Robotic manipulators are critical in many applications but are known to degrade over time. This degradation is influenced by the nature of the tasks performed by the robot. Tasks with higher severity, such as handling heavy payloads, can accelerate the degradation process. One way this degradation is reflected is in the position accuracy of the robot’s end-effector. In this article, we present a prognostic modeling framework that predicts a robotic manipulator’s remaining useful life (RUL) while accounting for the effects of task severity. Our framework represents the robot’s position accuracy as a Brownian motion process with a random drift parameter that is influenced by task severity. The dynamic nature of task severity is modeled using a continuous-time Markov chain (CTMC). To evaluate RUL, we discuss two approaches: 1) a novel closed-form expression for the residual life distribution (RLD) and 2) Monte Carlo (MC) simulations, commonly used in prognostics literature. Theoretical results establish the equivalence between these RUL computation approaches. We validate our framework through experiments using two distinct physics-based simulators for planar and spatial robot fleets. Our findings show that robots in both fleets experience shorter RUL when handling a higher proportion of high-severity tasks.

Advanced sensors and IoT devices have improved the monitoring and control of complex industrial enterprises. They have also created an interdependent fabric of geographically distributed process operations (clients) across these enterprises. Granger causality is an effective approach to detect and quantify interdependencies by examining how the state of one client affects the states of others over time. Understanding these interdependencies helps capture how localized events, such as faults and disruptions, can propagate throughout the system, potentially leading to widespread operational impacts. However, the large volume and complexity of industrial data present significant challenges in effectively modeling these interdependencies. This paper develops a federated approach to learning Granger causality. We utilize a linear state space system framework that leverages low-dimensional state estimates to analyze interdependencies. This helps address bandwidth limitations and the computational burden commonly associated with centralized data processing. We propose augmenting the client models with the Granger causality information learned by the server through a Machine Learning (ML) function. We examine the co-dependence between the augmented client and server models and reformulate the framework as a standalone ML algorithm providing conditions for its sublinear and linear convergence rates. We also study the convergence of the framework to a centralized oracle model. Moreover, we include a differential privacy analysis to ensure data security while preserving causal insights. Using synthetic data, we conduct comprehensive experiments to demonstrate the robustness of our approach to perturbations in causality, the scalability to the size of communication, number of clients, and the dimensions of raw data. We also evaluate the performance on two real-world industrial control system datasets by reporting the volume of data saved by decentralization.

We study a federated classification problem over a network of multiple clients and a central server, in which each client’s local data remains private and is subject to uncertainty in both the features and labels. To address these uncertainties, we develop a novel Federated Distributionally Robust Support Vector Machine (FDR-SVM), robustifying the classification boundary against perturbations in local data distributions. Specifically, the data at each client is governed by a unique true distribution that is unknown. To handle this heterogeneity, we develop a novel Mixture of Wasserstein Balls (MoWB) ambiguity set, naturally extending the classical Wasserstein ball to the federated setting. We then establish theoretical guarantees for our proposed MoWB, deriving an out-of-sample performance bound and showing that its design preserves the separability of the FDR-SVM optimization problem. Next, we rigorously derive two algorithms that solve the FDR-SVM problem and analyze their convergence behavior as well as their worst-case time complexity. We evaluate our algorithms on industrial data and various UCI datasets, whereby we demonstrate that they frequently outperform existing state-of-the-art approaches.

Modern fault diagnosis models that rely on machine learning and AI tools have been beset by three key challenges: first, highly uncertain training data due to sensor noise and potentially erroneous labeling by human operators, second, large numbers of fault classes, and third, highly imbalanced training data (normal versus fault data signatures). While various works in the literature attempt to address these challenges, they often leave at least one unaddressed. Indeed, some efforts leverage distributionally robust optimization to robustify common classifiers against data uncertainty. However, they commonly only focus on binary models, which can amplify data imbalance issues and fail to learn correlations between classes in multiclass problems. Alternatively, the existing multiclass diagnostic models do not commonly focus on hedging against data uncertainty and erroneous labels. Finally, data imbalance issues are commonly addressed via data augmentation or ensemble models. However, the former can lead to low-quality synthetic data, whereas the latter can be computationally expensive. This article introduces a Wasserstein distributionally robust multiclass support vector machine model designed to mitigate uncertainties in both data and label information. We utilize the multiclass Crammer–Singer (CS) loss to capture correlations between multiple classes. This also makes it more resilient to data imbalance. We prove key regularity properties of the CS loss and derive a convex and tractable reformulation and upper bound for the linear and kernel versions of our proposed model, respectively. Using extensive numerical experiments, we show that our model outperforms popular existing approaches, especially in settings with severe data imbalance, even when the synthetic minority oversampling technique is used to balance the data.

The optimization of preventive maintenance (PM) intervals traditionally follows predict-then-optimize (PTO) frameworks. These involve two sequential steps: training a statistical model to estimate the failure time distribution (FTD) and then integrating it into an optimization model for deciding the optimal PM interval. However, PTO models may have poor out-of-sample performance if the fitted FTD differs significantly from the true distribution or fails to capture covariate effects. To overcome these issues, this paper introduces a contextual distributionally robust optimization (DRO) model for computing PM intervals. The proposed model integrates empirical failure time data directly into the optimization framework without assuming specific distributions. Our setting assumes that the component FTD is affected by covariates. Therefore, our formulation seeks to exploit covariate knowledge to compute efficient PM decisions conditional on the observed covariates. We formulate a DRO model that accounts for potential misspecifications of the empirical FTD. This DRO formulation aims to minimize the long-term maintenance cost rate by optimizing PM decision policies over an infinite space, where these policies map covariate information to optimal PM intervals. We demonstrate that the proposed DRO model admits tractable mixed-integer linear programming reformulations in various practical cases. The efficacy of our model is demonstrated through computational studies involving simulated and real-world failure time data.

Nowadays, with the rapidly growing size of random orders and machine scales, cloud manufacturing (CMfg) is suffering from complex difficulties in scheduling enormous production services in real time. Especially when coupled with preventive maintenance (PM) scheduling for massive distributed machines, most current methods fail to promise agility and effectiveness simultaneously under this service-oriented paradigm. Therefore, this article aims to propose a collaborative scheduling algorithm to support real-time production and opportunistic maintenance (OM) in the highly dynamic CMfg environment. First, a novel collaborative scheduling model that combines real-time production and PM decision-making is formulated. Then, to address significant complexity in production scheduling, a dynamic suborder-oriented scheduling (DSS) strategy is designed to utilize frequent triggered events for allocating optimal manufacturing services in real time. Then, for improving computational capability in PM decision-making, a production-driven PM policy with two stage adjustment (TSA) is developed to adopt suborder changeovers as PM opportunities to formulate group PM decisions for massive machines. Finally, a collaborative DSS-TSA algorithm is designed to cyclically coordinate the real-time production scheduling with OM decision-making to avoid repetitively solving large-scale joint problems. Our collaborative scheduling algorithm is verified in a real CMfg scenario for gas turbine blades. The results prove the significant improvement in production timeliness, machine availability, and system profitability.

Large-scale power systems are composed of regional utilities with assets that stream sensor readings in real time. In order to detect cyberattacks, the globally acquired, real-time sensor data needs to be analyzed in a centralized fashion. However, owing to operational constraints, such a centralized sharing mechanism turns out to be a major obstacle. In this article, we propose a blockchain-based decentralized framework for detecting coordinated replay attacks with full privacy of sensor data. We develop a Bayesian inference mechanism employing locally reported attack probabilities that is tailor made for a blockchain framework. We compare our framework to a traditional decentralized algorithm based on the broadcast gossip framework both theoretically as well as empirically. With the help of experiments on a private Ethereum blockchain, we show that our approach achieves good detection quality and significantly outperforms gossip-driven approaches in terms of accuracy, timeliness, and scalability.

Supervisory control and data acquisition (SCADA) systems are widely used for industrial control of critical infrastructures, such as power plants and manufacturing systems. There is abundant evidence of SCADA systems being subject to cyberattacks. With increasing interest in industrial digitization, the cybersecurity of SCADA systems is poised to be even more important. Equipment faults and cyberattacks can manifest themselves in a similar fashion, i.e., they can exhibit similar signatures. This article focuses on methods that are capable of distinguishing equipment faults from bona fide cyberattacks. Especially, we consider a relatively sophisticated form of cyberattack known as the “replay attack” (RA). We derive mathematical formalisms that distinguish the RA from several classes of equipment faults and verify our methodology through an extensive numerical study.Note to Practitioners—This article is motivated by the problem of detecting replay cyberattacks in industrial control systems and differentiating it from equipment faults. Existing approaches mainly focus on the detection aspect but usually ignore the importance of differentiation. We an ensembled statistical process monitoring approach based on five statistical metrics. The statistical metrics are derived based on a theoretical analysis that shows the data characteristics under each system anomaly, including replay attack (RA), controller fault, and plant fault. We mathematically prove that the signatures generated by the derived metrics can be used to differentiate an RA from the equipment faults. We conduct a sensitivity analysis of the detection delay of our method regarding the magnitude of the cyberattack. Physical experiments on a rotating machinery setup show that the proposed approach applies to some simple real-world settings. In future research, we will address the scalability issue of our method as well as more generalized nonlinear system settings.

Cybersecurity of Industrial Control Systems (ICS) is drawing significant concerns as data communication increasingly leverages wireless networks. A lot of data-driven methods were developed for detecting cyberattacks, but few are focused on distinguishing them from equipment faults. In this paper, we develop a data-driven framework that can be used to detect, diagnose, and localize a type of cyberattack called covert attacks on smart grids. The framework has a hybrid design that combines an autoencoder, a recurrent neural network (RNN) with a Long-Short-Term-Memory (LSTM) layer, and a Deep Neural Network (DNN). This data-driven framework considers the temporal behavior of a generic physical system that extracts features from the time series of the sensor measurements that can be used for detecting covert attacks, distinguishing them from equipment faults, as well as localize the attack/fault. We evaluate the performance of the proposed method through a realistic simulation study on the IEEE 14-bus model as a typical example of ICS. We compare the performance of the proposed method with the traditional model-based method to show its applicability and efficacy.

The integration of condition monitoring with queueing systems to support decision making is not well explored. This paper addresses the impact of condition monitoring of the server on the system-level performance experienced by entities in a queueing system. The system consists of a queue with a single-server subject to Markovian degradation. The model assumes a Poisson arrival process with service times and repair times according to general distributions. We develop stability conditions and perform steady-state analysis to obtain performance measures (average queue length, average degradation, and so on). We propose minimizing an objective function involving four types of costs: repair, catastrophic failure, quality, and holding. The queue performance measures derived from steady-state analysis are benchmarked and compared to those from a discrete event simulation model. After verifying the queuing model, a sensitivity analysis is performed to determine the relationships between system performance and model parameters. Results indicate that the total cost function is convex and, thus, subject to an optimal repair policy. The model is sensitive to service time, quality costs, and failure costs for late-stage policy repairs decisions and sensitive to expected repair times and repair costs for early stage policy repair decisions.

Multistage manufacturing processes (MMPs) usually exhibit an interactive relationship between tool wear and product quality degradation. On one hand, the tool wear in a stage may result in the quality degradation of the products fabricated on that stage. On the other hand, the quality degradation at a preceding stage may lead to the change of the operational condition and thus affect the tool wear in subsequent stages. This interaction needs to be considered to accurately predict the residual life distribution (RLD) of MMPs, which will benefit condition-based maintenance and tool inventory management. In this paper, we propose an interaction model that utilizes a linear model to represent the impact of tool wear on quality degradation and a stochastic differential equation model to capture the impact of quality degradation on the instantaneous rate of tool wear. We then propose a Bayesian framework that incorporates real-time quality measurements to online update the RLD of MMPs. Our methodology is a generalization of an existing “QR-chain model,” which is dedicated into a similar research and application area. We conduct numerical studies to test the performance of our methodology and compare with the QR-chain model. The results show that our methodology outperforms the QR-chain model through capturing the impact of quality degradation on the process of tool wear and incorporating real-time quality measurements.

Complex systems often consist of multiple units that are required to work together in parallel to satisfy a specific engineering objective. As an example, in manufacturing processes, several identical machines may need to operate together to simultaneously fabricate the same products in order to meet the high production demand. This parallel configuration is often designed with some level of redundancy to compensate for unexpected events. In this way, when only a small portion of units fail to operate due to either unexpected machine downtime or scheduled maintenance, the remaining units can still achieve the engineering objective by increasing their workloads up to the designed capacities. However, the workload of a unit apparently impacts the unit's degradation rate as well as its failure time. Specifically, this paper considers the case that a higher workload assignment accelerates the unit's degradation and vice versa. Based on this assumption, we develop a method to actively control the degradation as well as the predicted failure time of each unit by dynamically adjusting its workloads. Our goal is to prevent the overlap of unit failures within a certain time period through taking advantage of the natural redundancy of the parallel structure, which may potentially lead to a better utilization of maintenance resources as well as a consistently ensured system throughput. A numerical study is used to evaluate the performance of the proposed method under different scenarios.

Prognostics involves the effective utilization of condition or performance-based sensor signals to accurately estimate the remaining lifetime of partially degraded systems and components. The rapid development of sensor technology, has led to the use of multiple sensors to monitor the condition of an engineering system. It is therefore important to develop methodologies capable of integrating data from multiple sensors with the goal of improving the accuracy of predicting remaining lifetime. Although numerous efforts have focused on developing feature-level and decision-level fusion methodologies for prognostics, little research has targeted the development of “data-level” fusion models. In this paper, we present a methodology for constructing a composite health index for characterizing the performance of a system through the fusion of multiple degradation-based sensor data. This methodology includes data selection, data processing, and data fusion steps that lead to an improved degradation-based prognostic model. Our goal is that the composite health index provides a much better characterization of the condition of a system compared to relying solely on data from an individual sensor. Our methodology was evaluated through a case study involving a degradation dataset of an aircraft gas turbine engine that was generated by the Commercial Modular Aero-Propulsion System Simulation (C-MAPSS).

This paper considers a fleet of identical units where each unit consists of the same critical components. The degradation state of each critical component is assumed to be monitored by an on-board sensor. The paper presents a methodology for identifying the top-k(thekmost reliable) units in a fleet using sensor-based prognostic information. Specifically, we develop a prognostics-based ranking (PBR) algorithm that combines stochastic degradation models with computer science database ranking algorithms. The stochastic degradation modeling framework is used to compute and update, in real-time, residual life distributions (RLDs) of the critical components of each unit. Using a base case exponential degradation model, we identify conditions necessary to establish stochastic ordering among the RLDs of similar components. A preference relationship, consistent with the stochastic ordering results, is then used to sort the units of the fleet based on the RLDs of their respective components. A database ranking algorithm, known as the threshold algorithm (TA), is then used to identify the top-kunits without necessarily computing all the RLDs. The paper concludes with an illustrative example.

Recent developments in degradation modeling have been targeted towards utilizing degradation-based sensory signals to predict residual life distributions. Typically, these models consist of stochastic parameters that are estimated with the aid of an historical database of degradation signals. In many applications, building a degradation database, where components are run-to-failure, may be very expensive and time consuming, as in the case of generators or jet engines. The degradation modeling framework presented herein addresses this challenge by utilizing failure time data, which are easier to obtain, and readily available (relative to sensor-based degradation signals) from historical maintenance/repair records. Failure time values are first fitted to a Bernstein distribution whose parameters are then used to estimate the prior distributions of the stochastic parameters of an initial degradation model. Once a complete realization of a degradation signal is observed, the assumptions of the initial degradation model are revised and improved for future predictions. This approach is validated using real world vibration-based degradation information from a rotating machinery application.

This paper presents a sensory-updated degradation-based predictive maintenance policy (herein referred to as the SUDM policy). The proposed maintenance policy utilizes contemporary degradation models that combine component-specific real-time degradation signals, acquired during operation, with degradation and reliability characteristics of the component's population to predict and update the residual life distribution (RLD). By capturing the latest degradation state of the component being monitored, the updating process provides a more accurate of the remaining life. With the aid of a stopping rule, maintenance routines are scheduled based on the most recently updated RLD. The performance of the proposed maintenance policy is evaluated using a simulation model of a simple manufacturing cell. Frequency of unexpected failures and overall maintenance costs are computed and compared with two other benchmark maintenance policies: a reliability-based and a conventional degradation-based maintenance policy (without any sensor-based updating).

The ability to accurately estimate the residual life of partially degraded components is arguably the most challenging problem in prognostic condition monitoring. This paper focuses on the development of a neural network-based degradation model that utilizes condition-based sensory signals to compute and continuously update residual life distributions of partially degraded components. Initial predicted failure times are estimated through trained neural networks using real-time sensory signals. These estimates are used to derive a prior failure time distribution for the component that is being monitored. Subsequent failure time estimates are then utilized to update the prior distributions using a Bayesian approach. The proposed methodology is tested using real world vibration-based degradation signals from rolling contact thrust bearings. The proposed methodology performed favorably when compared to other reliability-based and statistical-based benchmarks.

This paper presents a degradation modeling framework for computing condition-based residual life distributions of partially degraded systems and/or components functioning under time-varying environmental and/or operational conditions. Our approach is to mathematically model degradation-based signals from a population of components using stochastic models that combine three main sources of information: real-time degradation characteristics of component obtained by observing the component's in-situ degradation signal, the degradation characteristics of the component's population, and the real-time status of the environmental conditions under which the component is operating. Prior degradation information is used to estimate the model coefficients. The resulting generalized stochastic degradation model is then used to predict an initial residual life distribution for the component being monitored. In-situ degradation signals, along with real-time information related to the environmental conditions, are then used to update the residual life distributions in real-time. Because these updated distributions capture current health information and the latest environmental conditions, they provide precise lifetime estimates. The performance of the proposed models is evaluated using real world vibration-based degradation signals from a rotating machinery application.

This paper develops an integrated neural-network-based decision support system for predictive maintenance of rotational equipment. The integrated system is platform-independent and is aimed at minimizing expected cost per unit operational time. The proposed system consists of three components. The first component develops a vibration-based degradation database through condition monitoring of rolling element bearings. In the second component, an artificial neural network model is developed to estimate the life percentile and failure times of roller bearings. This is then used to construct a marginal distribution. The third component consists of the construction of a cost matrix and probabilistic replacement model that optimizes the expected cost per unit time. Furthermore, the integrated system consists of a heuristic managerial decision rule for different scenarios of predictive and corrective cost compositions. Finally, the proposed system can be applied in various industries and different kinds of equipment that possess well-defined degradation characteristics

Research on interpreting data communicated by smart sensors and distributed sensor networks, and utilizing these data streams in making critical decisions stands to provide significant advancements across a wide range of application domains such as maintenance management. In this paper, a stochastic degradation modeling framework is developed for computing and continuously updating residual life distributions of partially degraded components. The proposed degradation methodology combines population-specific degradation characteristics with component-specific sensory data acquired through condition monitoring in order to compute and continuously update remaining life distributions of partially degraded components. Two sensory updating procedures are developed and validated using real-world vibration-based degradation information acquired from rolling element thrust bearings. The results are compared with two benchmark policies and illustrate the benefits of the sensory updated degradation models proposed in this paper. Note for Practitioners-The proposed degradation-based prognostic methodology provides a comprehensive assessment of the current and future degradation states of partially degraded components by combining population-specific degradation or reliability information with real-time sensory health monitoring data. It is specifically beneficial for cases where degradation occurs in a cumulative manner and the degradation signal can be approximated by an exponential functional form. To implement this methodology, it is necessary: 1) to identify the physical phenomena associated with the evolution of the degradation process (spalling and wear herein); 2) choose the appropriate condition monitoring technology to monitor this phenomena (accelerometers); 3) identify a characteristic pattern in the sensory information to help develop a degradation signal (exponential growth); and 4) identify a failure threshold associated with the degradation signal. The first step in implementing this prognostic methodology is to obtain prior information related to stochastic parameters f the exponential model. This may require fitting some sample degradation signals with an exponential functional form and noting the values of the exponential parameters, or using subjective prior distributions. The second step is to acquire sensory information and begin updating the prior distribution. The updating frequency will dictate which expressions are used to compute the posterior distributions. Once the posterior means, variances, and correlation are computed, the truncated CDF of the residual life can be evaluated using (10) and (11). Note that the truncation is necessary to preclude negative values of the remaining life. Practitioners can implement this methodology using a simple spreadsheet. Since the residual life distributions are skewed, it is reasonable to utilize the median as a measure of the central tendency and, hence, an alternative estimate for the expected value of the remaining life

Automated tool sharing systems provide a technological response to the high cost of tools in flexible manufacturing systems. These systems allow different machines to use the same tools by automatically transferring them from machine to machine as tooling needs evolve. With these systems, tool allocation is a real-time issue that requires sophisticated control techniques to make the right allocation decisions. An essential property that tool sharing policies must guarantee is deadlock-free operation. Although manufacturing researchers have investigated the performance aspects of tool sharing through simulation, no work has yet addressed deadlock handling strategies for these real-time systems. In this paper, we characterize the structural and computational properties of the tooling deadlock problem. We develop polynomial algorithms that detect and avoid deadlock, and we investigate the safety implications of special structures appearing in tool sequences.