Condition Monitoring Program Research Articles

Explainable AI for Machine Fault Diagnosis: Understanding Features’ Contribution in Machine Learning Models for Industrial Condition Monitoring

Although the effectiveness of machine learning (ML) for machine diagnosis has been widely established, the interpretation of the diagnosis outcomes is still an open issue. Machine learning models behave as black boxes; therefore, the contribution given by each of the selected features to the diagnosis is not transparent to the user. This work is aimed at investigating the capabilities of the SHapley Additive exPlanation (SHAP) to identify the most important features for fault detection and classification in condition monitoring programs for rotating machinery. The authors analyse the case of medium-sized bearings of industrial interest. Namely, vibration data were collected for different health states from the test rig for industrial bearings available at the Mechanical Engineering Laboratory of Politecnico di Torino. The Support Vector Machine (SVM) and k-Nearest Neighbour (kNN) diagnosis models are explained by means of the SHAP. Accuracies higher than 98.5% are achieved for both the models using the SHAP as a criterion for feature selection. It is found that the skewness and the shape factor of the vibration signal have the greatest impact on the models’ outcomes.

Geosite condition monitoring in the UK 1999–2019

Geosite condition monitoring makes an assessment or value judgement about the current condition of a geosite compared to a defined desired baseline state. Such work has been undertaken in the UK for over 20 years. The original framework for monitoring nationally and internationally important geosites, first implemented in 1999, has been trialled independently in each of the four UK conservation agencies (Natural England, Northern Ireland Environment Agency, NatureScot, Natural Resources Wales and their predecessor bodies). Different adaptations and insights to geosite condition monitoring have been made depending on devolved circumstances, but the ability to report results on a UK basis, using some shared common standards, has been retained. This paper assesses the geosite monitoring programme in the UK, identifies the commonalities and differences in approach across the four UK countries and compares the results produced. As a long-running and widely tested national geosite condition monitoring programme, the lessons learned from the UK may be applicable to any geosite monitoring programme globally.

Nuclear Power Plant Instrumentation and Control Cable Prognostics Using Indenter Modulus Measurements

As the fleet of nuclear power plants (NPPs) approach their1 original qualified life (typically 40 years) and operators seek license extensions, regulators require assurance that they can continue to operate safely in the decades to come. Some of the most important, yet often overlooked components, are the cables that provide the signal paths for instrumentation and control (I&C) systems used to ensure safe and efficient operation of NPPs.In response to this, the authors explore the use of expanding indenter modulus (IM), an industry-accepted technique for cable condition monitoring, into a prognostic tool for predicting the remaining useful life (RUL) of I&C cables. Not only is this technique non-destructive, but it can be performed while NPP cables are in service, thus making it practical for adoption into existing cable condition monitoring programs. In this paper, the authors describe an accelerated aging cable test bed used to acquire several types of measurement parameters as cables age. Additionally, practical techniques are described in which simple IM measurements can be leveraged for condition monitoring and RUL estimation.Error analysis indicates the proposed method is superior to conventional RUL estimation techniques, such as simple trending and curve fitting. The authors demonstrate that using IM can potentially provide a non-destructive, in-situ estimation of RUL for I&C cables. As described in this paper, the IM data clearly shows trends as a function of cable age, and shows promising performance for RUL estimation especially compared with conventional techniques.

사물인터넷 기반 무선 상태 모니터링 시스템 개발

With the advancements in sensor technology and open source-based controllers, the development of a condition monitoring system has become relatively easier. However, there is not enough well-organized information to develop condition monitoring system. Therefore, we present a wireless condition monitoring system based on the Internet of Things (IoT) by using an open-source controller. The developed wireless condition monitoring system consists of a sensor module, data processing unit, and condition monitoring program. This system can measure 3-axis accelerations of frequency up to 1.5 kHz, with a frequency resolution of 0.05 Hz to 20 Hz, and temperature. The developed condition monitoring system provides statistical time information and frequency spectrum for advanced condition diagnosis. Furthermore, the condition diagnosis function for low-speed rotating machines was implemented effectively.

Analyzing the dynamics of the deadwood carbon pool in Spain through the European Level I Monitoring Programme

Forest deadwood is a key indicator for biodiversity, disturbances and for carbon storage assessments. However, the role of Mediterranean forests in deadwood carbon storage is unclear. In this regard, the European Regulation 2018/841 on the inclusion of greenhouse gas emissions and removals from land use, land use change and forestry in the 2030 climate and energy framework, requires reports by Member States on the average annual net emissions or removals of deadwood carbon. The main purpose of this work is to explore the trends of forest deadwood carbon to clarify the role of Spanish forest as deadwood carbon storage. We also investigate the effect of forest stand, climatic and physiographic variables on the deadwood carbon trends. For the purposes of our study, we used the deadwood data from two cycles of the European large-scale forest condition-monitoring programme Level I, which spanned from 2009 to 2017. Our findings indicate that there is statistical evidence of increments in deadwood carbon stocks in Spain over the last decade. We also identified certain spatial trends of mean annual changes in deadwood carbon: positive increment in the Mediterranean and Macaronesian regions, close to zero in the Atlantic region and negative in the Alpine region. We also found significant and positive relationships between the mean height and basal area on deadwood carbon stocks as well as a bell-shaped relationship between the mean diameter of living trees as a proxy of stand age on deadwood carbon stocks. The European Level I Monitoring Programme emerged as a useful tool to explore the dynamics of deadwood carbon. However, the temporal scheme and the lack of repeated measurements of dasometric inventories preclude forecasting future deadwood carbon stocks under alternative management regimes and climate conditions. In this regard, the deadwood data collected in future cycles, together with the deadwood data from the National Forest Inventory plots may allow the development of prediction models and therefore fulfill the international requirements on Member States for deadwood forecasts at national scale for the 2021 to 2030 period.

INTEGRATION AND OPTIMIZATION OF VARIOUS CONDITIONS MONITORING METHOD

Condition Monitoring technique acts as revolutionary mechanism because of effective, quick, reliable and cost effective application to uphold, maximize, optimize uninterrupted operation of high end equipment’s, in particular and all operating equipment’s, in general specially in power plant. Various methods are been adopted in this technique in which vibration analysis, wear debris analysis, thermography etc., are most common, widely recognized and in practice for carrying out condition monitoring program. These methods acts as preventive and proactive approach in identifying wear in machines and diagnose fault occurred in the machinery. Generally these condition monitoring techniques are isolated, carried out independently by respective departments, analyzed separately without establishing correlation between these techniques, leading to generation of separate reports for each technique. This leads to only a part of turbine faults typically get diagnosed. Therefore integration of these techniques can provide greater, deeper, and more reliable quick information leading to effective diagnosis of turbine in more accurate way. This paper deals with integrations and optimization of all various condition monitoring techniques that can be carried out simultaneously on turbine leading to early detection of problem and proper diagnosis to be carried out resulting in minimizing operating cost, avoid unplanned breakdown, and to maintain uninterrupted operation of turbine.

Evaluation of Current Signature in Bearing Defects by Envelope Analysis of the Vibration in Induction Motors

Motor current signature analysis (MCSA) enables non-invasive monitoring, without interruption of machine operation in a remote and online way, allowing the identification of various types of faults of electrical and mechanical nature without the need of accessing the motor itself, but only its supply cables. Despite its advantages, it has limitations in accurately diagnosing incipient roller bearing faults. For the detection of incipient roller bearing faults, envelope analysis of vibration signals is a well-known and stablished technique used by motor condition monitoring experts for a long time, overcoming MCSA for that purpose. Thus, it is proposed in this paper, that the fault characteristic frequencies of roller bearings are identified in the current spectrum with the aid of envelope analysis on the bearing vibration signal. After this aided identification, the fault related spectral components in the current spectrum can be correctly tracked over time for trending evaluation and decision-making. This approach can represent a significant economic value in a motor condition monitoring program, since vibration envelope analysis is performed only at a first step and, after that, its results can be applied for the MCSA monitoring of all same-model motor drivers in an industrial site. This approach is even more valuable considering the concept of the Self-Supplied Wireless Current Transducer (SSWCT) also proposed in this paper. The SSWCT is an Industrial Internet of Things (IIOT) device for MCSA application in an Industry 4.0 environment. This proposed device has wireless communication interface and wireless/battery less power supply, being supplied by the energy harvested from the magnetic field of the same currents it is transducing. So, it is a completely galvanic isolated monitoring device, without batteries and without any electric connections to the industry electric system, easily installable to the motor cables, not using precious space in the electric panels of the motor control centers and not having any physical contact to the monitored asset.

Predicting soil carbon saturation deficit and related properties of New Zealand soils using infrared spectroscopy

Conversion of soils supporting native vegetation to agricultural production has led to a loss of soil carbon stocks. Replacing a portion of the lost stocks will sequester atmospheric carbon with the concurrent benefit of enhancing soil sustainability. The ability of the fine fraction of soils (≤50-µm fraction) to adsorb organic carbon (OC) is considered a key mechanism capable of stabilising soil OC against loss. The difference between the current and maximum concentrations of OC in the soil fine fraction (FFC) has been termed the ‘saturation deficit’ (SatDef) and used to define the potential for a soil to sequester carbon. For New Zealand surface 0–15 cm soil layers, pedotransfer functions have been derived to quantify the soil carbon SatDef. The ability of combining infrared spectroscopy (IR) with partial least squares regression (PLSR) to derive predictive algorithms for soil properties included in these pedotransfer functions, the capacity of the soil fine fraction to stabilise carbon and the SatDef of the soil fine fraction were assessed in this study. A total of 168 air-dried and finely ground New Zealand surface soils representative of the major soil orders used for agricultural production were included. Principal components analysis of IR spectra showed a grouping by soil order that was related to mineralogy. Predictive IR/PLSR algorithms were derived for specific surface area, pyrophosphate-extractable aluminium, the FFC content, the 90th quantile regression of FFC and the SatDef of the fine fraction (R2 values ≥0.85; ratio of performance to interquartile range values ≥2.9). The results indicate that IR/PLSR provides a rapid and cost-effective mechanism for deriving information related to the amount of FFC in soils and the SatDef of the fine fraction. The IR/PLSR approach could be used to define the potential of soils to sequester carbon and identify the soil types to target for carbon sequestration technologies. The approach would also generate valuable data for soil carbon in national inventories or national soil condition monitoring programs.

Unbalance-induced rub between rotor and compliant-segmented stator

Today, turbomachinery designers are trying to attain zero clearances for optimizing performance and efficiency of machine by designing advanced compliant-segmented seals. This trend involves secondary issues such as rotor-to-stator contact interaction which is to be treated with more technical knowledge in rotordynamic design and condition monitoring programs. The aim of this paper is to gain insight into dynamic interaction between a flexible rotor and a set of compliant arc-shaped segments as stator system. Specifically, the unbalanced-induced forward rubbing response is investigated in tight clearance condition during resonance-passing situations. A Jeffcott rotor modeling is used for representing the flexible rotor. A set of linear elastically-mounted rigid bodies with arc-shaped inner surfaces, which are arranged circumferentially around the rotor, are assumed as the model of the stator. The effects of mass, rotary inertia, translational and tilting stiffnesses are taken into account in the segment modeling. To compute the contact dynamic response, the equations of motion are coupled via Lagrange multiplier technique. The constrained equation of motion is solved by an implicit predictor-corrector time-marching numerical algorithm. Investigations reveal that the steady-state coupled rubbing solutions exist for the rotor interacting with the segmented stator. The solutions are classified with respect to stator properties, e.g. the number of segments and the stator stiffness. The synchronous rubbing solution covers various feature including shift in resonance speed, lower amplitude of rotor's vibration, smooth rise and fall in case of soft segments, jump phenomena in case of stiff segments, chaotic orbiting motion under effects of segment mass, etc. as are discussed in detail.

The Role of Climatic Anomalies and Soil Moisture in the Decline of Drought-Prone Forests

Increased intensity and duration of droughts and high-temperature events have been associated with forest decline in many parts of the world, and these decline events are expected to become more common under climate change. There is, therefore, a need for monitoring and modeling of forest decline. We used a regional forest condition monitoring program (DEBOSCAT) to study the spatial distribution of decline events in 2012 in Catalonia (Northeastern Spain) and their relationship with climatic factors. In 2012, this dataset was collected after an extraordinarily dry summer, and allowed the study of decline events in eight dominant tree species. We fitted a logistic model to predict forest decline probability as a function of species, precipitation and temperature anomalies, solar radiation, and remotely sensed soil moisture data from the Soil Moisture and Ocean Salinity Mission (SMOS). Broadleaved species were more affected by decline events than conifers. The statistical model explained almost 40% of forest decline occurrence, wherein almost 50% of this variability was explained by species effect, with broadleaved trees being generally more sensitive to the studied factors than conifers. Climatically wetter areas and those more exposed to radiation were more likely to be affected, suggesting better adaptation of forests in dry areas. In general, more damaged forests were characterized by high-positive temperature anomalies, lower than average rainfall, and low soil moisture in summer 2012. The most vulnerable species was Fagus sylvatica, a Euro-Siberian species, contrasting with Pinus halepensis, a typically Mediterranean species, which showed low sensitivity to drought.

Incorporating Braking Distance Evaluation into Pavement Management System for Safe Road Operation

Maintaining a sufficiently short stopping distance is an important requirement for safe road operation. Stopping sight distances of road sections are decided during the highway alignment and roadway geometric design phase of the road development process. A pavement friction coefficient is used in the calculation of the stopping distances. Since pavement friction coefficient deteriorates with time under traffic action, and the available friction also reduces in wet weather, it is important for pavement maintenance engineers to ensure that sufficient skid resistance is maintained under actual operating conditions so that the stopping distances of road sections are less than the required safe stopping distance. This result requires pavement engineers to monitor regularly the available pavement skid resistance and the corresponding stopping distance of individual road sections. Because skid resistance is an accepted pavement condition to be measured in a pavement management system, it is proposed in this study that braking distance assessment be included as a part of the pavement condition monitoring program of a pavement management system. The proposed procedure consists of two parts: ( a) measurement of skid resistance and evaluation of braking distance and stopping distance and ( b) monitoring of the deterioration trends of individual road sections. The detailed procedure and the analytical tools necessary for the evaluation and prediction of braking and stopping distances, including a finite element skid resistance simulation model, are presented. A numerical example is presented to illustrate the concept and working of the proposed procedure.

Condition monitoring for the safe operation of offshore cranes

Offshore cranes work in harsher conditions than their onshore counterparts, with exposure to a highly corrosive environment and dynamic loading when conducting lifts from floating vessels. These degradation mechanisms can significantly reduce equipment life expectancy and increase maintenance/replacement costs without adequate mitigation programs in place. A systematic process must be followed to identify the potential for loss of integrity of components that could lead to failure, and strategies should be defined/implemented to ensure safe operation. Study outcomes feed into the maintenance and sparing philosophy, and cascade to site personnel as maintenance and operating procedures. The implementation of a robust condition-monitoring program prior to degradation occurring is a key factor in the success of this program, examples of which include the following. Slew bearing condition monitoring: in addition to regular visual inspection and recording of measurable readings, a well-structured grease sampling and analysis protocol provides early detection of moisture ingress, deterioration of lubrication and bearing failure. Structural integrity: reporting on integrity by exception does not provide confidence that an area may have been left out, or consistency in ongoing reporting. End of useful life definition: some components aren’t available for inspection; others can be inspected however they are susceptible to failures that are not practically detectable. These potential hidden failures can be prevented if a process is followed where the critical component’s operating life is defined, with appropriate supporting justification, and it is renewed accordingly regardless of visual appearance. An effective condition-monitoring program allows the operator to ascertain if the strategies in place are effective, pick up early detection of potential failure, and arrest the degradation while still inside a safe operating envelope.

Development of indentation techniques in support of cable condition monitoring programs

This paper summarizes work performed over the past few years at Atomic Energy of Canada Limited (AECL) to further develop indentation techniques in support of cable condition monitoring programs. Sensitivity to cable degradation of the commonly used specific compressive stiffness (or indenter modulus) is assessed based on laboratory testing of unaged, thermally aged, and irradiated cable samples. Another technique, which consists of measuring recovery time during the post-indentation phase, is also assessed for a variety of tested jacket and insulation materials. The document includes a review of the effect of indenter design and testing parameters on indentation results and provides input towards potential revisions of the IEC/IEEE 62582-2 Standard.

A myopic policy for optimal inspection scheduling for condition based maintenance

In this paper, we propose a decision policy for condition-based maintenance that schedules inspections according to the current health of the system, optimized myopically over the next inspection interval. In traditional condition-based maintenance practices, regular inspections are considered a given requirement. Most measurements, however, confirm that the machine in question is in good working order and that no maintenance is necessary. We consider the possibility that there are savings to be had in the number of inspections that are undertaken. As such, we use a proportional hazards model for risk of failure and a Markovian process to model the system covariates. The cost and time of inspections are incorporated into our model, and an optimal decision for one interval is made. This process can be repeated for each decision point, resulting in a decision policy that produces an inspect-or-replace result and an optimal time for the next inspection. Applied over the life of an asset, this policy can be used as a complete condition monitoring program supporting evidence based asset management.

Oil Analysis Cost-Effective Compressor Condition Computer Monitoring Technique

This paper describes the development of the use of lubricant analytical programmes and trend analysis to optimise oil change intervals and to predict compressor failure. The various analytical methods are covered, as are the most frequently occurring lubricant applications where such condition monitoring programmes are most appropriate.

Fault Diagnosis of Rolling Bearing Based on the PSO-SVM of the Mixed-Feature

The rolling bearing is one of the most important and widely used parts in the rotating machinery. It is necessary to establish a reliable condition monitoring program which can avoid serious fault in the runtime and diagnose failure timely and accurately when it happens. This paper puts forward to a fault diagnosis method of rolling bearing based on the PSO-SVM of the mixed-feature. Firstly, we extract features in time domain, frequency domain, and order quenfrency domain. Secondly, select both Support Vector Machine (SVM) parameters by Particle Swarm Optimization (PSO) algorithm and kernel function of SVM classification model. Finally, classification model of SVM is designed by using the extracted salient features, kernel function and optimal parameter of PSO. The result verifies the effectiveness of the proposed method.

A Quantitative Model of Wear Loss Based on On-Line Visual Ferrograph

On-line visual ferrograph (OLVF) is an efficient and real-time condition monitoring device. From the point of flow conservation, on the basis of the particle coverage area data collected by OLVF, this paper deduced two models about wear loss of the tribo-pairs in the wear process, one is general mathematical (GM) model including distribution impact factor of wear particle, and other simplified GM (SGM) model which does not contain the factor. The key factor affecting the accuracy of the two models is the three dimensional information of wear particles referring to particle area and thickness. This model using the disc and the ball whose materials were GCr15 were experimentally demonstrated on a pin-on-disc testing machine. And the OLVF was used to acquire the coverage area of the wear particles, which can reflect the wear loss. It shows that, in some cases, the approximate wear loss in the process was obtained on-line conveniently. Compared with experiment values derived from other wear measurement methods like weighing mass method and surface profilometry method, the SGM model can reflect tendency of wear loss about the tribo-pairs continuously. The deviations about wear loss by the model were discussed. Meanwhile, compared with the traditional means to compute the wear loss, this SGM model could be employed both for off-line analysis and on-line condition monitoring programs.

Addressing statistical and operational challenges in designing large-scale stream condition surveys

Implementing a statistically valid and practical monitoring design for large-scale stream condition monitoring and assessment programs can be difficult due to factors including the likely existence of a diversity of ecosystem types such as ephemeral streams over the sampling domain; limited resources to undertake detailed monitoring surveys and address knowledge gaps; and operational constraints on effective sampling at monitoring sites. In statistical speak, these issues translate to defining appropriate target populations and sampling units; designing appropriate spatial and temporal sample site selection methods; selection and use of appropriate indicators; and setting effect sizes with limited ecological and statistical information about the indicators of interest. We identify the statistical and operational challenges in designing large-scale stream condition surveys and discuss general approaches for addressing them. The ultimate aim in drawing attention to these challenges is to ensure operational practicality in carrying out future monitoring programs and that the resulting inferences about stream condition are statistically valid and relevant.

Optimizing a Wind Turbine Oil Condition Monitoring Program

Abstract The logistics and expense of obtaining oil samples from wind farms are among the conditions that make it challenging to establish a valuable condition monitoring program. Optimizing the oil condition monitoring program for a wind turbine is a critical step towards achieving forecasted ROI for the wind farm. To advance the current oil condition monitoring practices for most wind turbines, 150 identical wind turbine systems in the field for up to four years were studied. Several oil analysis tests were performed to understand the depletion of additives and changes in fluid chemistry. Basic statistical analysis was obtained on the data set to identify correlations in the data. The result of the analysis was a better understanding of the wind turbine oil’s formulation and the relationship to fluid degradation and gearbox condition. The statistical methodology that we performed can be applied to other wind turbine farms to improve the knowledge derived from their oil analysis tests, both for gear as well hydraulic oils. In addition, other innovative tests were suggested which may augment a current oil analysis test slate to improve the value of their program.

Guidelines for Alarm Limits and Trend Analysis

Abstract In 1999, CS96, In-Service Lubricant Testing and Condition Monitoring Services Industry Support, was formed to address the needs of monitoring in-service oils. The standards developed by this subcommittee provide equipment customers with a known basis for the quality of the data they are receiving [Improving Used Oil Analysis Standards: Recent Efforts of ASTM D02 Subcommittee CS96, Bryan Johnson, Practicing Oil Analysis, July–Aug 2006, pp 38–41]. However, test measurements have little meaning in a condition-monitoring program if they cannot be associated with a failure mechanism of the oil or the machine. Maximum reliability of in-service machine components and fluids requires that the condition-monitoringprogram provide timely indications of machinery and oil performance and remaining usable life. In order to address these critical aspects of condition monitoring, D02.96.4, Guidelines, initiated two task forces to develop Standard Guidelines to address Alarm Limits and Trends. Reliable alarm limits for fluid and equipment characteristics are required in order to properly interpret raw lubricant test data. However, these level alarms only state how much damage has occurred. In order to generate meaningful diagnostic and prognostic information on equipment and fluid condition, the rate of change must be trended. The predictive or forecasting nature of condition monitoring is based on trending, which determines the remaining useful life of the component and fluid. This paper presents an overview of practical alarm limit calculations using statistical analysis of equipment and fluid condition data and trend analysis of condition data in the dynamic equipment-operating environment. Various trending techniques and formulas will be presented with their associated benefits and limitations. These limit and trend calculations may be applied to all techniques that provide numerical test results and for all types of equipment (diesel, pumps, gas turbines, industrial turbines, hydraulics, etc.) whether large fleets or individual industrial machines.