simulated data interpreting them as low accuracy information to be intelligently collected offline and utilized online once the scenario is revealed to the user. GOLF supports real-time decision making on complex dynamical systems once a specific scenario is realized. We show preliminary results of the proposed techniques that motivate the authors in further pursuing the presented ideas.

We propose a random search algorithm for seeking the global optimum of an objective function in a simulation setting. The algorithm can be viewed as an extension of the MARS algorithm proposed in Hu and Hu (2011) for deterministic optimization, which iteratively finds improved solutions by modifying and sampling from a parameterized probability distribution over the solution space. However, unlike MARS and many other algorithms in this class, which are often population-based, our method only requires a single candidate solution to be generated at each iteration. This is primarily achieved through an effective use of past sampling information by means of embedding multiple nested stochastic approximation type of recursions into the algorithm. We prove the global convergence of the algorithm under general conditions and discuss two special simulation noise cases of interest, in which we show that only one simulation replication run is needed for each sampled solution. A preliminary numerical study is also carried out to illustrate the algorithm.

Traditionally, simulation has been used as a tool of design to estimate, compare and optimize the performance of different system designs. It is rarely used in making real-time decisions due to the long computation delay of executing simulation models. However, with the fast growth of computing capability, we have observed more and more works on reusing simulation efforts for repeated experiments with the help of data analytics tools, and the target of these works is to solve real-time decision problems. In this paper, we distill the important features of these works and summarize a new simulation framework, called offline-simulation-online-application (OSOA) framework, which treats simulation as a data generator, applies state-of-the-art analytics tools to build predictive models, and then uses the predictive models for real-time applications. In this paper, we illustrate how to apply the OSOA framework on estimation, ranking and selection and simulation optimization, and provide a prospect of this new framework.

We consider how to allocate simulation budget to estimate the risk measure of a system in a two-stage simulation optimization problem. In this problem, the first stage simulation generates scenarios that serve as inputs to the second stage simulation. For each sampled first stage scenario, the second stage procedure solves a simulation optimization problem by evaluating a number of decisions and selecting the optimal decision for the scenario. It also provides the estimated performance of the system over all sampled first stage scenarios to estimate the system’s reliability or risk measure, which is defined as the probability of the system’s performance exceeding a given threshold under various scenarios. Usually, such a two-stage procedure is very computationally expensive. To address this challenge, we propose a simulation budget allocation procedure to improve the computational efficiency for two-stage simulation optimization. After generating first stage scenarios, a sequential allocation procedure selects the scenario to simulate, followed by an optimal computing budget allocation scheme that determines the decision to simulate in the second stage simulation. Numerical experiments show that the proposed procedure significantly improves the efficiency of the two-stage simulation optimization for estimating system’s reliability.

While nearly all previous algorithms designed to solve simulation optimization problems have treated the outputs of simulation systems at a given design point (input parameter) as being independent of each other, this premise is flawed in that simulated outputs are generally correlated. We propose a decorrelation (DC) procedure that can effectively evaluate and remove the correlation of outputs of a simulation system. The proposed DC procedure is further integrated with STRONG, an improved framework of the well-known Response Surface Methodology (RSM), for tackling the simulation optimization problems with correlated outputs. This integration is particularly synergistic due to the fact that STRONG is a fully automated, response-surface-based procedure possessing appealing convergence properties and DC can take advantage of the concept of trust region as in STRONG to enable the removal of the correlation of outputs at the design points within the same trust region all at once. This is more efficient compared to the traditional approaches where a substantial number of observations are typically required for dealing with correlations. The resulting integrated method, which we call STRONG-DC, requires various adaptations so as to ensure the efficacy and efficiency of the overall framework. STRONG-DC preserves the desirable automation and convergence as STRONG, namely, it does not require human involvements and can be proved to achieve the truly optimal solution(s) with probability one (w.p.1) under reasonable conditions. Moreover, the effectiveness and efficiency of STRONG-DC are evaluated through extensive numerical analyses, along with a case study involving the well-known newsvendor problem.

Automated material handling systems (AMHS) have been widely used in semiconductor manufacturing. However, the performance of AMHS heavily hinges on vehicle fleet sizing, which is a complex yet crucial problem. For example, a small fleet size may increase the average wait time, but a large fleet size can also result in traffic congestion. This tradeoff is difficult and can be further exacerbated by profound uncertainty in the manufacturing process. In the literature, the existing models are focused on improving the mean-based performance of AMHS, where the resulting optimal vehicle fleet size is fixed, lacking the ability and flexibility to respond to the changes and/or special requirements that suddenly come up in the manufacturing process. Another drawback with the existing models is that they are not able to characterize the upside/downside risks associated with the resulting vehicle fleet size. This paper, motivated by a real project, presents a novel quantile-based decision model to fill the gap. The adjustment of $\alpha$ values in the proposed decision model allows for agile vehicle fleet sizing according to the production situations, resulting in the satisfactory performance of AMHS. We develop a simulation optimization solution method, called ES-AMHS in short, to enable the efficient derivation of the optimal vehicle fleet size. A comprehensive numerical analysis is conducted to evaluate the efficiency and efficacy of the solution method. Finally, an empirical study in cooperation with a wafer fab in Taiwan is presented to show the practical usefulness of this methodology in a real-world setting.

In order to control the charged jet and deposition area of the glass frit in the electrostatic spray system, the nozzles with/without sleeve-type auxiliary electrode are comparatively studied through simulations and experiments. The influence of the voltage applied on the auxiliary electrode and its diameter is analyzed. According to the vector diagram and the distribution of electric field intensity, the optimal values for the applied voltage and the diameter of the auxiliary electrode are determined for the present work, i.e. 1.5 kV and 6 mm, respectively. The experimental results further demonstrate the nozzle with the auxiliary electrode is much better than the one without the electrode, and under the above optimized condition the charged jet can be well controlled to deposit glass frit film on silicon wafer with a line width of 330 μm, which is much more narrow than the conventional nozzle (without the electrode). This work will offer another good way for better control of the deposition during electrostatic spray process.

Simulation is a powerful technique in evaluating and improving the performance of complex systems. In order to improve the efficiency of simulation experiment design, analysis and evaluation, auxiliary tools are required. Unfortunately, existing tools are usually not meeting the requirements of simulation. Moreover, the restricted interfaces, reusability and expandability influence their efficiency to a certain extent. In this paper, an integrated software environment, HIT-SEDAES, is designed for solving these problems. A process model of simulation experiment design, analysis and evaluation is introduced to guide the development of the software environment. And several solutions are proposed to solve key problems in this development. Finally, an application is used to illustrate how the software environment works for the problems of model validation, effectiveness evaluation and performance analysis.