Unscheduled Downtime Research Articles

Using Large Language Models to Recommend Repair Actions for Offshore Wind Maintenance

Abstract The Offshore Wind (OSW) industry is experiencing significant expansion, resulting in increased Operations & Maintenance (O&M) costs. Intelligent alarm systems offer the prospect of swift detection of component failures and process anomalies, enabling timely and precise interventions that could yield reductions in resource expenditure, as well as scheduled and unscheduled downtime. This paper introduces an innovative approach to tackle this challenge by capitalising on Large Language Models (LLMs). We present a specialised conversational agent that incorporates statistical techniques to calculate distances between sentences for the detection and filtering of hallucinations and unsafe output. This potentially enables improved interpretation of alarm sequences and the generation of safer repair action recommendations by the agent. Preliminary findings are presented with the approach applied to ChatGPT-4 generated test sentences. The limitation of using ChatGPT-4 and the potential for enhancement of this agent through re-training with specialised OSW datasets are discussed.

Enhancing Asset Performance: Maintenance Strategies for the GBC Pumping System at PT Freeport Indonesia

Focusing on Grasberg Block Caving (GBC) pumping system PT Freeport Indonesia, this study examines the factors of not achieving target and analyze the recommended solution. The research aims to identify the root causes of performance gaps and propose strategies to improve asset performance ultimately to achieve the company targets. The study employs a mixed method (quantitative approach as primary and qualitative as secondary) and analysis conducted with Pareto Chart identifies key factors contributing to performance gaps, Root Cause Analysis using Current Reality Tree to gather potential causes, as well as generating alternative solution and using Kepner-Tregoe Decision Analysis Evaluates and selects the recommended alternative solutions. The results are expected to provide actionable insights for the company. Implementing recommended alternative solution estimated saving 83% of the costs associated with component replacements and reduce overall unscheduled downtime 63.65%.

Predictive Maintenance using Machine Learning in Industrial IoT

The use of predictive maintenance Machine learning techniques aid systems or machines in lowering the occurrence of certain types of machine failures via prediction and the use of specific methods. An essential tactic for improving the efficiency and reliability of industrial equipment and optimizing maintenance operations is predictive maintenance (PdM). Machine learning-based predictive maintenance helps businesses reduce unscheduled downtime, maintenance expenses, and operational efficiency by identifying and fixing potential equipment issues in advance.

A machine learning approach for particle accelerator errant beam prediction using spatial phase deviation

Particle accelerators are extremely complex systems that are expected to operate on high availability. Predicting impending failures only by utilizing data collected from diagnostic equipment already on board can help operators to avoid installing expensive sensors, unscheduled downtime and associated costs. For this purpose we explore the predictive power of Machine Learning algorithms to detect faulty beams prior to the failure. In this study, we propose a Machine Learning approach to model mapping from a pair of sensors located across the accelerator. While the model is trained to represent normal operation, we evaluate the predictive performance on known faulty beam pulses. We also investigate the model performance on unseen data through k-fold cross-validation. Then we recap the analysis with a neural architecture search and hyperparameter optimization study to fine tune our initial model. This paper will also introduce a sustainable framework that can standardize Machine Learning workflow applied to particle accelerators.

Early Prediction of Remaining Useful Life for Rolling Bearings Based on Envelope Spectral Indicator and Bayesian Filter

On top of the condition-based maintenance (CBM) practice for rotating machinery, the robust estimation of remaining useful life (RUL) for rolling-element bearings (REB) is of particular interest. The failure of a single bearing often results in secondary defects in the connected structure and catastrophic system failures. The prediction of RUL facilitates proactive maintenance planning to ensure system reliability and minimize financial loss due to unscheduled downtime. In this paper, to acquire early and reliable estimations of useful life, the RUL prediction of REBs is formulated into nonlinear degradation state estimation tackled by the combination of the envelope spectral indicator (ESI) and extended Kalman filter (EKF). By fusing the spectral energy of the bearing fault characteristic frequencies (FCFs) in the averaged envelope spectrum, the ESI is crafted to remove the interference from rotor-dynamics and reveal the bearing deterioration process. Once the fault is identified, the recursive Bayesian method based on EKF is utilized for estimating the bearing end-of-life time via the exponential state-space model. The distinctive advantage of the proposed approach lies in its ability to make an early prediction of RUL using a small number of ESI observations, offering an efficient practice for predictive health management at the early stage of bearing fault. The performance of the proposed method is validated using publicly available experimental bearing vibration data across three different operating conditions.

A semi-supervised anomaly detection approach for detecting mechanical failures

Abstract Predictive maintenance attempts to prevent unscheduled downtime by scheduling maintenance before expected failures and/or breakdowns while maximally optimizing uptime. However, this is a non-trivial problem, which requires sufficient data analytics knowledge and labeled data, either to design supervised fault detection models or to evaluate the performance of unsupervised models. While today most companies collect data by adding sensors to their machinery, the majority of this data is unfortunately not labeled. Moreover, labeling requires expert knowledge and is very cumbersome. To solve this mismatch, we present an architecture that guides experts, only requiring them to label a very small subset of the data compared to today’s standard labeling campaigns that are used when designing predictive maintenance solutions. We use auto-encoders to highlight potential anomalies and clustering approaches to group these anomalies into (potential) failure types. The accompanied dashboard then presents the anomalies to domain experts for labeling. In this way, we enable domain experts to enrich routinely collected machine data with business intelligence via a user-friendly hybrid model, combining auto-encoder models with labeling steps and supervised models. Ultimately, the labeled failure data allows for creating better failure prediction models, which in turn enables more effective predictive maintenance. More specifically, our architecture gets rid of cumbersome labeling tasks, allowing companies to make maximum use of their data and expert knowledge to ultimately increase their profit. Using our methodology, we achieve a labeling gain of 90% at best compared to standard labeling tasks.

Application of statistical processes control for the performance improvement of a clinical engineering department

This article addresses the fundamental role of Statistical as a quality tool in the field of clinical engineering, to improve and optimize internal processes. This study describes the methodology used to apply the SPC in a reference hospital's clinical engineering department. Data was collected over an extensive period, involving multiple medical equipment and verification procedures. These data were analyzed using various statistical tools, such as control charts, Pareto charts, and descriptive statistics. The results showed stability in the department's processes, which made it possible to identify areas for potential improvement. Statistical analyses revealed behavior patterns and trends that were not previously apparent. Based on these conclusions, specific modifications were proposed in the department's processes to optimize efficiency, reduce costs, and improve service quality. The implementation of these modifications based on evidence suggests that they would positively impact the general performance of the clinical engineering department if applied Key indicators could improve significantly, reflecting increased medical equipment reliability and availability, decreased unscheduled downtime, and increased satisfaction for department staff and equipment users. In summary, this study highlights the importance of using SPC as a powerful improvement tool in clinical engineering. By adopting an approach based on data and scientific evidence, clinical engineering departments can achieve more efficient and effective management of their processes, contributing to higher-quality medical care and patient safety.

Влияние вида организации ремонта на безопасность труда и уровень травматизма

An important task of an industrial enterprise is to ensure the reliability of equipment at the operational stage. For this purpose, repair activities are carried out at the enterprise, which involve hazards and risks in the field of occupational safety and health protection. The hazards of emergency repairs are related to their unplanned, sudden, emergency and non-standardized operations. Often unforeseen situations arise, works are performed without prior detailed risk analysis and without elaborated procedures for their implementation. These circumstances are the hazards with which incident risks are associated. The least hazardous is the organization of preventive maintenance and technical condition repairs. And the lowest danger is characterized by the organization of predictive maintenance. Statistical data from the foreign sources confirm the existence of direct correlation between the volume of emergency repairs, backlog works and unscheduled downtime on the one hand and injury rate on the other. And vice versa, there is an inverse correlation of the level of injuries with the volume of preventive maintenance, the degree of compliance with the maintenance schedule. In this regard, it is advisable to use in the occupational safety and health management system the indicators known from practice as indicators of maintenance processes. They characterize production processes in terms of the share of emergency repairs, preventive maintenance, maintenance repairs in the total volume of work, as well as the degree of fulfillment of the maintenance plan-schedule, etc. On their basis, the criteria for risk assessment and performance indicators in the field of occupational safety and health can be established. Monitoring these indicators in the information system and measures to improve them will help to ensure the workplace safety by improving maintenance and repair processes, identifying and eliminating root causes hidden in these processes.

Stochastic Models and Processing Probabilistic Data for Solving the Problem of Improving the Electric Freight Transport Reliability

Improving the productivity and reliability of mining infrastructure is an important task contributing to the mining performance enhancement of any enterprise. Open-pit dump trucks that move rock masses from the mining site to unloading points are an important part of the infrastructure of coal mines, and they are the main transport unit used in the technological cycle during open-pit mining. The failure of any of the mining truck systems causes unscheduled downtime and leads to significant economic losses, which are associated with the need to immediately restore the working state and lost profits due to decreased site productivity and a disruption of the production cycle. Therefore, minimizing the number and duration of unscheduled repairs is a necessity. The most time-consuming operations are the replacement of the diesel engine, traction generator, and traction motors, which requires additional disassembly of the dump truck equipment; therefore, special reliability requirements are imposed on these units. In this article, a mathematical model intended for processing the statistical data was developed to determine the reliability indicators of the brush collector assembly and the residual life of brushes of electric motors, which, unlike existing models, allow the determination of the refined life of the brushes based on the limiting height of their wear. A method to predict the residual life of an electric brush of a DC electric motor is presented, containing a list of controlled reliability indicators that are part of the mathematical model. Using the proposed mathematical model, the reliability of the brush-collector assembly, the minimum height of the brush during operation, and the average rate of its wear were studied and calculated.

Relating bulk solid caking events to phase change events that happen during crystallization processes

Caking causes significant problems in the storage of bulk solids in process vessels. Caking results in large rathole formation, lumps in product, inability to empty bags, silos, hoppers, and poor flow in feeders. Caking problems reaches all levels of the production process from the large-scale silos to the small single use packages sold in stores.Caking is characterized as a gain of significant bulk strength during storage. It is responsible for many lost production hours and unscheduled downtimes in processes dealing with bulk solids – particularly powders. Caking is also mechanistic in its nature. In this paper we will deal with one of those causes. The goal of this work is to more clearly understand the relationship between the root cause of caking and the kinetics of the gain of strength observed when recrystallization results in powder caking. Detailed strength measurements seem to indicate that the caking of powders induces a gain in strength in accordance with two exponential growth terms. The first strength growth term is proportional to storage time to the 1st power. The second exponential growth term is proportional to storage time to 3rd power. It was found that this behavior was consistent with phase growth kinetics in accordance with the Johnson-Mehl-Avrami-Kolmogorov (JMAK) equation. This equation suggests that within a confined space where crystallization is occurring the fraction of space that transforms into a crystalline material has kinetics consistent with exponential functions that are proportional to time to the 1st power, 2nd power, and 3rd power depending on the dimension of the crystals that form. Rods or fibers correspond to kinetics proportional to time to the 1st power. Plates correspond to kinetics proportional to time to the 2nd power. The formation of spheres or 3D clusters correspond to kinetics proportional to time to the 3rd power. The bulk strength tends to have a one-to-one correspondence to exponential crystal growth to time to the 1st and 3rd powers where the 1st power exponential strength growth happens on a short time frame and the 3rd power exponential growth happens on a longer time frame. This suggests that crystal rods or fibers first form in the pendular volume between particles followed by 3D structure formation in the pendules during bulk solid caking events. Understanding the magnitude and kinetics of these events can give engineers the information to prevent or minimize this issue.

Anomaly Detection in a Smart Industrial Machinery Plant Using IoT and Machine Learning.

In an increasingly technology-driven world, the security of Internet-of-Things systems has become a top priority. This article presents a study on the implementation of security solutions in an innovative manufacturing plant using IoT and machine learning. The research was based on collecting historical data from telemetry sensors, IoT cameras, and control devices in a smart manufacturing plant. The data provided the basis for training machine learning models, which were used for real-time anomaly detection. After training the machine learning models, we achieved a 13% improvement in the anomaly detection rate and a 3% decrease in the false positive rate. These results significantly impacted plant efficiency and safety, with faster and more effective responses seen to unusual events. The results showed that there was a significant impact on the efficiency and safety of the smart manufacturing plant. Improved anomaly detection enabled faster and more effective responses to unusual events, decreasing critical incidents and improving overall security. Additionally, algorithm optimization and IoT infrastructure improved operational efficiency by reducing unscheduled downtime and increasing resource utilization. This study highlights the effectiveness of machine learning-based security solutions by comparing the results with those of previous research on IoT security and anomaly detection in industrial environments. The adaptability of these solutions makes them applicable in various industrial and commercial environments.

Production optimization in the automotive industry using check sheet to reduce unscheduled downtime

The study was developed in an automotive industry, due to the increase in unscheduled stops in the cylinder head production line and the lack of information and records. Therefore, the verification sheet was applied as a methodology, as it is a quality tool that aims to evaluate the production inefficiency index, in order to guarantee improvements in the indicators. With the help of indicators and in order to increase productivity and reduce costs and waste, the analysis of stops in the production process was divided into four sectors, namely maintenance activities, tooling problems, preparation and the production process itself. After this analysis, it was necessary to quantify the hours lost over the three months. Therefore, after the study, improvement was achieved in filling out the verification sheet, records and complete and correct information, trained employees in addition to a significant reduction in downtime and a significant increase in production.

Comprehensive Analysis of IoT with Artificial Intelligence to Predictive Maintenance Optimization for Indian Shipbuilding

The extensive review of the literature evaluation on predictive maintenance (PdM) in this work focuses on system designs, goals, and methodologies. In the business world, any equipment or system failures or unscheduled downtime would negatively affect or stop an organization's key operations, possibly incurring heavy fines and irreparable reputational damage. Traditional maintenance methods now in use are plagued by a variety of limitations and preconceptions, including expensive preventive maintenance costs, insufficient or incorrect mathematical deterioration procedures, and manual feature extraction. The PdM maintenance framework is suggested as a new method of maintenance framework to prevent any damage only after the analytical analysis shows specific malfunctions or breakdowns, which is in line with the growth of digital building and the advancement of the Internet of Things (IoT), and Artificial Intelligence (AI), and so on. We also present an overview of the three main types of fault diagnosis and prognosis methods used in PdM mechanisms: scientific, conventional Machine Learning (ML), and deep learning (DL). While offering a thorough assessment of DL-dependent techniques, we make a quick overview of the knowledge-based and conventional ML-dependent strategies used in various components or systems. Eventually, significant possibilities for further study are discussed.

Fault diagnosis of regenerative thermal oxidizer system via dynamic uncertain causality graph integrated with early anomaly detection

Regenerative thermal oxidizers (RTO) serve as pivotal equipment for the treatment of volatile organic compounds (VOCs). However, a diversity of faults of the RTO system may threaten personnel safety and lead to unscheduled downtime. In this work, we develop a traceability inference method based on dynamic uncertain causality graph (DUCG) to identify root fault cause of RTO at early anomaly alarm. A general expert knowledge base for the fault diagnosis of RTO system is constructed. And the anomaly alarm thresholds are modified according to historical data with n-sigma rules to improve the diagnostic accuracy. Results on different cases show advantages in both triggering early anomaly alarms and avoiding invalid alarms, which is crucial to improve the fault inference accuracy. By applying our method to 35 fault cases of a real RTO for pharmaceutical VOCs treatment, the fault detection rate reaches 94.29%, and the fault inference accuracy is 82.86%. The two scores are 3% and 21% higher than those with the conventional DUCG (with fixed alarms), respectively. This remarkable improvement underlines the significance of our proposed method for RTO maintenance, and it is expected to assure the system’s operational stability and reliability.

State-of-the-Art Techniques for Fault Diagnosis in Electrical Machines: Advancements and Future Directions

Electrical machines are prone to various faults and require constant monitoring to ensure safe and dependable functioning. A potential fault in electrical machinery results in unscheduled downtime, necessitating the prompt assessment of any abnormal circumstances in rotating electrical machines. This paper provides an in-depth analysis as well as the most recent trends in the application of condition monitoring and fault detection techniques in the disciplines of electrical machinery. It first investigates the evolution of traditional monitoring techniques, followed by signal-based techniques such as spectrum, vibration, and temperature analysis, and the most recent trends in its signal processing techniques for assessing faults. Then, it investigates and details the implementation and evolution of modern approaches that employ intelligence-based techniques such as neural networks and support vector machines. All these applicable and state-of-art techniques in condition monitoring and fault diagnosis aid in predictive maintenance and identification and have the highly reliable operation of a motor drive system. Furthermore, this paper focuses on the possible transformational impact of electrical machine condition monitoring by thoroughly analyzing each of the monitoring techniques, their corresponding pros and cons, their approaches, and their applicability. It offers strong and useful insights into proactive maintenance measures, improved operating efficiency, and specific recommendations for future applications in the field of diagnostics.

Intelligent control technology of engineering electrical automation for PID algorithm

Electrical device automation in smart industries assimilates machines, electronic circuits, and control systems for efficient operations. The automated controls provide human intervention and fewer operations through proportional-integral-derivative (PID) controllers. Considering these devices’ operational and control loop contributions, this article introduces an Override-Controlled Definitive Performance Scheme (OCDPS). This scheme focuses on confining machine operations within the allocated time intervals preventing loop failures. The control value for multiple electrical machines is estimated based on the operational load and time for preventing failures. The override cases use predictive learning that incorporates the previous operational logs. Considering the override prediction, the control value is adjusted independently for different devices for confining variation loops. The automation features are programmed as before and after loop failures to cease further operational overrides in this process. Predictive learning independently identifies the possibilities in override and machine failures for increasing efficacy. The proposed method is contrasted with previously established models including the ILC, ASLP, and TD3. This evaluation considers the parameters of uptime, errors, override time, productivity, and prediction accuracy. Loops in operations and typical running times are two examples of the variables. The learning process results are utilized to estimate efficiency by modifying the operating time and loop consistencies with the help of control values. To avoid unscheduled downtime, the discovered loop failures modify the control parameters of individual machine processes.

Fault propagation analysis of computer numerically controlled machine tools

Computer numerically controlled machine tools are the key equipment of the intelligent manufacturing industry, unscheduled shutdowns can substantially cause equipment damage and production loss. Fault propagation analysis, as a prominent basis of reducing unscheduled downtime, can locate the harmful parts and reveal the fault propagation path. To improve reliability, this paper proposes a novel fault propagation analysis method that can accurately and timely evaluate fault propagation coefficients of components in interior computer numerically controlled machine tools. To depict the latent cause–effect relationships of components, a quantitative causal diagram is built with the aid of design information and topology. An inverse PageRank algorithm is proposed to assess the fault propagation risks of components. Unlike the traditional PageRank algorithm, the idea of inverse PageRank is that a node is crucial if it directs to other important nodes. Subsequently, the fault propagation coefficients of components in the quantitative causal diagram are derived and the fault propagation path is identified. The proposed method is validated on the computer numerically controlled machine tool, and the results are basically consistent with the typical cascading events and conform to expert judgment.

Condition-Based Maintenance of Gensets in District Heating Using Unsupervised Normal Behavior Models Applied on SCADA Data

Increasing interest in natural gas-fired gensets is motivated by District Heating (DH) network applications, especially in urban areas. Even if they represent customary solutions, when used in DH, duty regimes are driven by network thermal energy demands resulting in discontinuous operation, which affects their remaining useful life. As such, the attention on effective condition-based maintenance has gained momentum. In this paper, a novel unsupervised anomaly detection framework is proposed for gensets in DH networks based on Supervisory Control And Data Acquisition (SCADA) data. The framework relies on multivariate Machine-Learning (ML) regression models trained with a Leave-One-Out Cross-Validation method. Model residuals generated during the testing phase are then post-processed with a sliding threshold approach based on a rolling average. This methodology is tested against nine major failures that occurred on the gas genset installed in the Aosta DH plant in Italy. The results show that the proposed framework successfully detects anomalies and anticipates SCADA alarms related to unscheduled downtime.

Tool Condition Monitoring with Convolutional Neural Network for Milling Tools and Turning Inserts

Tool wear is one of the cost drivers in the manufacturing industry because it directly affects the quality of the manufactured workpiece and production efficiency. Identifying the right time to replace the cutting tool is a challenge. If the tool is replaced too soon, the production time can be disrupted, causing unscheduled downtime. Conversely, if it is replaced too late, there will be an additional cost to replace raw materials damaged by broken tools. Therefore, researchers continue to develop tool condition monitoring (TCM) methods to analyze tool wear. A recent popular method is machine vision with convolutional neural networks (CNN). The present research aims to develop classification models that can categorize the image data of milling and turning inserts into GO (suitable for use) and NO GO (not suitable for use). Two approaches are selected for the modeling process, custom learning and transfer learning, with image data input from smartphones and microscope cameras. The experimental results show that the best model is the transfer learning approach using Inception-V3 architecture with a smartphone image. The model reaches 92.2% accuracy, hence demonstrating a relatively good performance in determining whether the tool is suitable for use or not.

Maintenance Scheduling Algorithm for Transformers in Tanzania Electrical Secondary Distribution Networks

The drive by the government of Tanzania to electrify every village has resulted into expansion of the electrical secondary distribution networks (ESDNs). Therefore, maintenance management is of the highest priority for the smooth operation of the ESDNs to reduce unscheduled downtime and unexpected mechanical failures. Studies show that condition-based predictive maintenance (CBPdM) method allows the utility company to monitor, analyze and process the information obtained from ESDNs transformers. Thus, this study adopts the CBPdM method to develop a maintenance scheduling algorithm that can predict the transformer state, forecast maintenance time based on transformer load profile and schedule its maintenance using a knowledge-based system (KBS). Applying the challenge driven education approach, the requirements for developing an algorithm were established through an extensive literature survey and engagement of the key stakeholders from the Tanzania utility company. Our study uses the Dissolved Gas Analysis tool to collect the transformer parameters used in algorithm design. The parameter analysis was performed using Statistical Package for Social Sciences software. Results show that the designed KBS algorithm minimizes human-related maintenance errors and lowers labour costs as the system makes all the maintenance decisions. Specifically, the proposed maintenance scheduling algorithm reduces downtime maintenance costs by 1.45 times relative to the classical inspection-based maintenance model while significantly saving the maintenance costs.
 Keywords: Electrical power network, Forecasted load consumption, Knowledge-Based System, Maintenance Scheduling, Predictive Maintenance, Secondary Distribution