Includes abstract. / Includes bibliographical references (p. 127-129). / Faults and inefficiencies in air conditioning installations account for between 2% and 11% of allenergy consumed by commercial buildings in the United States each year. Diagnostics systems havebeen proven to improve the performance of air conditioning plants but the high costs of purchasing,retrofitting and maintaining such a system results in limited market adoption of such systems.This thesis discusses the design, implementation and results of low-cost remote monitoring anddiagnostic platform for use in air conditioning installations. The design of the various hardwarecomponents is presented along with the structure of the framework developed for each device. The thesis also contains information regarding the selection, integration and installation of the various types of sensors required on the various installations. A specially-designed protocol was also developed to handle communication between the hardware devices. Both the physical configuration and details of the protocol structure are presented in detail in this thesis. The mechanism through which the device uploads data to a server is also described in this thesis and includes details on both the hardware and the server technologies used in the upload process. The system has been installed on two different sites in Cape Town, South Africa and has produced meaningful diagnostic information since November 2007.

The optic disc is the region of the retina where the optic nerve exits the back of the eye. A number of conditions can cause the optic disc to swell. Papilledema, optic disc swelling caused by raised intracranial pressure (ICP), and nonarteritic anterior ischemic optic neuropathy (NAION), swelling caused by reduced blood flow to the back of the eye, are two such conditions. Rapid, accurate diagnosis of the cause of disc swelling is important, as with papilledema the underlying cause of raised ICP could potentially be life-threatening and may require immediate intervention. The current clinical standard for diagnosing and assessing papilledema is a subjective measure based on qualitative inferences drawn from fundus images. Even with the expert training required to properly perform the assessment, measurements and results can vary significantly between clinicians. As such, the need for a rapid, accurate diagnostic tool for optic disc swelling is clear. Shape analysis of the structures of the retina has emerged as a promising quantitative tool for distinguishing between causes of optic disc swelling. Optic disc swelling can cause the retinal surfaces to distort, taking on shapes that differ from their normal arrangement. Recent work has examined how changes in the shape of one of these surfaces, Bruch's membrane (BM), varies between different types of optic disc swelling, containing clinically-relevant information. The inner limiting membrane (ILM), the most anterior retinal surface and furthest from BM, can take on shapes that are distinct from the more posterior layers when the optic disc becomes swollen. These unique shape characteristics have yet to be explored for their potential clinical utility. This thesis develops new shape models of the ILM. The ultimate goal of this work is to develop noninvasive, automated diagnostic tools for clinical use. To that end, a necessary first step in establishing clinical relevance is demonstrating the utility of retinal shape information in a machine learning classifier. Retinal layer shape information and regional volume measurements acquired from spectral-domain optical coherence tomography scans from 78 patients (39 papilledema, 39 NAION) was used to train random forest classifiers to distinguish between cases of papilledema and NAION. On average, the classifiers were able to correctly distinguish between papilledema and NAION 85.7±2.0% of the time, confirming the usefulness of retinal layer shapes for determining the cause of optic disc swelling. The results of this experiment are encouraging for future studies that will include more patients and attempt to differentiate between additional causes of optic disc edema.

Detecting abnormal operating conditions, which will lead to faults developing later, has important economic implications for industries trying to meet their performance and production goals. It is unacceptable to wait for failures that have potential safety, environmental and financial consequences. Moving from a “reactive” strategy to a “proactive” strategy can improve critical equipment reliability and availability while constraining maintenance costs, reducing production deferrals, decreasing the need for spare parts. Once the fault initiates, predicting its progression and deterioration can enable timely interventions without risk to personnel safety or to equipment integrity. This work presents an online-integrated condition monitoring and prognostics framework that addresses the above issues holistically. The proposed framework aligns fully with ISO 17359:2011 and derives from the I-P and P-F curve. Depending upon the running state of machine with respect to its I-P and P-F curve an algorithm will do one of the following: (1) Predict the ideal behaviour and any departure from the normal operating envelope using a combination of Evolving Clustering Method (ECM), a normalised fuzzy weighted distance and tracking signal method. (2) Identify the cause of the departure through an automated diagnostics system using a modified version of ECM for classification. (3) Predict the short-term progression of fault using a modified version of the Dynamic Evolving Neuro-Fuzzy Inference System (DENFIS), called here MDENFIS and a tracking signal method. (4) Predict the long term progression of fault (Prognostics) using a combination of Autoregressive Integrated Moving Average (ARIMA)- Empirical Mode Decomposition (EMD) for predicting the future input values and MDENFIS for predicting the long term progression of fault (output). The proposed model was tested and compared against other models in the literature using benchmarks and field data. This work demonstrates four noticeable improvements over previous methods: (1) Enhanced testing prediction accuracy, (2) comparable processing time if not better, (3) the ability to detect sudden changes in the process and finally (4) the ability to identify and isolate the problem source with high accuracy.

Detecting abnormal operating conditions, which will lead to faults developing later, has important economic implications for industries trying to meet their performance and production goals. It is unacceptable to wait for failures that have potential safety, environmental and financial consequences. Moving from a “reactive” strategy to a “proactive” strategy can improve critical equipment reliability and availability while constraining maintenance costs, reducing production deferrals, decreasing the need for spare parts. Once the fault initiates, predicting its progression and deterioration can enable timely interventions without risk to personnel safety or to equipment integrity. This work presents an online-integrated condition monitoring and prognostics framework that addresses the above issues holistically. The proposed framework aligns fully with ISO 17359:2011 and derives from the I-P and P-F curve. Depending upon the running state of machine with respect to its I-P and P-F curve an algorithm will do one of the following: (1) Predict the ideal behaviour and any departure from the normal operating envelope using a combination of Evolving Clustering Method (ECM), a normalised fuzzy weighted distance and tracking signal method. (2) Identify the cause of the departure through an automated diagnostics system using a modified version of ECM for classification. (3) Predict the short-term progression of fault using a modified version of the Dynamic Evolving Neuro-Fuzzy Inference System (DENFIS), called here MDENFIS and a tracking signal method. (4) Predict the long term progression of fault (Prognostics) using a combination of Autoregressive Integrated Moving Average (ARIMA)- Empirical Mode Decomposition (EMD) for predicting the future input values and MDENFIS for predicting the long term progression of fault (output). The proposed model was tested and compared against other models in the literature using benchmarks and field data. This work demonstrates four noticeable improvements over previous methods: (1) Enhanced testing prediction accuracy, (2) comparable processing time if not better, (3) the ability to detect sudden changes in the process and finally (4) the ability to identify and isolate the problem source with high accuracy.