Dissertation (Ph.D.)--Auburn University, 2006. / Abstract. Vita. Includes bibliographic references (p.149-156).

Where the role of software-intensive systems has shifted from the traditional one of fulfilling isolated computational tasks, larger collaborative societies with interaction as primary resource, is gradually taking its place. This can be observed in anything from logistics to rescue operations and resource management, numerous services with key-roles in the modern infrastructure. In the light of this new collaborative order, it is imperative that the tools (compilers, debuggers, profilers) and methods (requirements, design, implementation, testing) that supported traditional software engineering values also adjust and extend towards those nurtured by the online instrumentation of software intensive systems. That is, to adjust and to help to avoid situations where limitations in technology and methodology would prevent us from ascertaining the well-being and security of systems that assists our very lives. Coupled with most perspectives on software development and maintenance is one well established member of, and complement to, the development process. Debugging; or the art of discovering, localising, and correcting undesirable behaviours in software-intensive systems, the need for which tend to far outlive development in itself. Debugging is currently performed based on a premise of the developer operating from a god-like perspective. A perspective that implies access and knowledge regarding source code, along with minute control over execution properties. However, the quality as well as accessibility of such information steadily decline with time as requirements, implementation, hardware components and their associated developers, all alike fall behind their continuously evolving surroundings. In this thesis, it is argued that the current practice of software debugging is insufficient, and as precursory action, introduce a technical platform suitable for experimenting with future methods regarding online debugging, maintenance and analysis. An initial implementation of this platform will then be used for experimenting with a simple method that is targeting online observation of software behaviour.

The ability to diagnose deviations and predict faults effectively is an important task in various industrial domains for minimizing costs and productivity loss and also conserving environmental resources. However, the majority of the efforts for diagnostics are still carried out by human experts in a time-consuming and expensive manner. Automated data-driven solutions are needed for continuous monitoring of complex systems over time. On the other hand, domain expertise plays a significant role in developing, evaluating, and improving diagnostics and monitoring functions. Therefore, automatically derived solutions must be able to interact with domain experts by taking advantage of available a priori knowledge and by incorporating their feedback into the learning process. This thesis and appended papers tackle the problem of generating a real-world self-monitoring system for continuous monitoring of machines and operations by developing algorithms that can learn data streams and their relations over time and detect anomalies using joint-human machine learning. Throughout this thesis, we have described a number of different approaches, each designed for the needs of a self-monitoring system, and have composed these methods into a coherent framework. More specifically, we presented a two-layer meta-framework, in which the first layer was concerned with learning appropriate data representations and detectinganomalies in an unsupervised fashion, and the second layer aimed at interactively exploiting available expert knowledge in a joint human-machine learning fashion. Furthermore, district heating has been the focus of this thesis as the application domain with the goal of automatically detecting faults and anomalies by comparing heat demands among different groups of customers. We applied and enriched different methods on this domain, which then contributed to the development and improvement of the meta-framework. The contributions that result from the studies included in this work can be summarized into four categories: (1) exploring different data representations that are suitable for the self-monitoring task based on data characteristics and domain knowledge, (2) discovering patterns and groups in data that describe normal behavior of the monitored system/systems, (3) implementing methods to successfully discriminate anomalies from the normal behavior, and (4) incorporating domain knowledge and expert feedback into self-monitoring.

District heating delivers more than 70% of the energy used for heating and domestichot water in Swedish buildings. To stay competitive, district heating needs toreduce its losses and increase capabilities to utilise low grade heat. Finding faultysubstations is one way to allow reductions in supply temperatures in district heatingnetworks, which in turn can help reduce the losses. In this work three suggestedsymptoms of faults: abnormal quantization, drifting and anomalous values, are investigatedwith the help of hourly meter data of: heat load, volume flow, supplyand return temperatures from district heating substations. To identify abnormalquantization, a method is proposed based on Shannon’s entropy, where lower entropysuggests higher risk of abnormal quantization. The majority of the substationsidentified as having abnormal quantization with the proposed method has a meterresolution lower than the majority of the substations in the investigated districtheating network. This lower resolution is likely responsible for identifying thesesubstation, suggesting the method is limited by the meter resolution of the availabledata. To improve result from the method higher resolution and sampling frequencyis likely needed.For identifying drift and anomalous values two methods are proposed, one for eachsymptom. Both methods utilize a software for predicting hourly heat load, volumeflow, supply and return temperatures in individual district heating substations.The method suggested for identifying drift uses the mean value of each predictedand measured quantity during the investigated period. The mean of the prediction iscompared to the mean of the measured values and a large difference would suggestrisk of drift. However this method has not been evaluated due to difficulties infinding a suitable validation method.The proposed method for detecting anomalous values is based on finding anomalousresiduals when comparing the prediction from the prediction software to themeasured values. To find the anomalous residuals the method uses an anomalydetection algorithm called IsolationForest. The method produces rankable lists inwhich substations with risk of anomalies are ranked higher in the lists. Four differentlists where evaluated by an experts. For the two best preforming lists approximatelyhalf of the top 15 substations where classified to contain anomalies by the expertgroup. The proposed method for detecting anomalous values shows promising resultespecially considering how easily the method could be added to a district heatingnetwork. Future work will focus on reducing the number of false positives. Suggestionsfor lowering the false positive rate include, alternations or checks on theprediction models used.

Dimensionality reduction techniques such as t-SNE and UMAP are useful both for overview of high-dimensional datasets and as part of a machine learning pipeline. These techniques create a non-parametric model of the manifold by fitting a density kernel about each data point using the distances to its k-nearest neighbors. In dense regions, this approach works well, but in sparse regions, it tends to draw unrelated points into the nearest cluster. Our work focuses on a homotopy method which imposes graph-based regularization over the manifold parameters to update the embedding. As the homotopy parameter increases, so does the cost of modeling different scales between adjacent neighborhoods. This gradually imposes a more uniform scale over the manifold, resulting in a more faithful embedding which preserves structure in dense areas while pushing sparse anomalous points outward.

abstract: Imagery data has become important for civil infrastructure operation and maintenance because imagery data can capture detailed visual information with high frequencies. Computer vision can be useful for acquiring spatiotemporal details to support the timely maintenance of critical civil infrastructures that serve society. Some examples include: irrigation canals need to maintain the leaking sections to avoid water loss; project engineers need to identify the deviating parts of the workflow to have the project finished on time and within budget; detecting abnormal behaviors of air traffic controllers is necessary to reduce operational errors and avoid air traffic accidents. Identifying the outliers of the civil infrastructure can help engineers focus on targeted areas. However, large amounts of imagery data bring the difficulty of information overloading. Anomaly detection combined with contextual knowledge could help address such information overloading to support the operation and maintenance of civil infrastructures. Some challenges make such identification of anomalies difficult. The first challenge is that diverse large civil infrastructures span among various geospatial environments so that previous algorithms cannot handle anomaly detection of civil infrastructures in different environments. The second challenge is that the crowded and rapidly changing workspaces can cause difficulties for the reliable detection of deviating parts of the workflow. The third challenge is that limited studies examined how to detect abnormal behaviors for diverse people in a real-time and non-intrusive manner. Using video andii relevant data sources (e.g., biometric and communication data) could be promising but still need a baseline of normal behaviors for outlier detection. This dissertation presents an anomaly detection framework that uses contextual knowledge, contextual information, and contextual data for filtering visual information extracted by computer vision techniques (ADCV) to address the challenges described above. The framework categorizes the anomaly detection of civil infrastructures into two categories: with and without a baseline of normal events. The author uses three case studies to illustrate how the developed approaches can address ADCV challenges in different categories of anomaly detection. Detailed data collection and experiments validate the developed ADCV approaches. / Dissertation/Thesis / Doctoral Dissertation Civil, Environmental and Sustainable Engineering 2020

<p>In the first chapter, we illustrate a big picture of the printing systems and the concentration of this dissertation. </p><p><br></p><p>In the second chapter, we present a tone-dependent fast error diffusion algorithm for color images, in which the quantizer is based on a simulated linearized printer space and the filter weight function depends on the ratio of the luminance of the current pixel to the maximum luminance value. The pixels are processed according to a serpentine scan instead of the classic raster scan. We compare the results of our algorithm to those achieved using</p> <p>the fixed Floyd-Steinberg weights and processing the image according to a raster scan ordering. In the third chapter, we first design a defect generator to generate the synthetic abnormal</p> <p>printer sounds, and then develop or explore three features for sound-based anomaly detection. In the fourth chapter, we explore six classifiers as our anomaly detection models, and explore or develop six augmentation methods to see whether or not an augmented dataset can improve the model performance. In the fifth chapter, we illustrate the data arrangement and the evaluation methods. Finally, we show the evaluation results based on</p> <p>different inputs, different features, and different classifiers.</p> <p><br></p><p>In the last chapter, we summarize the contributions of this dissertation.</p>

The ability to look outward from your vehicle and assess dangerous peer behavior is typically a trivial task for humans, but not always. Distracted driving is an issue that has been seen on our roadways ever since cars have been invented, but even more so after the wide spread use of cell phones. This thesis introduces a new system for monitoring the surrounding vehicles with outside facing cameras that detect in real time if the vehicle being followed is engaging in distracted behavior. This system uses techniques from image processing, signal processing, and machine learning. It’s ability to pick out drivers with dangerous behavior is shown to be accurate with a hit count of 87.5%, and with few false positives. It aims to help make either the human driver or the machine driver more aware and assist with better decision making.

No description available.

Anomaly detection generally refers to algorithmic procedures aimed at identifying relatively rare events in data sets that differ substantially from the majority of the data set to which they belong. In the context of data series generated by sensors mounted on mobile devices for non-destructive inspection and monitoring, anomalies typically identify defects to be detected, therefore defining the main task of this class of devices. In this case, a useful way of operationally defining anomalies is to look at their information content with respect to the background data, which is typically noisy and therefore easily masking the relevant events if unfiltered. In this thesis, a Kullback-Leibler (KL) Divergence filter is proposed to detect signals with relatively high information content, namely anomalies, within data series. The data is generated by using the model of a broad class of proximity sensors that apply to devices commonly used in engineering practice. This includes, for example, sensory devices mounted on mobile robotic devices for the non-destructive inspection of hazardous or other environments that may not be accessible to humans for direct inspection. The raw sensory data generated by this class of sensors is often challenging to analyze due to the prevalence of noise over the signal content that reveals the presence of relevant features, as for example damage in gas pipelines. The proposed filter is built to detect the difference of information content between the data series collected by the sensor and a baseline data series, with the advantage of not requiring the design of a threshold. Moreover, differing from the traditional filters which need the prior knowledge or distribution assumptions about the data, this KL Divergence filter is model free and suitable for all kinds of raw sensory data. Of course, it is also compatible with classical signal distribution assumptions, such as Gaussian approximation, for instance. Also, the robustness and sensitivity of the KL Divergence filter are discussed under different scenarios with various signal to noise ratios of data generated by a simulator reproducing very realistic scenarios and based on models of real sensors provided by manufacturers or widely accepted in the literature.