Fault detection for residential photovoltaic power systems is an often-ignored problem. This thesis introduces a novel method for detecting power losses due to faults in solar panel performance. Five years of data from a residential system in Dalarna, Sweden, was applied on a random forest regression to estimate power production. Estimated power was compared to true power to assess the performance of the power generating systems. By identifying trends in the difference and estimated power production, faults can be identified. The model is sufficiently competent to identify consistent energy losses of 10% or greater of the expected power output, while requiring only minimal modifications to existing power generating systems.

The increasing numbers and complexity of spacecraft is driving a growing need for automated fault detection, isolation, and recovery. Anomalies and failures are common occurrences during space flight operations, yet most spacecraft currently possess limited ability to detect them, diagnose their underlying cause, and enact an appropriate response. This leaves ground operators to interpret extensive telemetry and resolve faults manually, something that is impractical for large constellations of satellites and difficult to do in a timely fashion for missions in deep space. A traditional hurdle for achieving autonomy has been that effective fault detection, isolation, and recovery requires appreciating the wider context of telemetry information. Advances in machine learning are finally allowing computers to succeed at such tasks. This dissertation presents an architecture based on machine learning for detecting, diagnosing, and responding to faults in a spacecraft attitude determination and control system. Unlike previous approaches, the availability of faulty examples is not assumed. In the first level of the system, one-class support vector machines are trained from nominal data to flag anomalies in telemetry. Meanwhile, a spacecraft simulator is used to model the activation of anomaly flags under different fault conditions and train a long short-term memory neural network to convert time-dependent anomaly information into a diagnosis. Decision theory is then used to convert diagnoses into a recovery action. The overall technique is successfully validated on data from the LightSail 2 mission. <br>

The overall objective of this thesis is to develop methods for fault detection and diagnosis for ground source heat pumps that can be used by servicemen to assist them to accurately detect and diagnose faults during the operation of the heat pump. The aim of this thesis is focused to develop two fault detection and diagnosis methods, sensitivity ratio and data-driven using principle component analysis. For the sensitivity ratio method model, two semi-empirical models for heat pump unit were built to simulate fault free and faulty conditions in the heat pump. Both models have been cross-validated by fault free experimental data. The fault free model is used as a reference. Then, fault trend analysis is performed in order to select a pair of uniquely sensitive and insensitive parameters to calculate the sensitivity ratio for each fault. When a sensitivity ratio value for a certain fault drops below a predefined value, that fault is diagnosed and an alarm message with that fault appears. The simulated faults data is used to test the model and the model successfully detected and diagnosed the faults types that were tested for different operation conditions. In the second method, principle component analysis is used to drive linear correlations of the original variables and calculate the principle components to reduce the dimensionality of the system. Then simple clustering technique is used for operation conditions classification and fault detection and diagnosis process. Each fault is represented by four clusters connected with three lines where each cluster represents different fault intensity level. The fault detection is performed by measuring the shortest orthogonal distance between the test point and the lines connecting the faults’ clusters. Simulated fault free and faulty data are used to train the model. Then, a new set of simulated faults data is used to test the model and the model successfully detected and diagnosed all faults type and intensity level of the tested faults for different operation conditions. Both models used simple seven temperature measurements, two pressure measurements (from which the condensation and evaporation temperatures are calculated) and the electrical power, as an input to the fault detection and diagnosis model. This is to reduce the cost and make it more convenient to implement. Finally, for each models, a user friendly graphical user interface is built to facilitate the model operation by the serviceman.

Recent refinements in Industry 4.0 and Machine Learning demonstrate the positive effects of using deep learning models for intelligent maintenance. The primary benefit of Deep Learning (DL) is its capability to extract attributes and make fast, accurate, and automated predictions without supervision. However, DL requires high computational power, significant data preprocessing, and vast amounts of data to make accurate predictions for intelligent maintenance. Given the considerable obstacles, meta-learning has been developed as a novel way to overcome these challenges. As a learning technique, meta-learning aims to quickly acquire knowledge of new tasks using theminimal available data by learning through meta-knowledge. There has been less research in the area of using meta-learning for Predictive Maintenance (PdM) and we considered it necessary to conduct this review to understand the applicability of meta-learning’s capabilities and functions to PdM since the outcomes of this technique seem to be rather promising. The review started with the development of a methodology and four research questions: (1) What is the taxonomy of meta-learning for PdM?, (2) What are the current state-of-the-art methodologies? (3) Which datasets are available for meta-learning in PdM?, and (4) What are the open issues, challenges, and opportunities of meta-learning in PdM?. To answer the first and second questions, a new taxonomy was proposed and meta-learnings role in predictive maintenance was identified from selected 55 papers. To answer the third question, we determined which types of datasets and their characteristics exist for this domain. Finally, the challenges, open issues, and opportunities of meta-learning in predictive maintenance were examined to answer the final question. The results of the research questions provided suggestions for future research topics.

Under de senaste åren fick maskininlärning mer och mer popularitet i samhället. Den implementeras i stor utsträckning inom många datavetenskapliga områden, t.ex. igenkänning av tal, video, objekt, sentimentanalys osv. Dessutom genererar moderna datorsystem och program stora filer med loggdata under deras körning och användning. Dessa loggfiler innehåller vanligtvis enorma mängder data, vilket leder till svårigheter att bearbeta all data manuellt. Således är användning av maskininlärningstekniker vid analys av loggdata för detektering av anomalibeteende av stort intresse för att uppnå skalbar underhåll av systemen. Syftet med detta arbete var att undersöka tillgängliga framträdande metoder för att implementera maskininlärning för upptäckning av loggfel och utvärdera en av dessa metoder. Uppsatsen fokuserade på att utvärdera DeepLog artificiella neurala nätverk som innehåller Long short-term memory algoritm. Utvärderingen omfattade mätning av den exekveringstid som behövdes och vilken precision, återkallande, noggrannhet och F1-index uppnåddes med modellen för maskininlärningsfelsdetektering vid användning av två olika loggdatamängder, en från OpenStack och en annan från Hadoop Distributed File System. Resultaten visade att DeepLog presterade bättre när man använde OpenStack-datamängd genom att uppnå höga resultat för alla index, särskilt recallsindex på cirka 90% som minimerade falska negativa förutsägelser, vilket är viktigt vid loggfelsdetektering. När DeepLog användes med HDFS-datamängd förbättrades körningstiden något men noggrannheten och recall av modellen tappades. Framtida arbete inkluderar att försöka och testa modellen med andra loggdatamängder eller andra ML-modeller för upptäckning av loggfel. / During the last years machine learning was gaining more and more popularity in the society. It is widely implemented in many fields of computer science, e.g. recognition of speech, video, objects, sentiment analysis, etc. Additionally, modern computer systems and programs generate large files with log data through their execution. These log files contain usually immense amount of data, which is a struggle for processing it manually. Thus, using machine learning techniques in the analysis of log data for detection of anomaly behavior is of a high interest for achieving scalable maintaining of the systems. The purpose of this work was to look into available prominent approaches of implementing machine learning for log fault detection and evaluate one of them. The paper focused on evaluating DeepLog artificial neural network that incorporates Long short-term memory. The evaluation included measuring the execution time needed and what precision, recall, accuracy and F1-index were achieved by the machine learning fault detection model when using two different log datasets, one from OpenStack and another from Hadoop Distributed File System. The results showed that DeepLog performed better when using OpenStack dataset by achieving high results for all indexes, especially the recall index of around 90%, minimizing the false negative predictions, which is important in the log fault detection. When using DeepLog with HDFS dataset the execution time was slightly improved but the accuracy and recall of the model were dropped. Future works includes trying another log datasets or ML models for log fault detection.

Indiana University-Purdue University Indianapolis (IUPUI) / The power transmission network is stretched thin in Western North America. When generators or substations fault, the resultant cascading failures can diminish transmission capabilities across wide regions of the continent. This thesis examined several methods of determining one-shot controls based on frequency decline in electrical generators to reduce the effect of one or more phase faults and tripped generators. These methods included criteria based on indices calculated from frequency measured at the controller location. These indices included criteria based on local modes and the rate of change of frequency. This thesis primarily used particle swarm optimization (PSO) with inertia to determine a well-adapted set of parameters. The parameters included up to three thresholds for indices calculated from frequency. The researchers found that the best method for distinguishing between one or more phase faults used thresholds on two Fourier indices. Future lines of research regarding one-shot controls were considered. A method that distinguished nearby tripped generators from one or more phase faults and load change events was proposed. This method used a moving average, a negative threshold for control, and a positive threshold to reject control. The negative threshold for the moving average is met frequently during any large transient event. An additional index must be used to distinguish loss of generation events. This index is the maximum value of the moving average up to the present time and it is good for distinguishing loss of generation events from transient swings caused by other events. This thesis further demonstrated how well a combination of controls based on both rate of change of frequency and local modes reduces instability of the network as determined by both a reduction in RMSGA and control efficiency at any time after the events. This thesis found that using local modes is generally useful to diagnose and apply one-shot controls when instability is caused by one or more phase faults, while when disconnected generators or reduced loads cause instability in the system, the local modes did not distinguish between loss of generation capacity events and reduced load events. Instead, differentiating based on the rate of change of frequency and an initial upward deflection of frequency or an initial downward deflection of frequency did distinguish between these types of events.

No description available.

Fault detection and diagnosis is a critical component of operations management systems. The goal of FDD is to identify the occurrence and causes of abnormal events. While many approaches are available, data-driven approaches for FDD have proven to be robust and reliable. Exploiting these advantages, the present study applied functional principal component analysis (FPCA) to carry out feature extraction for fault detection in internal combustion engines. Furthermore, a feature subset that explained 95% of the variance of the original vibrational sensor signal was used in a multilayer perceptron to carry out prediction for fault diagnosis. Of the engine states studied in the present work, the ending diagnostic performance shows the proposed approach achieved an overall prediction accuracy of 99.72 %. These results are encouraging because they show the feasibility for applying FPCA for feature extraction which has not been discussed previously within the literature relating to fault detection and diagnosis.

Early failure detection has been an integral part of condition monitoring of critical systems, such as wind turbines and helicopter rotor drivetrains. An unexplored application of early failure detection is fatigue testing of driveline components. On many occasions, driveline components fail catastrophically, leaving no evidence of the root cause of failure and causing extensive damage to test equipment. This can be prevented by detecting failure in its early stages. Test specimen would be preserved, enabling correlation of test results with design predictions. In this thesis, a method for early failure detection of gear fatigue is proposed. The gears in questions are parts of driveline components undergoing fatigue tests. The proposed method includes generation of an autoregressive model from a healthy, time synchronously averaged vibration signal. The parameters of the generated model are then used to construct a filter, which predicts deviations from the healthy signal. The output of this filter is then processed to detect failure. Vibration data from four run to failure tests were analysed. While the proposed method detected failure in all four data sets, performance was better in tests carried out at high torque and low speed in comparison to tests carried out under low torque and high speeds. Finally, potential improvements in the proposed method to increase its effectiveness are proposed. / “Early Failure Detection” (tidig detektion av utmattningsbrott) har länge varit en viktig del av tillståndsövervakning av kritiska system, som till exempel vindkraftverk och drivsystem för rotorblad på helikoptrar. Ett mindre utforskat område av “Early Failure Detection” är utmattningstestning av komponenter för transmissionssystem. Ofta går komponenterna sönder på ett sådant sätt att grundorsaken till haveriet inte går att fastställa, och som riskerar att skada testriggarna. Detta kan förebyggas om haveriet kan upptäckas i ett tidigt skede innan komponenten gar sönder helt och hållet. Testobjeket kan då bevaras, vilket ger möjligheter att korrelera testresultatet till utmattningsberäkningar av konstruktionen.  I den här uppsatsen föreslås en metod för Early Failure Detection för drevsatser i växlar. Växlarna ingår i transmissionssystem som utmattningsprovas. Den föreslagna metoden innebär att en autoreggresiv modell skapas från en tids-synkron medelvärdesbildning på den uppmätta signalen för den oförstärda komponenten. Parametrarna från den modellen används sedan för att skapa ett filter som predikterar avvikelser mot den oförstörda komponenten. Slutligen behandlas utsignalen fran det filteret för att upptäcka utmattningsskador pa drevsatsen i växeln.  Vibrationsdata fran fyra utmattningsprov har analyserats. I samtliga prov har provet körts tills brott har konstaterats. Utmattningsskador kunde konstateras tidigt, innan brottet inträffade, i tre av de fyra fallen. Slutligen föreslås förslag på utveckling av den använda metoden for att förbättra predikteringarna.

During the past years an increased focus on the development of autonomous solutions has resulted in driverless vehicles being used in numerous industries. Volvo Construction Equipment is currently developing the TA15, an autonomous hauler part of a larger transport solution. The transition to autonomous haulers have further increased the need for improved system condition monitoring in the strive for increased operational time. A method aiming to identify and isolate faults in the hydraulic steering system on the TA15 was therefore investigated in this thesis. Using fault tree analysis, five faults considered to be of importance regarding steering performance were selected. Two different methods for detecting the faults were compared to each other, data-driven and model based. Out of the two, data-driven was selected as the method of choice due to high modularity and relative simplicity regarding implementation. The data-driven approach consisted of Feed-Forward and Long Short Term Memory networks where the suitable inputs were decided to be a combination of pressure and position signals. Utilizing a simulation model of the steering system validated against the TA15, the selected faults were induced in the simulated system with various severity. Training the networks to classify and estimate fault severity in the simulated model resulted in satisfactory results using both networks. It was however concluded that in contrary to the Feed-Forward network, the LSTM network could achieve good performance using less amount of sensors. Although the diagnostic method showed promising result on a simulation model, test on the real TA15 needs to be performed in order to properly evaluate the method. The advantage of using a data-driven approach was specially noticeable when comparisons were made to the model based approach. The data-driven approach relies on labeling data rather than complete system knowledge. Meaning that the method developed therefore could be applied on practically any hydraulic system in construction equipment by changing the training data.