Unreliable Machine Research Articles

A New Approach to Capacity Scaling Augmented with Unreliable Machine Learning Predictions

Modern data centers suffer from immense power consumption. As a result, data center operators have heavily invested in capacity-scaling solutions, which dynamically deactivate servers if the demand is low and activate them again when the workload increases. We analyze a continuous-time model for capacity scaling, where the goal is to minimize the weighted sum of flow time, switching cost, and power consumption in an online fashion. We propose a novel algorithm, called adaptive balanced capacity scaling (ABCS), that has access to black-box machine learning predictions. ABCS aims to adapt to the predictions and is also robust against unpredictable surges in the workload. In particular, we prove that ABCS is [Formula: see text] competitive if the predictions are accurate, and yet, it has a uniformly bounded competitive ratio even if the predictions are completely inaccurate. Finally, we investigate the performance of this algorithm on a real-world data set and carry out extensive numerical experiments, which positively support the theoretical results. Funding: This work was partially supported by the Division of Computing and Communication Foundations [Grant 2113027]. The authors also acknowledge financial support for this project from the Algorithm and Randomness Center–Transdisciplinary Research Institute for Advancing Data Science Fellowship at Georgia Tech.

Reliability and Sensitivity Analyses of an Unreliable Retrial Machine Repair System with Warm Standby Switching Failure

In this study, an unreliable machine repair system with warm standby switching failure and retrial behavior is examined. An application of the Ceph storage cluster system is described to illustrate the applicability of the proposed model and establish its results. To analyze the reliability of the application system, two-state-transition rate diagrams and their associated differential-difference equations are provided. Laplace transforms and the matrix-analytical method are employed to obtain the steady-state probabilities. Explicit expressions for the reliability function, mean time to failure, and steady-state availability of the examined system are then derived. With these system performance measures, we perform a sensitivity analysis of the system parameters to identify critical variables. Furthermore, the numerical results for various parameter settings are tabulated and presented graphically. Finally, managerial insights and constructive results are provided as a guide for managers in determining an effective strategy for performance enhancement.

Young children's use of probabilistic reliability and base-rates in decision-making.

Children are skilled reasoners who readily use causal, reliability, and base-rate (i.e., prior probability) information in their decisions. Though these abilities are typically studied in isolation, children often must consider multiple pieces of information to make an informed decision. Four experiments (N = 320) explored the development of children’s ability to use reliability and base-rate information when making decisions about draw outcomes. Experiment 1 examined the age at which children can first compare and choose between probabilistically reliable machines. Three- and 4-year-old children saw machines that were probabilistically reliable at obtaining objects while sampling from uniform distributions (i.e., all target or non-target objects). Although 4-year-old children correctly used reliability in their decisions, 3-year-olds did not. In Experiment 2a, 4- to 6-year-olds were presented with the same probabilistically reliable machines, although they sampled from a mixture of target and non-target items. Here, children tended to choose the machine with the better proportion of targets, regardless of reliability. This was replicated in Experiment 2b. In Experiment 3, children were presented with one perfectly reliable machine and one probabilistically unreliable machine. Here, children continued to mostly choose the machine with the better proportion of targets. These results raise questions about base-rate overuse early in development and highlight the need for additional work on children’s ability to use multiple pieces of information in decision-making.

A multi-station unreliable machine model with working vacation policy and customers’ impatience

ABSTRACT Working vacation queues with breakdowns and customers impatience have many applications in different real-life situations. The development of these models to determine their performance is extremely important. In this paper, we deal with a finite population Markovian multi-server machine system with breakdowns, repairs, Bernoulli feedback, balking, reneging, and retention of reneged customers, under multiple synchronous working vacations. The investigated model has a potential application in the real-world machine systems, such as the automated manufacturing systems with limited space. For the analysis purpose, we employ Q-matrix method in order to obtain the steady-state probabilities as well as closed-form expressions for several system characteristics. Particular cases of the current study are presented. We construct the expected cost function and develop an optimization problem to determine the optimum cost. The direct search method and the Quasi–Newton method are applied to find the optimum system capacity, the minimum number of servers, and the optimum service rates during both working vacation and regular busy periods at minimum cost. Lastly, a sensitivity analysis for the numerical simulation is carried out.

Capacity Scaling Augmented With Unreliable Machine Learning Predictions

Modern data centers suffer from immense power consumption. As a result, data center operators have heavily invested in capacity scaling solutions, which dynamically deactivate servers if the demand is low and activate them again when the workload increases. We analyze a continuoustime model for capacity scaling, where the goal is to minimize the weighted sum of flow-time, switching cost, and power consumption in an online fashion. We propose a novel algorithm, called the Adaptive Balanced Capacity Scaling (ABCS) algorithm, that has access to black-box machine learning predictions. ABCS aims to adapt to the predictions and is also robust against unpredictable surges in the workload. In particular, we prove that the ABCS algorithm is (1 + ")-competitive if the predictions are accurate, and yet, it has a uniformly bounded competitive ratio even if the predictions are completely inaccurate.

Production systems with cycle overrun: modelling, analysis, improvability and bottlenecks

Production systems literature usually attributes throughput losses to two reasons: unreliable equipment and random part processing time (also referred to as machine cycle time). In practice, however, one more reason for throughput losses is observed: cycle overrun. The specificity of cycle overrun is that not all parts may require more time than allotted by a fixed cycle time and, if the overrun does occur, its duration is coupled with the part processing time: typically, it is a fraction or a small multiple of the cycle time. This paper is indented to develop methods for analysis and improvement of production systems with unreliable machines and cycle overrun. Specifically, it introduces a mathematical model of an unreliable machine with cycle overrun, develops its simplified version, explores the efficacy of machines' improvability by reducing either downtime or cycle overrun, and discusses the issue of bottleneck identification. These results are obtained under the exponential assumption on machine reliability and overruns and then extended to the non-exponential case. Finally, the methods developed are illustrated by a case study based on an automotive transmission case machining line.

Solving a hybrid batch production problem with unreliable equipment and quality reassurance

A hybrid batch fabrication plan involving an outsourcing option is often established to deal with the in-house capacity constraint and/or meet timely demand with a reduced cycle time. Besides, the occurrences of unpredictable equipment malfunction and imperfect product quality should also be effectively managed during in-house fabrication to meet the production schedule and the required quality level. To address these concerns, we examine a hybrid economic production quantity (EPQ) problem with an unreliable machine and quality reassurance; wherein, an outside provider helps supply a portion of the batch at a requested timing and desirable quality, but at the price of a higher than in-house unit cost. Corrective action is performed immediately when a Poisson-distributed breakdown occurs. Once the equipment repairing task completes, the interrupted lot’s fabrication resumes. Random nonconforming products are identified, and the re-workable items among them are separated from the scraps. A rework task follows the regular process in each cycle at an extra cost. A portion of reworked items fails and are scrapped. A model portraying the problem’s characteristics is built, and an optimization methodology is utilized to find the optimal runtime solution to the problem. A numerical example reveals our result’s applicability, and specific managerial implications are suggested.

Learning From Mislabeled Training Data Through Ambiguous Learning for In-Home Health Monitoring

Data are widely collected via the IoT for machine learning tasks in in-home health monitoring applications and mislabeled training data lead to unreliable machine learning models in in-home health monitoring. Researchers have proposed a wide arrangement of algorithms to deal with mislabeled training data, in which one straightforward and effective solution is to directly filter noise from training data so that the negative effects of mislabeled data can be minimized. In essence, noise filtering might be a suboptimal solution because the mislabeled data are not completely useless. The features and distributions of mislabeled data are still useful for learning, especially when training data are insufficient. In this work, we propose a novel framework to learn from mislabeled training data through ambiguous learning (LeMAL). LeMAL mainly consists of two parts. First, it converts the original training data to ambiguous data. Second, an ambiguous learning algorithm is applied to the ambiguous data. In this work, we propose a novel distance-based ambiguous learning algorithm so that the ambiguous data can be used in a better way. Finally, we demonstrate that LeMAL can effectively improve learning performance over existing noise filtering methods.

Optimal hybrid inventory replenishment runtime for a vendor- buyer coordinated system with breakdowns and rework/disposal of nonconforming items

Most multinational/transnational manufacturers that own internal supply chains and operate in turbulent world markets apply strategies related to product quality, low-cost, and timely delivery. This study aims to assist such firms with making accurate decisions to enable their competitive strategies and cope with the realities of limited capacity and unreliable equipment. We examine a vendor-buyer coordinated system featuring batch fabrication, outsourcing, quality reassurance, discontinuous deliveries, and an unreliable machine. The system outsources a portion of a lot to reduce manufacturing uptime, and the in-house fabrication system experiences undesirable defective items and Poisson distributed breakdowns. In each cycle, corrective action and rework/disposal of defective stocks are undertaken as these incidents occur, and upon receipt of outsourced products and when the entire batch is quality ensured, it makes multiple deliveries of the end products. Using modeling, formulation, derivation, and an optimization methodology, we obtain the problem’s cost function and justify its convexity. We then apply differential calculus and propose a recursive algorithm to derive the problem’s optimal replenishment runtime. A numerical illustration is offered to show the applicability of the result that reveals various important system characteristics/ capabilities, such as the distinct and combined influences of breakdowns, outsourcing, rework, scrap, and delivery-frequency factors on various system parameters, performance, and optimal runtime. The methods proposed here can facilitate managerial operations planning and strategic decision making in an intra-supply chain setting in practice.

A robust algorithm for explaining unreliable machine learning survival models using the Kolmogorov–Smirnov bounds

A new robust algorithm based on the explanation method SurvLIME called SurvLIME-KS is proposed for explaining machine learning survival models. The algorithm is developed to ensure robustness to cases of a small amount of training data or outliers of survival data. The first idea behind SurvLIME-KS is to apply the Cox proportional hazards model to approximate the black-box survival model at the local area around a test example due to the linear relationship of covariates in the model. The second idea is to incorporate the well-known Kolmogorov–Smirnov bounds for constructing sets of predicted cumulative hazard functions. As a result, the robust maximin strategy is used, which aims to minimize the average distance between cumulative hazard functions of the explained black-box model and of the approximating Cox model, and to maximize the distance over all cumulative hazard functions in the interval produced by the Kolmogorov–Smirnov bounds. The maximin optimization problem is reduced to the quadratic program. Various numerical experiments with synthetic and real datasets demonstrate the SurvLIME-KS efficiency.

Feature-Based Explanations Don't Help People Detect Misclassifications of Online Toxicity

We present an experimental assessment of the impact of feature attribution-style explanations on human performance in predicting the consensus toxicity of social media posts with advice from an unreliable machine learning model. By doing so we add to a small but growing body of literature inspecting the utility of interpretable machine learning in terms of human outcomes. We also evaluate interpretable machine learning for the first time in the important domain of online toxicity, where fully-automated methods have faced criticism as being inadequate as a measure of toxic behavior.We find that, contrary to expectations, explanations have no significant impact on accuracy or agreement with model predictions, through they do change the distribution of subject error somewhat while reducing the cognitive burden of the task for subjects. Our results contribute to the recognition of an intriguing expectation gap in the field of interpretable machine learning between the general excitement the field has engendered and the ambiguous results of recent experimental work, including this study.

Improved Hadoop Cluster Performance by Dynamic Load and Resource Aware Speculative Execution and Straggler Node Detection

The big data is one of the fastest growing technologies, which can to handle huge amounts of data from various sources, such as social media, web logs, banking and business sectors etc. In order to pace with the changes in the data patterns and to accommodate the requirements of big data analytics, the platform for storage and processing such as Hadoop, also requires great advancements. Hadoop, an open source project executes the big data processing job in map and reduce phases and follows master-slave architecture. A Hadoop MapReduce job can be delayed if one of its many tasks is being assigned to an unreliable or congested machine. To solve this straggler problem, a novel algorithm design of speculative execution schemes for parallel processing clusters, from an optimization perspective, under different loading conditions is proposed. For the lightly loaded case, a task cloning scheme, namely, the combined file task cloning algorithm, which is based on maximizing the overall system utility, a straggler detection algorithm is proposed based on a workload threshold. The detection and cloning of tasks assigned with the stragglers only will not be enough to enhance the performance unless cloning of tasks is allocated in a resource aware method. So, a method is proposed which identifies and optimizes the resource allocation by considering all possible aspects of cluster performance balancing. One main issue arises due to the pre configuration of distinct map and reduce slots based on the number of files in the input folder. This can cause severe under-utilization of slot as map slots might not be fully utilized with respect to the input splits. To solve this issue, an alternative technique of Hadoop Slot Allocation is introduced in this paper by keeping the efficient management of slots model. The combine file task cloning algorithm combines the files which are less than the size of a single data block and executes them in the highly performing data node. On implementing these efficient cloning and combining techniques on a heavily loaded cluster after detecting the straggler, machine is found to reduce the elapsed time of execution to an average of 40%. The detection algorithm improves the overall performance of the heavily loaded cluster by 20% of the total elapsed time in comparison with the native Hadoop algorithm.

Joint optimization of production and maintenance strategies considering a dynamic sampling strategy for a deteriorating system

This paper presents a control policy of integrated production, maintenance and quality control planning for a continuous production system subject to quality deterioration. The system under study is composed of an unreliable machine that produces one part type satisfying customers demand. The machine can fail at any time and is subject to quality deterioration, and so a preventive maintenance and quality control policies are proposed to decrease the rate of defectives and to increase the system availability. The proposed production policy incorporates production thresholds, which regulate the machine production rate. Traditional sampling inspections standards such as ANSI/ASQC Z1.4 and ISO 2859 have addressed a dynamic quality control level to face quality deterioration. However, these standards are based only on quality considerations, disregarding the economic aspects and the interactions with production and maintenance management in the design of sampling plans. Thus, the proposed integrated model analyses in detail the effect of such dynamic sampling strategy and relevant interactions with production and maintenance strategies. The main objective of the paper is to determine an appropriate production policy as well as the preventive maintenance and the quality control rates in order to minimize the expected average incurred cost and satisfy at the same time a quality constraint. Given the high flexibility and capacity to model complex manufacturing systems, a combination of simulation modeling and optimization techniques are used to determine a solution for this stochastic and constrained problem. In addition, numerical examples and an extensive sensitivity analysis are conducted to illustrate the proposed control approach. Furthermore, we compared the proposed integrated policy with three common policies from the literature. Such study serve us to highlight the effectiveness of the approach, since the proposed integrated policy led to considerable cost savings.

A vendor-buyer coordinated system featuring an unreliable machine, scrap, outsourcing, and multiple shipments

A vendor-buyer coordinated system featuring an unreliable machine, scrap, outsourcing, and multiple shipments

Optimal delivery due date for a supplier with an unreliable machine under outsourced maintenance

In this paper, we analyse the due-date-to-promise problem for a supplier who supplies parts to his customer. The parts are produced on an unreliable machine with constant production rate. The machine is leased from a contractor who is also entrusted with its maintenance. The supplier's decision is the optimal due date to promise to his customer, taking into account the holding and tardiness costs on the customer side, as well as the transfer payment to the contractor for the lease and maintenance of the equipment. The contractor in turn has to decide on the frequency of pre-ventive maintenance. Both maximize their own profits. For this finite horizon optimization problem, we employ a Non-Renewal-Theory based approach from the literature to derive the performance characteristics in the transient regime - first for a general setting involving general distributions and then, for a special case. We then use Game Theory to analyse the underlying two-member game for both the non-cooperative and cooperative cases. A numerical example drawn from the literature is used to illustrate the special case. We show that for the special case, the game has a unique Nash equilibrium. Another significant result is that the Nash solution dominates the Stackelberg solution when the supplier is the leader. The supplier is worse off when the contractor is the leader. The example further shows that the supplier and the contractor stand to gain under cooperation which is well documented in the literature. We point to relevant literature for strategies to enforce this cooperation.

Determining optimal uptime considering an unreliable machine, a maximum permitted backorder level, a multi-delivery plan, and disposal/rework of imperfect items

To remain competitive in unstable business environments, production managers of contemporary manufacturing firms must simultaneously achieve some important operating goals while planning batch fabrication. These goals may include ensuring high product quality, satisfying buyer demand with timely deliveries and acceptable stock availability, lowering cost, and keeping the batch fabrication process undisruptive. In production planning, permitting shortage backordering is an effective policy to reduce stock holding cost, given that stock availability meets an acceptable level of customer’s satisfaction. In addition, in an actual manufacturing environment, random equipment breakdowns and imperfect items are inevitable due to unpredictable issues in manufacturing processes. This study develops a decision support model to explore a fabrication uptime problem considering an unreliable machine, a maximum permitted backorder level, a multi-delivery plan, and disposal/rework of imperfect items to assist manages with the aforementioned operating goals. With the help of mathematical modeling, an optimization method, and a solution-seeking algorithm, the proposed precise model is capable of not only determining the optimal fabrication uptime policy, but also revealing the individual and joint impact(s) of various key system parameters on the problem, thus facilitating managerial decision-making.

Economic production quantity with the presence of imperfect quality and random machine breakdown and repair based on the artificial bee colony heuristic

This article considers a production-inventory system consisting of a single imperfect unreliable machine. The items manufactured by the system are either perfect items or imperfect items, which require a rework to be restored to perfect quality. The rework rate is permitted to be different from the production rate if the rework process is different from the main manufacturing process. The fraction of the number of imperfect items is random following a general distribution function. The time to failure of the machine is random, following a general distribution function. If the machine fails before the lot is completed, the production is interrupted and the machine repair is started immediately. A random machine repair time is assumed, with a general distribution function. Unlike a common assumption in the literature, after the repair of the machine is completed, the production resumes. During the machine repair, a shortage can occur. A single-variable expected average cost function is derived to find the optimal lot size. Because of the complexity in the model, the ABC heuristic is proposed and implemented to find a near optimal value for the lot size. The article also provides a sensitivity analysis of the model's key parameters. It has been observed that the lot interruption-resumption policy leads to smaller lot sizes. • An imperfect production system has a random time to failure and a random time to repair. • Shortages are permitted during the machine repair. • The model determines the optimal production lot size. • The lot interruption-resumption strategy leads to smaller lot sizes. • The lot size reacts significantly to changes in the machine failure rate and the repair rate.

Joint production and repair efficiency planning of a multiple deteriorating system

This paper presents an integrated model for the simultaneous production and repair activity planning of a manufacturing system whose performance output is subject to progressive deterioration. In this context, an appropriate joint control strategy is critical to reduce costs and remain competitive. The obtained control policy balances the amount of maintenance activities needed to increase the availability and reduce defects against the increase in the total cost from downtime and deterioration. The production system consists of an unreliable machine that produces one product type and where unmet demand is backlogged. The rate of defects of the machine depends on its level of deterioration, which is defined through a set of multiple operational states and the age of the machine. Additionally, an intensity control model is adapted to define the repair efficiency applied to the system, aiming to mitigate the effect of deterioration that is mainly observed on the failure intensity of the system. The solution is obtained numerically through the formulation of a Hamilton–Jacobi–Bellman equation. A numerical example is provided and an extensive sensitivity analysis is conducted to validate the obtained results.

Joint Production and Replacement Planning for an Unreliable Manufacturing System Subject to Random Demand and Quality

We consider a problem of optimal production and replacement control for a deteriorating manufacturing system containing a single unreliable machine that produces one type of products. The random phenomena are the breakdowns and repairs of the machine, the quality deteriorations and the demand of customers. The machine is subject to deteriorations and minimal repairs, which affect its failure rate and the defective rate of the parts produced, both increasing with the age of the machine. Due to the aging process, the option to replace the machine should be taken to eliminate these effects in order to be able to meet long-term demand. The objective is to find the simultaneous production and replacement control policies in order to minimize the total cost over an infinite planning horizon. The optimality conditions are developed in the form of the second-order Hamilton-Jacobi-Bellman (HJB) due to randomness in demand and quality. Numerical methods are used to obtain the optimal control policies. Finally, to illustrate the contribution of the paper and the robustness of the obtained control policies, a numerical example and a sensitivity analysis are presented.

Distribution of time to buffer overflow in a finite-buffer manufacturing model with unreliable machine

One of the most important characteristic of each queueing model is time to buffer overflow. In the paper the distribution of that time in a single-channel manufacturing system, modelled by a Markovian finite-buffer queue with machine breakdowns, is studied. By using the analytical approach based on the idea of embedded Markov chain, total probability law and linear algebra, the formula for Laplace transform of the time to buffer overflow conditional distribution is found. Numerical illustration is presented.