•Automatic classification of WOs enhances and structures occupant-generated data.•Category prediction accuracy 57–86% for all problem categories and 80%+ for the top 3.•Prediction accuracy up to 90% ...was achieved for the most frequent subcategories.•Class prediction accuracies above 95% were consistently achieved for multiple classes.•When the problem category is known, subcategory prediction accuracy increases above 90%
Occupant-generated work orders are recognized as a good potential data to support Facility Management (FM) activities, however they are unstructured and rarely contain the specific information engineers require to resolve the reported issues. Instead, this often requires multiple trips are often needed to identify the required trade, identify the problem and required parts/tools, and resolve. A key challenge is data quality: free-form (unstructured) text is collected that frequently lacks necessary detail for problem diagnosis. Machine Learning provides new opportunities within the FM domain to improve the quality of information collected through online work order reporting systems by automatically classifying WOs and prompting building occupants with appropriate FM team-developed questions in real time to gather the required specific information in structured form. This paper presents the development, comparison, and application of two sets of supervised machine learning models to perform this classification for WOs generated from occupant complaints. A set of ∼150,000 historical WOs was used for model development and textual classification using with various term and itemset frequency approaches was tested. Classifier prediction accuracies ranged from 46.6% to 81.3% for classification by detailed subcategory; this increased to between 68% (simple term frequency) to 90% (random forest) when the dataset only included the ten most common (accounting for 70% of all WOs) subcategories. Hierarchical classification decreased performance. An FM-BIM integration approach is finally presented using the resultant classifiers to provide facilities management teams with spatio-temporal visualization of the work order categories across a series of buildings to help prioritize and streamline operations and maintenance task assignments.
The transition to Industry 4.0 has improved factories by improving the manufacturing process. With increasing automation, awareness of the role of humans in industrial maintenance management is also ...important for the realization of the Industrial Internet of Things (IIoT). Today’s smart factories use data from various sources for extraction of valuable insights to improve manufacturing processes and avoid failures. These improvements also add to the complexity of resolving different maintenance issues faced during the manufacturing process. There is a need to leverage untapped human knowledge in Maintenance Work Orders (MWOs) to handle these complex challenges using state-of-the-art Natural Language Processing (NLP) techniques. The development of Industry 4.0 technologies is leading to a growing interest in using digital twins in many sectors. Digital twin-based services are revolutionizing design, manufacturing, product use, and maintenance (diagnosis, prognosis, and decision-making). This paper proposes a human knowledge centered intelligent maintenance decision support. The proposed service can find solutions to new maintenance problems using knowledge in past maintenance records in a digital twin environment. The architecture of the proposed service and its connections with Physical Space (PS), Virtual Space (VS) and Digital Twin Data (DTD) are presented in this paper. The performance of the service is validated using a case study on an open-source dataset of real MWOs from mining excavators. Results indicate that state-of-the-art NLP techniques can be used to process human knowledge in MWOs and generates interesting patterns. This study is also a step forward towards application of Technical Language Processing (TLP) in a smart manufacturing setup.
•Human knowledge centered maintenance decision support.•Maintenance in manufacturing process cycle for digital twin based environment.•Service architecture to process human knowledge in Maintenance Work Orders (MWOs).•Role of state-of-the-art NLP techniques in maintenance decision support.
•Determine the period between measurements in non-continuous predictive maintenance.•The Time Interval Between Measurements (TIBeM) is calculated dynamically.•The TIBeM changes depending on the ...criticality, reliability and machine status.•Current machine status is based on automatic assessment of the measurements.•The TIBeM can be immediately implemented in the ERP system of the plant.
In practice, on a large number of machines in industrial plants, predictive maintenance relies on periodic measurements to diagnose the condition of the equipment, rather than continuous monitoring of vibrations. In those cases, choosing an appropriate period between measurements is the key to success. Setting a long period implies taking a serious risk of breakdown, while a very short time interval between measurements can unnecessarily increase the costs of the maintenance plan.
This work shows a methodology to determine and manage this Time Interval Between Measurements (TIBeM) dynamically adapted to each machine and situation. Depending on the criticality of each machine and its reliability, more specifically, its present diagnosed functional condition and the history of failures and measurements, the most appropriate TIBeM is recalculated each time a new measurement and diagnostic is performed.
The described method has been implemented and validated in a large process plant and has led to a considerable improvement in costs and the management of its predictive maintenance plan.
En este trabajo de investigación se llevó a cabo la implementación de la herramienta Google Site para gestionar las actividades y proyectos de mantenimiento en una empresa dedicada a la producción de ...elementos cárnicos localizada en el departamento del cesar en Colombia. Para lograr el correcto uso de la herramienta, fue necesario analizar la gestión de los activos actuales de la empresa, mediante la recopilación de información a través de órdenes de trabajo de mantenimiento, manuales, fichas técnicas, conocimiento de inspectores, operadores mecánicos, e información obtenida de internet. Considerando lo anterior, todos los procedimientos utilizados por la empresa fueron integrados en la herramienta online de Google Site basándose en la exactitud y confianza del trabajo realizado por el grupo de mantenimiento. Lo anterior con la finalidad de garantizar la visualización de la gestión de mantenimiento de formaremota y en cualquier lugar evaluado de forma correcta la trazabilidad de la información lo que contribuye al futuro de mejores prácticas en las actividades de mantenimiento, gestión de activos y documentos, lo que se ve reflejado en la mejora continua para la empresa.
Maintenance work orders are commonly used to document information about wind turbine operation and maintenance. This includes details about proactive and reactive wind turbine downtimes, such as ...preventative and corrective maintenance. However, the information contained in maintenance work orders is often unstructured and difficult to analyze, presenting challenges for decision-makers wishing to use it for optimizing operation and maintenance. To address this issue, this work compares three different approaches to calculating reliability key performance indicators from maintenance work orders. The first approach involves manual labeling of the maintenance work orders by domain experts, using the schema defined in an industrial guideline to assign the label accordingly. The second approach involves the development of a model that automatically labels the maintenance work orders using text classification methods. Through this method, we are able to achieve macro average and weighted average F1-scores of 0.75 and 0.85 respectively. The third technique uses an AI-assisted tagging tool to tag and structure the raw maintenance information, together with a novel rule-based approach for extracting relevant maintenance work orders for failure rate calculation. In our experiments, the AI-assisted tool leads to an 88% drop in tagging time in comparison to the other two approaches, while expert labeling and text classification are more accurate in KPI extraction. Overall, our findings make extracting maintenance information from maintenance work orders more efficient, enable the assessment of reliability key performance indicators, and therefore support the optimization of wind turbine operation and maintenance.
The classification of work orders for livelihood service play an important role in the e-government system, which goal is to push these work orders to the corresponding processing department. ...Although many studies have realized the classification of work orders automatically, these work orders are either related to many subjects or related to many departments. Therefore, the problem of work order classification is not a single label classification problem, but a multi-label classification problem. In view of the continuous development of deep neural networks, the goal of this paper is to modify a MultilabelMarginLoss function and design a MultilabelCrossEntropyLoss function to realize the multi-classification system of work orders, which can push the work orders to the related departments(one or more departments) automatically according to their appeal contents (consisting of short texts). We perform deep neural network for text classification including TextCNN, TextRCNN, TextRNN, TextRNN_Att, FastText and DPCNN on the real government hotline data set and compare the performances of these models. The results reveal that TextCNN and TextRCNN have a more stable performance on different loss functions, especially on the MultilabelCrossEntropyLoss function.
Reliability modelling requires accurate failure time of an asset. In real industrial cases, such data are often buried in different historical databases which were set up for purposes other than ...reliability modelling. In particular, two data sets are commonly available: work orders (WOs), which detail maintenance activities on the asset, and downtime data (DD), which details when the asset was taken offline. Each is incomplete from a failure perspective, where one wishes to know whether each downtime event was due to failure or scheduled activities.
In this paper, a text mining approach is proposed to extract accurate failure time data from WOs and DD. A keyword dictionary is constructed using WO text descriptions and classifiers are constructed and applied to attribute each of the DD events to one of two classes: failure or nonfailure. The proposed method thus identifies downtime events whose descriptions are consistent with urgent unplanned WOs. The applicability of the methodology is demonstrated on maintenance data sets from an Australian electricity and sugar processing companies. Analysis of the text of the identified failure events seems to confirm the accurate identification of failures in DD. The results are expected to be immediately useful in improving the estimation of failure times (and thus the reliability models) for real-world assets.
Unstructured technical texts are a rich resource of engineering knowledge underutilised for data analysis. Maintenance work orders (MWO), for example, capture valuable information to inform what work ...was done on an asset and why. Data in MWO short text fields is unstructured, terse and jargon-rich, complicating the ability of both humans and machines to read it. Our challenge is to efficiently extract technical information from the MWO short text field and combine it with data in structured fields such as dates, functional location, make and model of the asset. In this paper we present a technical language processing-based solution for this problem. Echidna is an intuitive query-enabling interface that visualises historic asset data in the form of a knowledge graph. This knowledge graph is produced by MWO2KG, which uses deep learning supported by annotated training data to automatically construct knowledge graphs from unstructured technical text combined with data from structured fields. The tools are tested on maintenance work order and delay accounting data provided by industry partners. These tools provide reliability engineers with an efficient way to find information in historic asset data for failure modes and effects analysis, maintenance strategy validation and process improvement work. Source code for both tools is available on GitHub under the Apache 2.0 License.
Purpose
The aim of this paper was to study current practices in FM work order processing to support and improve decision-making. Processing and prioritizing work orders constitute a critical part of ...facilities and maintenance management practices given the large amount of work orders submitted daily. User-driven approaches (UDAs) are currently more prevalent for processing and prioritizing work orders but have challenges including inconsistency and subjectivity. Data-driven approaches can provide an advantage over user-driven ones in work-order processing; however, specific data requirements need to be identified to collect and process the functional data needed while achieving more consistent and accurate results.
Design/methodology/approach
This paper presents the findings of an online survey conducted with facility management (FM) experts who are directly or indirectly involved in processing work orders in building maintenance.
Findings
The findings reflect the current practices of 71 survey participants on data requirements, criteria selection, rankings, with current shortcomings and challenges in prioritizing work orders. In addition, differences between criteria and their ranking within participants’ experience, facility types and facility sizes are investigated. The findings of the study provide a snapshot of the current practices in FM work order processing, which aids in developing a comprehensive framework to support data-driven decision-making and address the challenges with UDAs.
Originality/value
Although previous studies have explored the use of selected criteria for processing and prioritizing work orders, this paper investigated a comprehensive list of criteria used by various facilities for processing work orders. Furthermore, previous studies are focused on the processing and prioritization stage, whereas this paper explored the data collected following the completion of the maintenance tasks and the benefits it can provide for processing future work orders. In addition, previous studies have focused on one specific stage of work order processing, whereas this paper investigated the common data between different stages of work order processing for enhanced FM.
Sensors and mathematical models have been used since the 1990’s to assess the health of systems and diagnose anomalous behavior. The advent of the Internet of Things (IoT) increases the range of ...assets on which data can be collected cost effectively. Cloud-computing and the wider availability of data and models are democratizing the implementation of prognostic health (PHM) technologies. Together, these advancements and other Industry 4.0 developments are creating a paradigm shift in how maintenance work is planned and executed. In this new future, maintenance will be initiated once a potential failure has been detected (using PHM) and thus completed before a functional failure has occurred. Thus corrective work is required since corrective work is defined as “work done to restore the function of an asset after failure or when failure is imminent.” Many metrics for measuring the effectiveness of maintenance work management are grounded in a negative perspective of corrective work and do not clearly capture work arising from condition monitoring and predictive modeling investments. In this paper, we use case studies to demonstrate the need to rethink maintenance terminology. The outcomes of this work include 1) definitions to be used for consistent evaluation of work management performance in an Industry 4.0 future and 2) recommendations to improve detection of work related to PHM activities.
PDF
