Equipment Failure Prediction Research Articles

An overview of the application of machine learning in predictive maintenance

With the rise of industrial artificial intelligence (AI), smart sensing, and the Internet of Things (IoT), companies are learning how to use their data not only for analysing the past but also for predicting the future. Maintenance is a crucial area that can drive significant cost savings and production value around the world.
 Predictive maintenance (PdM) is a technique that collects, cleans, analyses, and utilises data from various manufacturing and sensing sources like machines usage, operating conditions, and equipment feedback. It applies advanced algorithms to the data, automatically compares the fed data and the information from previous cases to anticipate or predict equipment failure before it happens, thus helping optimise equipment utilisation and maintenance strategies, improve performance and productivity, and extend equipment life. Robust PdM tools enable organisations to leverage and maximise the value of their existing data to stay ahead of potential breakdowns or disruptions in services, and address them proactively instead of reacting to issues as they arise. Therefore, it has attracted more and more attention of specialists in recent years.
 This paper provides a comprehensive review of the recent advancements of machine learning (ML) techniques widely applied to PdM by classifying the research according to the ML algorithms, machinery and equipment used in data acquisition. Important contributions of the researchers are highlighted, leading to some guidelines and foundation for further studies. Currently, BIENDONG POC is running some pilot PdM projects for critical equipment in Hai Thach - Moc Tinh gas processing plant.

Predicting Equipment Failure in Manufacturing Plants: An AI-driven Maintenance Strategy

Failure of equipment can lead to costly repairs and downtime for manufacturing facilities. Traditional maintenance techniques can result in unexpected failures, which can cause revenue losses and production delays. Artificial Intelligence (AI)-driven methods use machine learning technologies to predict equipment malfunctions before they happen, which has gained widespread interest. Artificial intelligence-based maintenance strategies can help improve the reliability of equipment and reduce the cost of repairs in manufacturing facilities. This study explores the use of Kaggle's predictive maintenance dataset to predict equipment failure in a manufacturing plant. The research involves collecting and evaluating data, selecting machine learning models, and evaluating metrics. The findings of this study revealed the application of AI-driven maintenance techniques to predict equipment failure, highlighting their potential to improve the efficiency of the manufacturing process and reduce costs. This study serves as a valuable contribution to the field of predictive maintenance and provides relevant implications for the industry. The paper explores the application of AI-based maintenance strategies to predict the failure of equipment in manufacturing facilities. The study utilizes the Kaggle dataset, which contains machine sensor readings and process variables from a manufacturing facility that processes batches of a specific product. The research methodology includes data gathering, engineering, and model selection. It also compared the effectiveness of AI-driven and traditional maintenance approaches. The results indicated that the former led to better predicted equipment failure rates and lower costs. The study's findings provide valuable insight into the subject of predictive maintenance and the implications it has for the manufacturing sector. It also offers future research directions.

Predictive asset availability optimization for underground trucks and loaders in the mining industry

With an increased focus on operational optimization, cost rationalization and increased advances in sensors, data storage and machine learning, mining Original Equipment Manufacturers are in need of intelligent machines to create an unassailable competitive advantage. To address this, Sandvik Mining and Rock Technology (Sandvik) and IBM have developed a customized service for fleet management and predictive maintenance for Sandvik’s mining equipment for the improvement of Overall Equipment Efficiency. In this endeavour, this paper aims to discuss the application of analytics to reduce machine downtime by predicting equipment failure in advance thus improving asset utilization of the fleet used in the mining industry. It details the journey of the analysis using the IoT data, data exploration, statistical approximations employed, various machine learning algorithms considered, and final selection of the techniques based on the industry recommended criteria. Predictive models for component failures (engine, brakes and transmission) and predictive models for equipment time to failure were built for underground mining trucks and loaders. With increased availability of additional sensor data and the need to interpret the outcomes that are actionable, supervised machine learning algorithms like decision trees were considered. Our work highlights various challenges encountered, the workarounds and solutions used to overcome them. The resulting models (built with IBM’s predictive analytics capability) of this work are augmented with Sandvik’s analytical offering, OptiMine® Analytics. This paper also highlights as to how our work has made a significant impact in financial terms and the client testimonials received.

Effects of imperfect IoT-enabled diagnostics on maintenance services: A system design perspective

Although many firms have deployed Internet of Things (IoT)-based diagnostics to predict equipment failures and maintenance requirements, they receive imperfect demand signals due to inadequate data quality and the use of ineffective predictive tools. A strategic queuing model was developed in this study to investigate a maintenance service provider’s optimal capacity allocation and pricing decisions in the presence of imperfect IoT-based diagnostics. In addition, we considered the heterogeneous service rates resulting from the accelerating effect of IoT-based diagnostics. The results of the study revealed that, on the one hand, the optimal service rate always increases with the diagnostic quality; on the other hand, if customer error cost is higher than that of the service provider, the optimal price increases with the diagnostic quality; otherwise, it decreases with the diagnostic quality. Furthermore, we found that the error costs for the two major stakeholders—the service provider and the customer—may affect the equilibrium in different ways. Finally, in the presence of the accelerating effect, we found that although the accelerating effect can improve the average service rate, system congestion actually increases if the effect becomes increasingly obvious; however, the service provider’s profit improves in this case.

Set-theoretical approach to finding the range of possible values of the size of a group of objects in a situation of multiple indistinguishability

In situations where it is not possible to timely use new methods for assessing hazardous objects, traditional methods of risk assessment can be supplemented with new methods based on possible assessments of objective primary information. Evaluation of the destructive capacity and consequences of the use of high-risk systems becomes increasingly important as their structural and functional complexity grows.Despite the existence of the considered works on certain aspects, the known approaches to the consideration of the problem of indistinguishability of influencing factors are characterized by limited functionality. These capabilities do not fully provide the required level of reliability when making management decisions. In view of this, the problem of predicting equipment failure caused by an unfavourable coincidence of circumstances due to the indistinguishability of the initial data is one of the urgent tasks of risk management. Its solution is of significant theoretical and practical interest for many complex heterogeneous dynamical systems. The paper deals with a problem from the field of probabilistic modeling. The relevance of the performed research lies in the fact that the existing methods for predicting equipment failure in systems of potentially hazardous objects do not take into account the causes of accidents that occurred due to unfavourable circumstances.The aim of the work is to develop a method for the set-theoretic modeling of the occurrence of problems with equipment due to an unfavourable coincidence of circumstances, which provides a range of possible values for the size of a group with a given composite very dangerous sign. To achieve this goal, the following tasks are solved:– finding the range of possible values of the number of groups of objects modeled by the intersection of indistinguishable sets and finding the range of possible values of the number of groups of objects modeled by the union and the difference of indistinguishable sets;– finding possible values of the number of groups of objects modeled by an arbitrary composite property and generalization of the problem, features of the set-theoretic apparatus of its solution.In the course of the study, it was found that a set of objects with a compound very dangerous feature of an arbitrary type can be associated with a set representing the union of intersections (intersection of unions) of sets of objects with simple features and their complements. Moreover, the operations of intersection, union, and complement constitute a complete set of operations for the most common version of Boolean algebra of sets.

Predicting anomaly conditions of energy equipment using neural networks

In modern conditions for complex thermal power facilities, the issue of developing methods for predicting equipment failures is especially relevant. Methods based on the intellectualization of diagnostic systems and allowing to obtain predictive models based on the use of both current data received in real time from measuring equipment and retrospective information are considered promising. Intellectualization of the system in terms of the ability to learn allows to quickly adjust the parameters of forecasting models under changing conditions of equipment operation, to determine new deadlines for scheduled repairs and minimize equipment downtime. A limitation of the use of methods is the incompleteness of failure statistics, ie when equipment failures are rare or non-existent. Such diagnostics of energy equipment, especially thermal power facilities, contributes to a more environmentally friendly production.

Application research of grey prediction model in rocket launcher fault prediction

This paper puts forward the importance of artillery equipment failure prediction, introduces the basic principles of grey prediction modeling, and finds the traditional GM (grey model) (1,1)deficiency. Aiming at this deficiency, the GM(1,1) model is improved based on residual correction and adaptive learning. The fault prediction of a rocket launcher automatic leveling system is taken as an application example, and the grey prediction GM(1,1) is compared and analyzed. Basic model and improved prediction of GM(1,1) model. The analysis shows that the improved GM(1, 1) grey prediction method improves the prediction accuracy and achieves a faster convergence speed, which indicates that the model is feasible and effective, and can provide an effective technical basis for the rocket gun automatic leveling system fault prediction.

Микро-потоки работ: сочетание потоков работ и потоковой обработки данных для поддержки цифровых двойников технологических процессов

Recently, there has been an explosive growth in the development of the concept of Industry. One of the most important elements of this concept is the application of methods of mathematical modeling and intelligent data analysis to create models of production processes and final products based on the processing of signals coming from intelligent sensors. A set of such models, which represent a virtual representation of industrial processes, systems and equipment, is called twin. Digital twins use data from sensors installed on production lines or on the basis of final products to predict equipment failures, optimize product quality and reduce the environmental impact of production processes. The model sets, underlying digital twins, can be described as workflows consisting of a set of computational services, each of which is a model of a process step. To provide flexible support for cloud computing to perform digital twins, we offer the concept of Micro-Workflows, which combines the power of the Scientific Workflows concept, the flexibility of container technology and the sustainability of the Stream Processing approach in distributed computing systems.

Guest Editorial: Production 4.0 Brings Autonomous Management to the Well

Guest editorial Most oil and gas operators can reduce their work s mantra down to a simple quest for excellence: To safely produce more barrels for fewer dollars. For decades, this pursuit has driven technological advances, including fracturing in completions and significant efficiency gains in drilling. However, each of these disciplines has essentially plateaued and further advancement is expected to be incremental at best. Production, however, holds the last remaining frontier for true efficiency gains. Industry 4.0 is the innovation revolution across all industries. For our own, it now presents a massive opportunity to elevate oilfield production performance to new standards of efficiency, control, and recovery. Early results indicate that operators who harness the power of Production 4.0 capitalize on an easy-to-reach and plentiful resource their own data. Building Production 4.0 This concept leverages the next industrial superpower-big data to empower an organization with operations-level decision making with minimal human involvement. Building Production 4.0 requires four elements, including the Internet of Things (IoT), cloud computing, edge computing, and advanced data analytics. IoT comprises digitally connected physical devices that add systematic support and efficiencies through empirical data and remote controls. Cloud computing allows these devices to network through Internet-hosted servers that store and manage data while reducing infrastructure and complexity. Edge computing connects intelligent devices to current and historical data so that autonomous decisions can be made where they matter most at the wellsite. Advanced analytics brings the concepts of IoT, cloud computing, and edge computing together to create an interconnected, intelligent ecosystem that helps oilfield equipment to act as intelligent machines that learn and teach themselves. Armed with powerful advanced analytics, intelligent ecosystems can glean meaningful, actionable insight from a pristine data source. Personnel no longer have the burden of managing lower-level transactions. Analytics help identify trends and anomalies, which provides opportunities for enhanced efficiency and higher performance. For the oilfield industry, data analytics enables intelligent production machines to improve themselves on the fly. These self-governing systems also predict equipment failures for just-in-time maintenance and allow operators to manage their entire asset by exception. Software: The Backbone of Production 4.0 Technology A key building block is an IoT platform, which is evolved from a traditional SCADA system. These are data workhorses that bring all the data streams from every part of an oil and gas operation into a single and unified platform. Production-optimization software then uses this data to act as a field-wide intelligence system to optimize artificial-lift and surface-facility efficiency. Leveraging this data empowers operators to not only increase equipment run life, but also maximize production while making it as profitable and economically viable as possible. These systems continuously monitor asset performance, give operators remote control over component-level assets, and enable more informed production decisions.

Multi-agent system architectures for collaborative prognostics

This paper provides a methodology to assess the optimal multi-agent architecture for collaborative prognostics in modern fleets of assets. The use of multi-agent systems has been shown to improve the ability to predict equipment failures by enabling machines with communication and collaborative learning capabilities. Different architectures have been postulated for industrial multi-agent systems in general. A rigorous analysis of the implications of their implementation for collaborative prognostics is essential to guide industrial deployment. In this paper, we investigate the cost and reliability implications of using different multi-agent systems architectures for collaborative failure prediction and maintenance optimization in large fleets of industrial assets. Results show that purely distributed architectures are optimal for high-value assets, while hierarchical architectures optimize communication costs for low-value assets. This enables asset managers to design and implement multi-agent systems for predictive maintenance that significantly decrease the whole-life cost of their assets.

An IIoT-based architecture for decision support in the aeronautic industry

The Industry 4.0 movement is driving innovation in manufacturing through the application of digital technologies, leading to solid performance improvements. In this context, this paper introduces a real-time analytical framework based on predictive, simulation and optimization technologies applied to decision support in manufacturing systems, enabled by an underlying reference implementation of an open Industrial Internet of Things (IIoT) platform. This architecture integrates critical equipment, manufacturing and corporate systems through a Unified IIoT Cloud Platform. A real case study on the aeronautic industry demonstrates the proposal feasibility of this architecture to enhance productivity, predict equipment failures and bring agility to react to unexpected events. In this case study, the monitoring tool displays the current status of the critical resources and the predictive tool calculates a probability of failure. When this probability reaches a certain threshold, the simulation tool is triggered to evaluate the impact of the disruption in the system’s productivity. Results from the tools are displayed online through an alert system so that each stakeholder is informed timely and in a contextualized way.

The Techniques for Achieving High Power Equipment Reliability with Distributed Informational System

The problem of power generating equipment reliability and safety of thermal (TPP), hydro (HPP) and nuclear (NPP) power plants at each its life cycle stage is set and the solution approach is proposed. The blockchain distributed data storage system to unite different members of the energy system is described. It is shown that the suggested technology allows decentralized data storage reliability increase due to the data exists till its last participant leaves and safety is ensured by the electric market participant consensus algorithm. The algorithm for the decentralized blockchain system is developed. The new technique for reliability design calculation based on using both constant and time-varying failure rate is introduced. It is suggested to use control action method expressed as the design, technological and operational parameters from the normative documents. The generalized model of reliability design calculation representing product of three components of failure free operation probability is developed. It is shown that developing new algorithms of statistical data obtaining and designing full repair processes allow planning equipment repair, obtaining and analyzing the corresponding reliability indeces when the equipment is in operation and then choose the most fitting repair time and amount optimization, operation mode selection and power plant long term equipment time in operation forecasting solutions. The technique for the power equipment condition forecasting by archival data stored in the distributed system, the data can be used to predict equipment failures and decide whether it should be repaired. It is shown that the desired prediction accuracy can be achived by using neural network due to its feature to reveal complex relations between input and output values.

Using Fuzzy FMEA approach to improve decision-making process in CNC machine Electrical and control equipment failure prediction

Using Fuzzy FMEA approach to improve decision-making process in CNC machine Electrical and control equipment failure prediction

Real-time big data analytics for hard disk drive predictive maintenance

Effective and reliable cloud services depend on the quality of service provided by large-scale data centers, and data center equipment reliability issues can cause significant data and financial loss to cloud service clients. Corrective maintenance is a reactive approach that only corrects problems once they occur, resulting in unwanted downtime, while preventative maintenance relies on replacing equipment which may yet still have considerable effective operating lifetime, thus raising maintenance costs. In contrast, predictive maintenance can potentially predict equipment failure in advance, thus reducing unplanned downtime and extending equipment lifetimes, thus reducing maintenance costs while increasing system reliability. This research aims to develop a real-time predictive maintenance system, HDPass, based on Apache Spark for the detection of imminent hard disk drive (HDD) failures in data centers.

Using Machine Learning Algorithms on data residing in SAP ERP Application to predict equipment failures

Asset intensive Organizations have searched long for a framework model that would timely predict equipment failure. Timely prediction of equipment failure substantially reduces direct and indirect costs, unexpected equipment shut-downs, accidents, and unwarranted emission risk. In this paper, the author proposes a model that can predict equipment failure by using data from SAP Plant Maintenance module. To achieve that author has applied data extraction algorithm and numerous data manipulations to prepare a classification data model consisting of maintenance records parameters such as spare parts usage, time elapsed since last completed maintenance and the period to the next scheduled maintained and so on. By using unsupervised learning technique of clustering, the author observed a class to cluster evaluation of 80% accuracy. After that classifier model was trained using various machine language (ML) algorithms and subsequently tested on mutually exclusive data sets with an objective to predict equipment breakdown. The classifier model using ML algorithms such as Support Vector Machine (SVM) and Decision Tree (DT) returned an accuracy and true positive rate (TPR) of greater than 95% to predict equipment failure. The proposed model acts as an Advanced Intelligent Control system contributing to the Cyber-Physical Systems for asset intensive organizations.

An analysis of methods and approaches to evaluating reliability when predicting equipment failures

THE QUALITY OF EQUIPMENT plays an important role in ensuring its reliability. The production of poor-quality goods at manufacturing facilities can be avoided when the manufacturer observes the requirements of state standards, construction regulations and rules, technical specifications. etc. However, checking that the manufacturer is complying with these requirements is rather complicated. The authors of this article have analysed approaches to creating a mathematical tool capable of making a quantitative and qualitative prediction of equipment failures, evaluating current operational readiness, and determining the optimum degree of load to ensure principles of maximum reliability and conservability. Approaches to quantitative predicting equipment failures, developed and described using the research results, could be introduced into the current system of conformity assessment, given that there is a sufficient amount of statistical data regarding equipment functioning. These data have been collected by Transneft PJSC, over the last decade.

Predicting Equipment Failure Using IoT and AI

Predicting Equipment Failure Using IoT and AI

Operational Supply Chain Planning Method for Integrating Spare Parts Supply Chains and Intelligent Maintenance Systems

The lack of spare parts and inefficient maintenance leads to low equipment availability, thus damaging service level and potentially increasing production costs in the industrial sector. Spare parts demand is sporadic and erratic, which impairs its forecasting with available statistical tools. The use of Intelligent Maintenance Systems (IMS) has been studied to support the process of predicting equipment failures and thus contributing to the competitiveness of productive systems. In this context, this study proposes a method, which aims to integrate information provided by Intelligent Maintenance Systems into the operational planning of a spare parts supply chain. The obtained results indicate that the developed method is able to support spare parts supply chain cost reduction, while ensuring predetermined service levels.

A Study on Prediction Model of Equipment Failure Through Analysis of Big Data Based on RHadoop

With the development of the internet of things, which is widely applied not only to everyday objects but also to industrial areas, the production of big data is accelerating. To provide intelligent services without human intervention in the internet of things environment, intelligent communication between objects becomes the key, and since the failure of the mechanical equipment attached to the sensor causes malfunction of the object and product failure, big data analysis to predict equipment failure is becoming more important. The purpose of this study is to propose a model for predicting mechanical equipment failure from various sense data collected in the manufacturing process. This study constructed a RHadoop-based big data platform to distribute a large amount of datasets for research, and performed logistic regression modeling to predict the main variables causing the failure from various collected variables. As a result of the study, the main variables in the manufacturing process that cause equipment failure were derived from the collected sense data, and the fitness and performance evaluation for the prediction model were made using the ROC curve. As a result of the performance evaluation of the prediction model, the ROC curve showed a fairly high prediction accuracy with AUC close to 1. The results of this study are expected to be applicable to the prediction of malfunctions, product failure, or abnormal communication between objects due to miscellaneous product faults in our daily lives in the internet of things environment.

Sequential anomalies: a study in the Railway Industry

Concerned with predicting equipment failures, predictive maintenance has a high impact both at a technical and at a financial level. Most modern equipments have logging systems that allow us to collect a diversity of data regarding their operation and health. Using data mining models for anomaly and novelty detection enables us to explore those datasets, building predictive systems that can detect and issue an alert when a failure starts evolving, avoiding the unknown development up to breakdown. In the present case, we use a failure detection system to predict train door breakdowns before they happen using data from their logging system. We use sensor data from pneumatic valves that control the open and close cycles of a door. Still, the failure of a cycle does not necessarily indicates a breakdown. A cycle might fail due to user interaction. The goal of this study is to detect structural failures in the automatic train door system, not when there is a cycle failure, but when there are sequences of cycle failures. We study three methods for such structural failure detection: outlier detection, anomaly detection and novelty detection, using different windowing strategies. We propose a two-stage approach, where the output of a point-anomaly algorithm is post-processed by a low-pass filter to obtain a subsequence-anomaly detection. The main result of the two-level architecture is a strong impact in the false alarm rate.