Rozprawy doktorskie na temat „Performance monitoring”

Performance monitoring in communication networks, which isthe main topic of this thesis, is an essential part ofperformance and network management. The applications cover anumber of different areas such as daily operations andmaintenance, usage information for billing purposes, customerreports, traffic control, planning, and dimensioning.

The main purpose of the thesis is to develop a single methodfor measurement of the most significant network performanceparameters in IP networks. It is a continuation from a previouslicentiate thesis that focused on performance monitoring incell-based networks.

The development of a measurement method that combines activeand passive approaches using monitoring blocks is the maincontribution of this work. A traffic flow performance meterbased on these principles has been implemented as an extensionto the open source flow meter NeTraMet and tested. Theresolution and precision of the results are determined by thesize of the monitoring block, which is the method’s mainparameter. Relevant statistical issues regarding packet lossesand delays are analysed. Finally, the measurement method isdiscussed in the context of applications, as well as networkand service management systems in general.

With the ascendency of data and the rise of interest in analytics, organizations are becoming more interested in the use of data to make their business processes more intelligent and reactive. BI applications are a common way that organizations integrate analytics in their processes. However, it can be days, weeks or even months before a manual response is undertaken based on a human interpreting a report. Even when information technology supports automatic responses within an organization, it is often implemented in an ad hoc manner without following a systematic framework. In this thesis, we present a reactive performance monitoring (RPM) framework which aims at automating the link from the analytical (how well is the operational achieving the strategic) to the operational (the particular process steps implemented within an organization that determine its behavior) aspects of businesses to bypass the strategic (the high level and long term goals an organization is trying to achieve) as needed and reduce the latency between knowledge and action. Our RPM framework is composed of an architecture, a methodology, and a rule environment which permits the redaction of rules possessing relevant conditions and actions. In addition, we present an OLAP rule engine which is demonstrated to be effective in our framework where events are streamed in, reacted upon in real-time, and stored in an OLAP database. To develop and evaluate our framework, two case studies were undertaken. The first was done using IBM technologies implementing an application made to identify patients at high risk of cancer recurrence. The second was done using open source technologies. With this second implementation, we created an application that has the goal of informing women from at risk populations of the different stages of pregnancy on a weekly basis.

Sand production is a common complex problem in the oil and gas industry, and choke valves is typically suffering for this in form of erosive damage. The degree of erosive damage is decided by many different factors where the flow rate velocity and the sand rate are the most important ones. Much effort has been spent on ways of reducing the choke erosion to be able to maintain the oil and gas production at an optimal level with attention to increased profit, safety and availability. Use of Computational Fluid Dynamics (CFD) has been essential in this work by simulating flow through the choke valve for optimizing the choke design, choosing the optimal erosion resistant material, coming up with improved erosion-related models, and optimal operational procedures of the choke. Producing with Acceptable Sand Rate (ASR), which means allowing a certain degree of sand erosion in chokes, have proven to be a successful way of maintaining the oil and gas production at an optimal level. To satisfy ASR-production, demands are made on an optimal use of condition and performance monitoring equipment and tools. The use of the condition and performance monitoring tool INSIGHT (from ABB) has in general proven to be successful for satisfying the ASR-production on different Statoil fields, including Statfjord which is in this thesis the area of focus regarding the use of INSIGHT. Important condition monitoring data such as sand rate, flow rate and pressure necessary to say something about the choke erosion status in INSIGHT must be as good as possible, because the quality of the results are limited by quality of the input data. In this thesis, INSIGHT has been presented, discussed and tested to be able to come up with possible limitations and improvements with special attention to condition monitoring (input) data used in INSIGHT.

For many applications a common source of performance degradation is excessive processor stalling from high memory latencies or poor data placement. Performance degradations from program and memory hierarchy interactions are often difficult for programmers and compilers to correct due to a lack of run-time information or limited knowledge about the underlying problem. By leveraging the Pentium 4 processor's performance monitoring hardware, specific run-time information can be provided, allowing code modifications to reduce or even eliminate problematic code, resulting in reduced execution times.
Furthermore, many tools currently available to aid programmers are program counter centric. These tools point out which area of the code produce slowdowns, but they do not directly show where the problem data structures are. This is a common problem in programs that dynamically allocate memory. By creating a ?malloc-centric? tool, we can develop an interesting perspective of the memory behavior of the system, providing better insight into the sources of performance problems.

The aim of this work was to develop sensory devices and data acquisition system to facilitate investigations into the mechanics of the rowing system, comprising the rower(s), boat and oars. As such, the parameters to be measured were: boat and seat position, velocity and acceleration; oar force; foot force; oar angle and rower heart rate. An oar force sensor was designed that fitted into the cavity of a modified oarlock. This sensor design is cheap, yields sound results and its presence is almost not noticeable to the rower. A review of previously applied methods of oar force measurement, predating 1900, is included. Foot force is of interest to many different fields of research, thus there is a large amount of literature on the subject of foot force measurement. A comprehensive review of this literature is used to aid in the design of the required sensor. The combination of a non-simple dynamic loading (i.e. time varying spatially distributed normal and shear forces), with static foot position distinguishes the problem of measuring the force under the feet during rowing from most previously considered cases. A strain gauge-based force sensing plate was designed to measure both normal force distribution and unidirectional shear force under the feet. Sample results are presented from a study with international class New Zealand rowers on a rowing ergometer. The sensor, performs well under normal force loadings, but needs modification to measure shear accurately. Possible modifications are suggested. While only a single oar angle, known as the sweep angle, was required to be measured, a sensor combination capable of measuring the spatial orientation of the oar relative to the boat was conceived. A new method of relative orientation estimation, via approximation of the Rodrigues' vector, which allows relative weighting of sensory data, was derived. Unfortunately, calibration issues prevented the gathering of meaningful data in the time available. A full theoretical development, including a new calibration scheme, which should alleviate the encountered problems, is included. While the motion of the rower within the boat is an important consideration in the dynamics of the rowing system, few previous researchers have measured it. These previous methods are briefly described, before the sensor used in this study, the optical rotary encoder, is detailed. Differentiation of the encoder signal to obtain seat velocity and acceleration relative to the boat was achieved using a purpose designed simple Kalman filter. The kinematic parameters of the boat, i.e. position, velocity and acceleration were measured using a combination of accelerometer and submerged impeller. The information from these two sensors was combined using a variant of the Kalman filter used in the differentiation of the encoder signal. The combination of the seat and boat kinematics allows study of the motion of the system centre of mass. Supplying power to, and collecting data from the above sensory devices was a purpose built data acquisition system dubbed ORAC (On-the-water Rowing Acquisition Computer). ORAC was designed to transmit the collected information, in real-time, to a remote laptop computer via wireless LAN, but the system used proved to have insufficient range, and hence ORAC was used as a standalone computer.

Continuous Integration (CI) is standard practice in Agile software development for source code management, including handling changes between development branches and master branch, and providing feedback for stakeholders. In the Continuous Integration the desirable practice is for development branches the be merged in to master branch daily. Before development branches is to be merged to master branch, the automated test tests builds are run on them and feedback of the tests is provided to developer and possible other stakeholders. If the tests are run without errors, the development branch is to be merged to master branch. The problems may arise, when the automated test build durations have variance. This leads developers to experience volatile service level causing irregularity to development process. The cause for the build duration variance can be the variance in the available system resources in the CI server. In some cases, the resource variation can cause false failures on the test build results as for example when there is defined maximum test build duration time limiter or when individual build step operation produces timeouts. To be able to resolve the resource starvation on the CI server, there needs to be means for identifying the resource starvation events in relation to test build statistics. By being able to identify relation between build statistics and resource starvation events, root cause, as starving resource, can be identified. By identifying the starving resource or resources, the system resources can be updated by adding more specific resource or handling the starvation through reserving the specific resource for key processes. The objective for this research is to implement monitoring tool for monitoring the Jenkins server’s builds and system resources during the builds. The thesis was conducted in the case company’s CI project including large-scale development project as customer project. The problem with case company was volatile test build durations and very insufficient data about system resources in the starvation events. The implemented monitoring tool achieved for providing extensive system resource data with combining it with the test build statistics to provide information about system resource availability in relation to test build statistics. The results indicate that the monitoring tool can be used for identifying resource starvation events and possible root causes for the events. The limitation of the implemented monitoring tool is the dependency for the database server as Elasticsearch. Elasticsearch was chosen to be used as database to establish adaptive monitoring data presenting functionality via Kibana server. This restricts portability of the monitoring tool to Jenkins instances as the monitoring tool needs mentioned instances for full functionality.

Rotary blasthole drills operating in a western Canadian surface coal mine were instrumented with microprocessor-based monitoring equipment. During routine production drilling, the performance parameters of penetration rate, torque, rotary speed, pulldown and bailing air pressure were monitored at sampling intervals of 10 centimeters (4 inches). The acquired digital data were subsequently correlated with both geological and geomechanical rock properties permitting a detailed examination of machine-rock interaction. The results from a statistical analysis of the drill data identified unique ranges of performance parameters for the sandstone, siltstone, mudstone and coal units encountered at the mine. Correlation of these ranges with geophysical logs in the monitored boreholes, enabled a further calibration of drilling parameter variation to rock type. A relationship between drill performance parameter responses and rock compressive and shear strength was established. This correlation demonstrated the possibility of estimating rock strength properties based on drill performance data. Trend analysis techniques applied to the monitored drill data permitted a further understanding of the nature of tricone bit wear for the particular study environment. The applications of drill monitoring techniques are illustrated in terms of geological exploration, mine planning, tricone bit selection and wear evaluation, and drill automation and control.

Multimedia applications require certain guarantees from the underlying system concerning quality of various parameters, such as the throughput, delay, loss rate etc. We assume that such quality of service (QoS) parameters are normally agreed upon for the duration of a session, and should be maintained throughout that session. If a QoS violation occurs, possibly due to the temporary overload of one of the system components, either the user will notice this and may request a renegotiation of the quality and cost parameters, or the system will automatically do some internal reconfiguration in order to recover from the problem. In the latter case, the system needs to continuously monitor the relevant performance parameters. In this thesis, we develop means for performing such monitoring in the context of a News-on-Demand application. We have designed and implemented a distributed QoS monitoring facility that can be used by distributed multimedia applications for QoS monitoring and QoS violation detection. The system consists of a measurement component and an administrative component. The measurement component is responsible for continuously measuring the end-to-end QoS parameters of connections between the server and client application, while the administrative component is responsible for the overall administration of the monitoring system, which includes raising QoS violation notifications if the quality of service is violated.

A P&H 2800XP electric mining shovel working in a Western Canadian surface coal mine was instrumented with microprocessor-based monitoring equipment, and its performance monitored as it excavated a series of test blasts. Hoist and crowd motor armature voltages and currents, dipper trajectories, cycle times and load weights were recorded. A diggability index has been established based on the responses of the hoist motor. Values of the diggability index correlated well with digging conditions as observed during monitoring and with muckpile fragmentation size distributions as determined by a photographic survey. This correlation establishes the ability of an instrumented shovel to diagnose the efficiency of ground preparation practices (blasting) by identifying variations in muckpile diggability. The data on dipper trajectory has demonstrated that variations in digging practices do exist amongst an experienced group of shovel operators, and that variation in trajectory significantly influences values of the recorded motor performance parameters. An approach is described to account for variations in digging practice on assessments of diggability. An allied investigation into the ability of time studies to define diggability, revealed dig cycle times to be operator dependant, and unrelated to levels of digging effort.

The British National Health Service (NHS) is one of the largest public services in the world and consequentially in 2004 it produced 25% of the total public sector carbon emissions for England. To meet national carbon targets the NHS must reduce its emissions; 26% by 2020, 64% by 2030, 80% by 2050 and is therefore interested in the development of strategies for reducing carbon dioxide emissions from buildings. The NHS building stock consists of a range of building archetypes constructed over the past 100 years. The energy used for heating and cooling hospital premises is the source of 22% of all NHS carbon emissions. The individual buildings are distributed across hospital sites that often have centralised energy plants, which make it difficult to monitor energy consumption on an individual building level. This thesis develops a method for monitoring the energy consumption of individual hospital buildings. The method was implemented on three case study buildings at Bradford Royal infirmary (BRI); a 1920s Nightingale, a nucleus and a modern modular building. Lessons were gathered from these studies to advance the knowledge on monitoring in UK hospitals. One of the key findings was that empirical models based on measured data are useful for estimating individual buildings annual heating energy consumption. The results show that the mechanically ventilated nucleus building had the highest energy consumption (808.7kWh/m2), followed by the naturally ventilated Nightingale building (420.7kWh/m2) and then the mixed-mode modular building (289.0kWh/m2). The internal environment was optimal in the nucleus building, but the Nightingale and modular buildings underperformed, with the modular overheating in summer and both buildings failing to meet air quality recommendations. Taking energy consumption and summer thermal resilience into consideration the Nightingale building had the best performance, demonstrating the longevity of the traditional design. The work identified a number of useful hospital design features; well-insulated heavyweight building fabric, well-controlled space heating, use of heat recovery ventilation and installation of localised monitoring equipment. Further useful research into this area could involve: using dynamic thermal simulation to test recommended building design features, investigating the monitoring method on a wider sample of sites and investigating air quality monitoring in hospitals.

Wastewater treatment plants (WWTP) are often the cause of malodor. The compounds which are the main causes of the odor is hydrogen sulfide (H2S), ammonia(NH3), mercaptans (RSH) and volatile organic compounds (V OC) [1]. The odorous air can be analyzed to determine the concentration of the odorants. The odorous can also be analyzed by measuring the odor. The odor is measured, a test panel of people smells the odorous air and determines how many times greater the concentration of the odorants is compared to the odor threshold which is the concentration at which a compound or a mixture is detectable by smell [3]. Measurements were done at three systems for air purication at three different locations, the Vimmerby WWTP, the Alvim WWTP and Renova's biological waste treatment facility in Gothenburg. The odor was measured at the inlet and the outlet and the concentration of H2S and ozone (O3) were measured at all the sampling points of the systems. The system at Vimmerby consisted of three CIFs followed by an UV-reactor and an AC-reactor. In Alvim there were two system which used UV and AC. The system at Renova consisted of a barrier filter followed by UV and AC. The system at the Vimmerby WWTP had a conversion rate between 87-97% of H2.S The CIFs had conversions between 50-64% of the H2S. H2S was not detected at any of the other systems. O3 was only detected at Renova where ground level O3 was present at the inlet, 0.16 ppm. The concentration increased to 0.20 ppm after the UV-reactor. The activated carbon could adsorb all the incoming O3. The odor at the Vimmerby WWTP was determined to 27500 Ou/m3 at the inlet and 19071 Ou/m3 at the outlet. The odor conversion over the system was 31%. The odor conversion at the Alvim WWTP was 99.8%. With an odor of 5490 Ou/m3 atthe inlet and 11 Ou/m3 at the outlet. The ingoing air at the system at Renova had an odor of 434 Ou/m3 and was reduced to 36 Ou/m3 at outlet. The odor conversion at Renova was 92%. To increase the accuracy of the measurements the time between the sampling and measurements should have been minimized. The test panels should also have been larger and the panelists should have been screened in advance, so results from panelists which were over and under sensitive to odors were not included in the final results.
Vattenreningsverk är ofta en källa till dåligt lukt. De kemiska föreningar som är den huvudsakliga orsaken till dålig lukt är svavelväte (H2S) ammoniak (NH3),merkaptaner (RSH) och lättflyktiga organiska föreningar (V OC) [1]. Det går att mäta lukt genom att mäta koncentrationen av föreningarna som orsakar den dåliga lukten. Det går även att avgöra hur stark lukt är genom att en testpanel får lukta på utspädda luktprover. På så vis går det att bestämma hur många gånger luften måste spädas ut för att nå lukttröskeln. Lukttröskeln definieras som den koncentration som ett ämne eller en blandning går att detektera med hjälp av lukt [3]. Matningar gjordes vid system för luftrening vid tre anläggningar, två vattenreningsverk, Vimmerby och Alvim, och Renovas behandlingsanläggning för biologåkt avfall i Göteborg. Lukten mättes vid inloppet och utloppet av sytemen. H2S och O3 mättes vid alla av systemens mätpunkter. Systemet i Vimmerby bestodav tre katalytiska järnfilter i serie följt av en UV-reaktor och aktivt kol. I Alvim fanns det två system med UV följt av aktivt kol. Hos Renova bestod systemet av ett partikelfilter följt av UV och aktivt kol. Systemet vid Vimmerbys vattenreningsverk hade en omsättning mellan 87-97% av HSS Omsättningen av H2S  över de katalytiska järnfilterna var mellan 50-64%. H2S detekterades inte vid några av de andra systemen. O3 detekterades endast vid Renovas anläggning där marknära O3 fanns i det ingående flödet, 0.16 ppm. Koncentrationen ökade till 0.20 ppm efter UV -reaktorn. Det aktiva kolet klarade avatt adsobera allt det inkommande. Lukten vid systemet vid Vimmerby mättes till 27500 Ou/m3 vid inloppet och vid utloppet till 19071 Ou/m3. Omsättningen av lukten bestämdes till 31% över systemet. Omsättningen av lukten vid Alvim bestämdes till 99.8%. Lukten vidinloppet var 5490 Ou/m3 och 11 Ou/m3 vid utloppet. Luften vid inloppet vidsystemet hos Renova var 434 Ou/m3 och 36 Ou/m3. Omsättningen av lukten hos Renova var 92%. Exaktheten för mätningar kunde ha förbättrats om tiden mellan provtagningen och analysen av provet hade minimerats. Testpanelen skulle också haft flera deltagare som skulle varit gallrade i förväg så att resultat från paneldeltagare som var över- och okänsliga för lukt inte fanns med i det slutgiltiga resultatet.

Performance monitoring counters are implemented in most recent microprocessors. In this thesis, we describe various performance measurement experiments for a program and a system that we conducted on a Linux operating system using the Pentium performance counters. We carried out our performance measurements on a Pentium II microprocessor. The Pentium II performance counters can be configured to count events such as cache misses, TLB misses, instructions executed etc. We used a low intrusive overhead technique to access these performance counters. We used these performance counters to measure the cache miss overheads due to context switches in Linux system. Our methodology involves sampling the hardware counters every 50ps. The sampling was set up using signals related to interval timers. We describe an analytical cache performance model under multiprogrammed condition from the literature and validate it using the performance monitoring counters. We next explores the long term performance of a system under different workload conditions. Various performance monitoring events - data cache h, data TLB misses, data cache reads or writes, branches etc. - are monitored over a 24 hour period. This is useful in identifying activities which cause loss of system performance. We used timer interrupts for sampling the performance counters. We develop a profiling methodology to give a perspective of performance of the different functions of a program, not only on the basis of execution-time but also on the data cache misses. Available tools like prof on Unix can be used to pinpoint the regions of performance loss of programs, but they mainly rely on an execution-time profiles. This does not give insight into problems in cache performance for that program. So we develop this methodology to get the performance of each function of the program not only on the basis of its execution time but also on the basis of its cache behavior.

Performance monitoring counters are implemented in most recent microprocessors. In this thesis, we describe various performance measurement experiments for a program and a system that we conducted on a Linux operating system using the Pentium performance counters. We carried out our performance measurements on a Pentium II microprocessor. The Pentium II performance counters can be configured to count events such as cache misses, TLB misses, instructions executed etc. We used a low intrusive overhead technique to access these performance counters. We used these performance counters to measure the cache miss overheads due to context switches in Linux system. Our methodology involves sampling the hardware counters every 50ps. The sampling was set up using signals related to interval timers. We describe an analytical cache performance model under multiprogrammed condition from the literature and validate it using the performance monitoring counters. We next explores the long term performance of a system under different workload conditions. Various performance monitoring events - data cache h, data TLB misses, data cache reads or writes, branches etc. - are monitored over a 24 hour period. This is useful in identifying activities which cause loss of system performance. We used timer interrupts for sampling the performance counters. We develop a profiling methodology to give a perspective of performance of the different functions of a program, not only on the basis of execution-time but also on the data cache misses. Available tools like prof on Unix can be used to pinpoint the regions of performance loss of programs, but they mainly rely on an execution-time profiles. This does not give insight into problems in cache performance for that program. So we develop this methodology to get the performance of each function of the program not only on the basis of its execution time but also on the basis of its cache behavior.

The wide range of applications for single input single output controllers have encouraged interest in monitoring their performance. Over the past two decades researchers in the area have found many performance enhancement opportunities by applying these techniques. These are most evident in large operational plants with hundreds of controllers being monitored at the same time. Early performance measures were based on minimum variance control as a benchmark for controller performance. Many other procedures have since emerged that have improved the level of accuracy in these performance measures. In addition, these improvements made it easier to implement control loop performance monitoring in large industrial settings. This thesis looks at the performance measures in use for single input single output controllers. The work here looks at incorporating these different measures for a specific manufacturing plant. Ways of identifying the goals and objectives of controllers in a system are presented. Furthermore, measures are proposed that most accurately indicate if these goals and objectives are being met. The concept is demonstrated on a distillation system in a gas plant. It is shown how using these objective driven techniques can provide the user with sound results. These results do not require much user analysis to identify sources of problems and areas of improvement.

Compressors are widely used in the oil and gas industry, and have been so for many years. Previously, gas turbines where the standard driver for big offshore compressors, but as an ever increasing amount of attention is put on the emission of CO2 and NOx and the electric motor technology is developing this option is becoming more popular. This shift of drive technology introduces new possibilities for condition monitoring by the use of electrical signals from the motor. That is the background for this report. The objective of this report was to describe state-of-the-art condition monitoring techniques using electrical signals and to suggest new approaches to condition monitoring made possible by these techniques. In addition monitoring approaches was to be demonstrated on example data from Statoil facilities. This report starts with a brief introduction to the different maintenance techniques and the benefits gained by using condition based maintenance. Then a short description of a generic compression system and the main components is given. The most frequent failure modes for each component are presented along with the condition monitoring approaches capable of detecting them. Vibration analysis and performance analysis is shortly described before the condition monitoring techniques using electrical signals are presented. The dominating method is named Electrical Signature Analysis. This is a non-intrusive method which analyses the current and voltage frequency spectra. Many failure modes will appear as specific frequencies in this spectra, thus failure detection and diagnostics are possible by this method. Some examples of failure frequencies are given. By the use of this method mechanically and electrically related problems can be detected, not only in the drive, but also in the gear and driven equipment. Some examples of available ESA tools are presented, including MCM and ALLSAFE PRO. A new approach for condition monitoring by the use of classical methods and ESA combined is proposed. At the end a tool named Early Fault and Disturbance Detection, developed by ABB, is presented and a short analysis of process data is performed.

Parallel architectures, like the transputer-based multicomputer network, offer potentially enormous computational power at modest cost. However, writing programs on a multicomputer to exploit parallelism is very difficult due to the lack of tools to help users understand the run-time behavior of the parallel system and detect performance bottlenecks in their programs. This thesis examines the performance characteristics of parallel programs in a multicomputer network, and describes the design and implementation of a real-time performance monitoring tool on transputers. We started with a simple graph theoretical model in which a parallel computation is represented as a weighted directed acyclic graph, called the execution graph. This model allows us to easily derive a variety of performance metrics for parallel programs, such as program execution time, speedup, efficiency, etc. From this model, we also developed a new analysis method called weighted critical path analysts (WCPA), which incorporates the notion of parallelism into critical path analysis and helps users identify the program activities which have the most impact on performance. Based on these ideas, the design of a real-time performance monitoring tool was proposed and implemented on a 74-node transputer-based multicomputer. Major problems in parallel and distributed monitoring addressed in this thesis are: global state and global clock, minimization of monitoring overhead, and the presentation of meaningful data. New techniques and novel approaches to these problems have been investigated and implemented in our tool. Lastly, benchmarks are used to measure the accuracy and the overhead of our monitoring tool. We also demonstrate how this tool was used to improve the performance of an actual parallel application by more than 50%.
Science, Faculty of
Computer Science, Department of
Graduate

The use of batch processes is widespread across the manufacturing industries, dominating sectors such as pharmaceuticals, speciality chemicals and biochemicals. The main goal in batch production is to manufacture consistent, high quality batches with minimum rework or spoilage and also to achieve the optimum energy and feedstock usage. A common approach to monitoring a batch process to achieve this goal is to use a recipe-driven approach coupled with off-line laboratory analysis of the product. However, the large amount of data generated during batch manufacture mean that it is possible to monitor batch processes using a statistical model. Traditional multivariate statistical techniques such as principal component analysis and partial least squares were originally developed for use on continuous processes, which means they are less able to cope with the non-linear and dynamic behaviours inherent within a batch process without being adapted. Several approaches to dealing with batch behaviour in a multivariate framework have been proposed including multi-way principal component analysis. A more advanced approach designed to handle the typical characteristics of batch data is that of model-based principal component. It comprises of a mechanistic model combined with a multivariate statistical technique. More specifically, the technique uses a mechanistic model of the process to generate a set of residuals from the measured process variables. The theory being that the non-linear behaviour and the serial correlation in the process will be captured by the model, leaving a set of unstructured residuals to which principal component analysis (PCA) can be applied. This approach is benchmarked against the more standard approaches including multiway principal components analysis, batch observation level analysis. One limitation identified of the model-based approach is that if the mechanistic model of the process is of reduced complexity then the monitoring and fault detection abilities of the technique will be compromised. To address this issue, the model-based PCA technique has been extended to incorporate an additional error model which captures the differences between the mechanistic model and the process. This approach has been termed super model-based PCA (SMBPCA). A number of different error models are considered including partial least squares (linear, non-linear and dynamic), autoregressive with exogenous (ARX) variables model and dynamic canonical correlation analysis. Through the use of an exothermic batch reactor simulation, the SMBPCA approach has been investigated with respect to fault detection and capturing the non-linear and dynamic behaviour in the batch process. The robustness of the technique for application in an industrial situation is also discussed.

This thesis investigates advanced optical performance monitoring approaches for future all-optical networks using the synchronous sampling technique. This allows for improved signal quality estimation, fault management and resource allocation through improved control of transmission at the physical layer level. Because of the increased transparency in next generation networks, it is not possible to verify the quality of the signal at each node because of the limited number of optical-electrical-optical conversions, and therefore new non-intrusive mechanisms to achieve signal quality monitoring are needed. The synchronous sampling technique can be deployed to estimate the bit error rate, considered an important quality measure, and hence can be utilised to certify service level agreements between operators and customers. This method also has fault identification capabilities by analysing the shapes of the obtained histograms. Each impairment affects the histogram in a specific way, giving it a unique shape that can be used for root cause analysis. However, chromatic dispersion and polarisation mode dispersion (PMD) can have similar signatures on the histograms obtained at decision times. A novel technique to unambiguously discriminate between these two sources of degradation is proposed in this work. It consists of varying the decision times so that sampling also occurs at both edges of the eye diagram. This approach is referred to as three-section eye sampling technique. In addition, it is shown that this method can be used to accurately assess first order polarisation mode dispersion and can simultaneously estimate the differential group delay (DGD) and the power splitting ratio between the two states of polarisation. Since synchronous sampling is employed, the effect of PMD on the sampling times is also investigated. For the first time, closed form relationship between the shift in sampling time, the DGD and the power splitting ratio between the polarisation states is obtained. Three types of high-Q filter based clock recovery circuits are considered: without pre-processing circuits that can be used for RZ format and with an edge detector or a squarer pre-processing circuits suitable for NRZ format. Moreover, this technique can be used to monitor chromatic dispersion and a large monitoring range of more than 1750ps/nm is experimentally demonstrated at 10Gbit/s. Since it can monitor PMD and dispersion, this method can be deployed to control dynamic PMD or dispersion compensators. Furthermore, this technique offers easy and quick inline eye mask testing and timing jitter assessment.

Energy management has been an important function of organisations since the oil crisis of the mid 1970’s led to hugely increased costs of energy. Although the financial costs of energy are still important, the growing recognition of the environmental costs of fossil-fuel energy is becoming more important. Legislation is also a key driver. The UK has set an ambitious greenhouse gas (GHG) reduction target of 80% of 1990 levels by 2050 in response to a strong international commitment to reduce GHG emissions globally. This work is concerned with the management of energy consumption in buildings through the analysis of energy consumption data. Buildings are a key source of emissions with a wide range of energy-consuming equipment, such as photocopiers or refrigerators, boilers, air-conditioning plant and lighting, delivering services to the building occupants. Energy wastage can be identified through an understanding of consumption patterns and in particular, of changes in these patterns over time. Changes in consumption patterns may have any number of causes; a fault in heating controls; a boiler or lighting replacement scheme; or a change in working practice entirely unrelated to energy management. Standard data analysis techniques such as degree-day modelling and CUSUM provide a means to measure and monitor consumption patterns. These techniques were designed for use with monthly billing data. Modern energy metering systems automatically generate data at half-hourly or better resolution. Standard techniques are not designed to capture the detailed information contained in this comparatively high-resolution data. The introduction of automated metering also introduces the need for automated analysis. This work assumes that consumption patterns are generally consistent in the short-term but will inevitably change. A novel statistical method is developed which builds automated event detection into a novel consumption modelling algorithm. Understanding these changes to consumption patterns is critical to energy management. Leicester City Council has provided half-hourly data from over 300 buildings covering up to seven years of consumption (a total of nearly 50 million meter readings). Automatic event detection pinpoints and quantifies over 5,000 statistically significant events in the Leicester dataset. It is shown that the total impact of these events is a decrease in overall consumption. Viewing consumption patterns in this way allows for a new, event-oriented approach to energy management where large datasets are automatically and rapidly analysed to produce summary meta-data describing their salient features. These event-oriented meta-data can be used to navigate the raw data event by event and are highly complementary to strategic energy management.

Wastewater treatment plants (WWT P ) are often the cause of malodor. The com-pounds which are the main causes of the odor is hydrogen sulfide (H2S), ammonia (NH3), mercaptans (RSH) and volatile organic compounds (V OC) [1]. The odorous air can be analyzed to determine the concentration of the odorants. The odorous can also be analyzed by measuring the odor. The odor is measured, a test panel of people smells the odorous air and determines how many times greater the concentration of the odorants is compared to the odor threshold which is the concentration at which a compound or a mixture is detectable by smell [3].Measurements were done at three systems for air purification at three different locations, the Vimmerby WWT P , the Alvim WWT P and Renova’s biological waste treatment facility in Gothenburg. The odor was measured at the inlet and the outlet and the concentration of H2S and ozone (O3) were measured at all the sampling points of the systems. The system at Vimmerby consisted of three CIF s followed by an UV-reactor and an AC-reactor. In Alvim there were two system which used UV and AC. The system at Renova consisted of a barrier filter followed by UV and AC.The system at the Vimmerby WWT P had a conversion rate between 87-97% of H2S. The CIF s had conversions between 50-64% of the H2S. H2S was not detected at any of the other systems. O3 was only detected at Renova where ground level O3 was present at the inlet, 0.16 ppm. The concentration increased to 0.20 ppm after the UV-reactor. The activated carbon could adsorb all the incoming O3.The odor at the Vimmerby WWT P was determined to 27500 Ou/m3 at the inlet and 19071 Ou/m3 at the outlet. The odor conversion over the system was 31%. The odor conversion at the Alvim WWTP was 99.8%. With an odor of 5490 Ou/m3 at the inlet and 11 Ou/m3 at the outlet. The ingoing air at the system at Renova had an odor of 434 Ou/m3 and was reduced to 36 Ou/m3 at outlet. The odor conversion at Renova was 92%.To increase the accuracy of the measurements he time between the sampling and measurements should have been minimized. The test panels should also have been larger and the panelists should have been screened in advance, so results from panelists which were over and under sensitive to odors were not included in the final results.

Application Performance Monitoring (APM) används i allt större utsträckning för att samla in data om mjukvarusystems prestanda och om hur användare interagerar med systemen. Detta för att säkerställa tillgänglighet och robusthet samt förbättra slutanvändarupplevelser. Syftet med denna studie är att undersöka hur införandet av APM kan gå till, vilka utmaningar som finns samt vilka kostnader och prestandaförsämringar som införandet innebär. Detta görs genom en fallstudie där APM implementeras i Enterprise Resource Planning-systemet MONITOR G5 som utvecklats av det svenska företaget Monitor ERP System AB. Systemet är utvecklat i Microsofts mjukvaruramverk .NET Framework och APM-tjänsten som används är Microsofts Application Insights. Studien resulterade i ett förslag på en APM-lösning där data om användarinteraktioner, prestanda och uppkomna fel samlas in och visualiseras. Sex instrumentpaneler skapades som visar olika aspekter av insamlade data, bland annat prestandamätvärden såsom processor- och minnesanvändning, uppkomna fel, laddningstider för vyer samt vilka delar av systemet som används mest och minst. Kostnadsanalysen visar att kostnaderna kan bli väldigt höga, men förslag på strategier för att hålla kostnaderna nere ges. Prestandatesterna som utfördes för att undersöka APM-lösningens påverkan på systemets prestanda gav otillförlitliga resultat, men det är troligt att prestandakostnaden för APM är liten.
The use of Application Performance Monitoring (APM) for collecting data about performance and end-user behaviors in complex software systems is increasing. APM is used to ensure availability and robustness and to enhance end-user experiences. This study aims to investigate how the adoption of APM can be done, what challenges organizations face during the implementation as well as costs and performance overhead associated with APM. This is achieved through a case study in which APM is introduced into the Enterprise Resource Planning (ERP) system MONITOR G5, developed and maintained by the Swedish software company Monitor ERP System AB. The system is developed in Microsoft’s .NET Framework and the APM service used is Microsoft’s Application Insights. The study resulted in a proposed APM solution wherein data regarding user interactions, performance and errors are collected and visualized. Six dashboards were created, showing different aspects of the collected data, for instance: which parts of the system is most and least frequently used, errors, load times and performance metrics such as processor and memory usage. The cost analysis shows that monetary costs can be very high, but strategies for suppressing costs are proposed. The performance tests that were conducted to determine the performance overhead of APM are inconclusive, but it is likely that the performance penalty of using APM is small.

This dissertation develops a new concept for performance-based monitoring (PBM) of instrumented buildings subjected to earthquakes. This concept is achieved by simultaneously combining and advancing existing knowledge from structural mechanics, signal processing, and performance-based earthquake engineering paradigms. The PBM concept consists of 1) optimal sensor placement, 2) dynamic response reconstruction, 3) damage estimation, and 4) loss analysis. Within the proposed concept, the main theoretical contribution is the derivation of a nonlinear model-based observer (NMBO) for state estimation in nonlinear structural systems. The NMBO employs an efficient iterative algorithm to combine a nonlinear model and limited noise-contaminated response measurements to estimate the complete nonlinear dynamic response of the structural system of interest, in the particular case of this research, a building subject to an earthquake. The main advantage of the proposed observer over existing nonlinear recursive state estimators is that it is specifically designed to be physically realizable as a nonlinear structural model. This results in many desirable properties, such as improved stability and efficiency. Additionally, a practical methodology is presented to implement the proposed PBM concept in the case of instrumented steel, wood-frame, and reinforced concrete buildings as the three main types of structural systems used for construction in the United States. The proposed methodology is validated using three case studies of experimental and real-world large-scale instrumented buildings. The first case study is an extensively instrumented six-story wood frame building tested in a series of full-scale seismic tests in the final phase of the NEESWood project at the E-Defense facility in Japan. The second case study is a 6-story steel moment resisting frame building located in Burbank, CA, and uses the recorded acceleration data from the 1991 Sierra Madre and 1994 Northridge earthquakes. The third case is a seven-story reinforced concrete structure in Van Nuys, CA, which was severely damaged during the 1994 Northridge earthquake. The results presented in this dissertation constitute the most accurate and the highest resolution seismic response and damage measure estimates obtained for instrumented buildings. The proposed PBM concept will help structural engineers make more informed and swift decisions regarding post-earthquake assessment of critical instrumented building structures, thus improving earthquake resiliency of seismic-prone communities.

The data generated in a chemical industry is a reflection of the process. With the modern computer control systems and data logging facilities, there is an increasing ability to collect large amounts of data. As there are many underlying aspects of the process in that data, with its proper utilization, it is possible to obtain useful information for process monitoring and fault diagnosis in addition to many other decision making activities. The purpose of this research is to utilize the data driven multivariate techniques of Principal Component Analysis (PCA) and Independent Component Analysis (ICA) for the estimation of process parameters. This research also includes analysis and comparison of these techniques for fault detection and diagnosis along with introduction, explanation and results from a new methodology developed in this research work namely Hybrid Independent Component Analysis (HICA).The first part of this research is the utilization of models of PCA and ICA for estimation of process parameters. The individual techniques of PCA and ICA are applied separately to the original data set of a waste water treatment plant (WWTP) and the process parameters for the unknown conditions of the process are calculated. For each of the techniques (PCA and ICA), the validation of the calculated parameters is carried out by construction of Decision Trees on WWTP dataset using inductive data mining and See 5.0. Both individual techniques were able to estimate all parameters successfully. The minor limitation in the validation of all results may be due to the strict application of these techniques to Gaussian and non-Gaussian data sets respectively. Using statistical analysis it was shown that the data set used in this work exhibits Gaussian and non-Gaussian behaviour.In the second part of this work multivariate techniques of Principal Component Analysis (PCA) and Independent Component Analysis (ICA) have been used for fault detection and diagnosis of a process along with introduction of the new technique, Hybrid Independent Component Analysis (HICA). The techniques are applied to two case studies, the waste water treatment plant (WWTP) and an Air pollution data set. As reported in literature, PCA and ICA proved to be useful tools for process monitoring on both data set, but a comparison of PCA and ICA along with the newly developed technique (HICA) illustrated the superiority of HICA over PCA and ICA. It is evident from the fact that PCA detected 74% and 67% of the faults in the WWTP data and Air pollution data set respectively. ICA successfully detected 61.3% and 62% of the faults from these datasets. Finally HICA showed improved results by the detection of 90% and 81% of the faults in both case studies. This showed that the new developed algorithm is more effective than the individual techniques of PCA and ICA. For fault diagnosis using PCA, ICA and HICA, contribution plots are constructed leading to the identification of responsible variable/s for a particular fault. This part also includes the work done for the estimation of process parameters using HICA technique as was done with PCA and ICA in the first part of the research. As expected HICA technique was more successful in estimation of parameters than PCA and ICA in line with its working for process monitoring.

With improving professionalism of sports around the world, the volume and frequency of training required for competitive performances at the elite level has increased concurrently. With this amplification in training load comes an increased need to closely monitor the associated fatigue responses, since maximising the adaptive response to training is also reliant on avoiding the negative consequences of excessive fatigue. The rationale for the experimental chapters in this thesis was established after considering survey responses regarding current best practice for monitoring fatigue in high performance sporting environments (Chapter 3). On the basis of the results, vertical jump assessments were selected for further investigation regarding their utility in determining neuromuscular fatigue responses. Outcomes from the subsequent series of studies aimed to provide practitioners working in high performance sport with guidelines for using vertical jumps to monitor athletic fatigue. The results from Chapter 4 indicate using the mean value of at least six jumps enhances the ability to detect small but practically important changes in performance from week to week. This study also highlighted large differences (4-6%) in morning and afternoon performance, indicating that the time of day performance is assessed needs to be accounted for when monitoring changes in jump performance. Chapter 5 explored the theory that the time of day effect observed in Chapter 4 can be explained by internal temperature differences. This theory was supported by demonstrating that an extended warm-up period can negate differences in jump performance in the morning and the afternoon. Researchers who are unable to standardise the time of day that assessment occurs are able, therefore, to control for performance differences by manipulating the warm-up protocols. The third study examined changes in vertical jump performance over a three month training period and produced several novel outcomes. A major finding was that unloaded jumps were more sensitive to neuromuscular fatigue during intensive training than loaded jumps (Chapter 6). Furthermore, this set of results showed that all subjects changed their jump technique via a reduction in the amplitude of the countermovement when they were highly fatigued. Using the same data, an analysis was performed to quantify individual differences in within-subject variation (Chapter 7) during normal and intensive training. These results provided the first indication that within-subject variability in vertical jump performance is substantially different between individuals and between different training phases, an important consideration for interpreting the practical importance of performance changes. In Chapter 8 the relationship between vertical jump performance and electrically elicited force of the knee extensors was examined to better understand the mechanism(s) of changes in jump performance associated with neuromuscular fatigue during intensive overload training. The results showed that the fatigue assessed by vertical jump performance was likely not only peripheral in origin as previously suggested by other authors. Further research is required to further understand the mechanisms of reduced performance during overload training, although the preliminary evidence presented implicates central mechanisms. To conclude the thesis, the findings presented in the experimental chapters are summarised, with a series of practical recommendations for using vertical jumps to monitor athletic fatigue presented.

The focus of this Thesis is on examining and explaining the effects of performance monitoring on employee performance and well-being outcomes. The use of monitoring, organisational systems and practices for managing employees' performance behaviours, is extensive, despite an ongoing and unresolved debate over its effects on employees. Job design, goal-setting and feedback, leadership and organisational justice theories were used as a basis for constructing a more comprehensive and integrative research model on performance monitoring and its effects than has been examined to date. Herein, characteristics of the monitoring process were hypothesised to motivate cognitive and attitudinal reactions in employees, in turn affecting outcomes. The role of managers was further positioned as antecedent to the employee experience of monitoring. To test the model, survey data from around one thousand call centre employees and managers were collected within two longitudinal field studies, and analysed statistically using multiple regression, structural equation modelling and latent growth modelling. Findings from the main cross-sectional analyses supported that performance monitoring had both direct and indirect effects on employee performance and well-being as a function of its utility, and that manager support was a critical factor. Learning, mental-effort, and perceptions of the fairness and privacy invasiveness of monitoring were established as explanatory mechanisms. This was the first field research to document a monitoring-performance relationship. The model was further explored on a longitudinal basis, providing limited support for the direct effects of monitoring over time. Overall, performance monitoring that developed employees was found to have the most global benefits. The theoretical and practical implications of this research are discussed, and directions for future research presented.

Performance monitoring is commonly portrayed as providing a uniquely objective, rational foundation for decisions, based on a single-stranded feedback loop between setting objectives and measuring results. In this thesis, I investigate whether this portrayal is accurate. I analyse whether the linear model underlying performance monitoring provides an adequate basis for understanding decisions about access to rural water supply in Ethiopia. My examination focuses primarily on the politics of knowledge production from three angles. First, I examine whether the assumptions underlying the definition of ‘access' to rural water used in performance monitoring in Ethiopia, adequately represent the divergent notions of access among the relevant actors. My findings show that formal framings of access, codified in national and international guidelines and benchmarks, focus on technical aspects of the water supply infrastructure. I bring to light that the goal of performance monitoring in relation to achieving ‘access' is driven by the methods used to measure it, mainly the parameters of infrastructure, volume, distance and quality, suggesting a circularity between framings of the inputs to and objectives of appraisal. In this self-referential process, a particular image of the world determines the meaning of performance, which is used as a yardstick. The power of this dynamic is apparent in Ethiopian stakeholders' characterisations of access, which, even when critical, revolve narrowly around these dominant parameters. This one-dimensional and technical framing of access, constantly reproduced in self-referential monitoring circles, contrasts starkly with the multi-dimensional and dynamic nature of the water access experiences of local residents in Ethiopia. Second, I test whether, in reality, monitoring processes conform to the linearity assumed by the feed-back function of performance monitoring. The process tracing method used to illuminate the political and power dynamics of monitoring processes, shows that sector government actors at different administrative levels, with different rationales, provide different stories of ‘access'. Viewed from this perspective, performance monitoring can be seen not, as conventionally asserted, as a uniquely rational appraisal of performance, but rather as being about ‘performing monitoring' – the playing out on a management stage of certain politically-necessary performances. At the same time, I find that numerous less formalised monitoring practices proceed in parallel with the formal PM process, which, together, form a body of largely ‘tacit' knowledge that informs sector stakeholders' daily work. It is this wider body of knowledge, rather than only formalised PM results, that informs decisions. Third, I investigate the mechanisms that led to the formulation of specific decisions associated with rural water access and the role in these of performance monitoring. I find that particular decisions, such as repairs to rural water schemes, have multiple causes, among which performance monitoring is a contributory and necessary, but not sufficient factor. My investigation of criteria affecting budget allocations highlights that sector offices' limited control over them contributes to making strategic planning a rubber stamping exercise whose processes can be characterised as ‘muddling through' as opposed to adhering to the linear model suggested by Results-based Management. My findings highlight the need to break the self-referential cycle of narrowly framed performance monitoring exercises. They suggest greater attention to the ‘tacit' monitoring practices in local settings, and a focus on the process of monitoring and the power relations within it, to complement the dominant focus on monitoring targets and indicators.

Wireless personal area networks have emerged as an important communication infrastructure in areas such as at-home healthcare and home automation, independent living and assistive technology, as well as sports and wellness. Wireless personal area networks, including body sensor networks, are becoming more mature and are considered to be a realistic alternative as communication infrastructure for demanding services. However, to transmit data from e.g., an ECG in wireless networks is also a challenge, especially if multiple sensors compete for access. Contention-based networks offer simplicity and utilization advantages, but the drawback is lack of predictable performance. Recipients of data sent in wireless sensor networks need to know whether they can trust the information or not. Performance measurements, monitoring and control is of crucial importance for medical and healthcare applications in wireless sensor networks. This thesis focuses on development, prototype implementation and evaluation of a performance management system with performance and admission control for wireless sensor networks. Furthermore, an implementation of a new method to compensate for clock drift between multiple wireless sensor nodes is also shown. Errors in time synchronization between nodes in Bluetooth networks, resulting in inadequate data fusion, are also analysed.

QC 20120529

In this study, a new gas turbine monitoring system being able to carry out appropriate run process is set up for a gas turbine with 250 kW power rating and its accessories. The system with the mechanical and electrical connections of the required sub-parts is transformed to a kind of the test stand. Performance test result calculation method is described. In addition that, performance evaluation software being able to apply with the completion of the preliminary performance tests is developed for this gas turbine. This system has infrastructure for the gas turbine sub-components performance and aerothermodynamics research. This system is also designed for aviation training facility as a training material for the gas turbine start and run demonstration. This system provides the preliminary gas turbine performance research requirements in the laboratory environment.

Most of the advanced microprocessors today incorporate on-chip hardware performance counters. These counters are capable to count various events in a non-invasive way, while executing real workloads. Events such as the number of instructions, memory accesses, cache and TLB misses are the most common ones that can be precisely measured. The primary accomplishment of this work was to implement a performance monitoring tool, which could be used to evaluate system behaviour on high-end processing platforms. The tool is able to collect data from hardware performance counters and present them in an interpretable way. Moreover, it has support for two different platforms and two operating systems. As a secondary objective, several measurements were carried out on both supported platforms and operating systems to demonstrate the tool’s capabilities, and to solve the potential use-cases.
De flesta av dagens mikroprocessorer innehåller prestandaräknare direkt i hårdvaran. Dessa räknare kan räkna olika typer av händelser på ett icke störande sätt medans hårdvaran är under last. Händelser såsom instruktioner, minnesaccesser, cache och TLB missar är de vanligast förekommande räknarna som kan göras precisa. Det främsta genomförandet i denna uppgift var att implementera att verktyg för att övervaka prestanda som kan användas för att beräkna ett systems beteende på högprestanda-plattformar. Verktyget kan hämta prestandaräknarna och presentera dem i ett läsbart format. Dessutom har verktyget stöd för två olika plattformar och det översattes till två olika operativsystem. Som ett sekundärt mål gjordes många mätningar på de båda plattformarna samt operativsystemen som stöddes för att visa verktygets funktion och lösa potentiella användningsfall.

Beach protection is today emerging as one of the most relevant environmental issues at the global level. The increasing vulnerability of beaches to human impact and the effect of climate change are determining an increasing risk which implies a significant socioeconomic threat. In fact, erosion of beaches and coastlines is observed with increasing frequency, with implications on the societal resilience to natural hazards. Several different techniques can be applied for protecting beaches and the seashore, ranging from structural methods to green and innovative solutions that are the subject of increasing attention in recent times. The aim of this thesis is to study innovative defense works against the erosion of the coast: in particular, the study focused on a submerged concrete barrier prototype, called WMESH, whose first modules have been recently installed in an area of the sea bad of the Emilia-Romagna coast. It is a permeable submerged barrier made up of reinforced concrete with a geometry that can dissipate a substantial portion of the wave energy, to promote the sand flow down to the shore and counteract its return. The first monitoring of the positioned modules and statistical tests on the wave and tide conditions in the affected area were performed in order to see the structure's response. These tests allowed to perform computer simulations with the 2D MIKE 21 program: the obtained results show that the structure responds positively, creating a reduction in the significant wave height and generating a sediment accumulation area in the area surrounding it.

This thesis examined the benefits of physiological and performance testing of elite swimmers. The study considered the following research questions: the degree to which physiological and performance measures in training contribute to swimming performance; sources and magnitude of variability in testing, training and competition performance; the magnitudes of changes in test measures during routine training; and the reliability, validity and utility of miniaturised and automated smart sensor technology to monitor the stroke and performance times of swimmers in training. The experimental approach involved the retrospective analysis of five years of physiological and performance testing of elite level swimmers, the development of a new accelerometry-based smart sensor device to monitor swimmers in the pool, a cross-sectional study comparing the physiological and performance responses of swimmers of different levels, and the effects of an intensive 14-day training program on submaximal physiological and performance measures. Collectively, the outcomes of these studies provide a strong justification for the physiological and performance testing of elite swimmers, a quantitative framework for interpreting the magnitude of changes and differences in test scores and sources of variation, and highlight the potential utility of new smart sensor technology to automate the monitoring of a swimmer�s training performance. The first study (Chapter 2) characterises the changes and variability in test performance, physiological and anthropometric measures, and stroke mechanics of swimmers within and between seasons over their elite competitive career. Forty elite swimmers (24 male, 16 female) performed a 7 x 200-m incremental swimming step test several times each 6-month season (10 � 5 tests, spanning 0.5 to 6.0 y). Mixed linear modeling provided estimates of change in the mean and individual responses for measures based on submaximal performance (fixed 4-mM lactate), maximal performance (the seventh step), and lean mass (from skinfolds and body mass). Submaximal and maximal swim speed increased within each season from the pre to taper phase by ~2.2% for females and ~1.5% for males (95% confidence limits �1.0%), with variable contributions from stroke rate and stroke length. Most of the gains in speed were lost in the off-season, leaving a net average annual improvement of ~1.0% for females and ~0.6% for males (�1.0%). For submaximal and maximal speed, individual variation between phases was �2.2% and the typical measurement error was �0.8%. In conclusion, step test and anthropometric measures can be used to confidently monitor progressions in swimmers in an elite training program within and between seasons. The second study (Chapter 3) quantified the relationship between changes in test measures and changes in competition performance for individual elite swimmers. The primary question addressed was whether test measures could predict a swimmers performance at the major end-of-season competition. The same sample group as in Study 1 was examined. A 7 x 200-m incremental swimming step-test and anthropometry were conducted in up to four training phases each season. Correlations of changes in step-test and anthropometric measures between training phases between and within seasons, with changes in competition performance between seasons, were derived by repeated-measures mixed modeling and linear regression. Changes in competition performance were best tracked by changes in test measures between taper phases. The best single predictor of competition performance was skinfolds for females (r = -0.53). The best predictor from the step-test was stroke rate at 4-mM lactate (females, r = 0.46; males, r = 0.41); inclusion of the second-best step-test predictor in a multiple linear regression improved the correlations marginally (females, r =0.52 with speed in the seventh step included; males, r = 0.58 with peak lactate concentration included). Changes in test measures involving phases other than the taper provided weak and inconclusive correlations with changes in performance, possibly because the coaches and swimmers took corrective action when tests produced poor results. In conclusion, a combination of fitness and techniques factors are important for competitive performance. The step test is apparently a useful adjunct in a swimmer�s training preparation for tracking large changes in performance. These initial studies identified stroke mechanics as a major determinant of a swimmer�s performance. Chapter 4 details the development of a small tri-axial accelerometry-based smart sensor device (the Traqua) that enables continual monitoring of various performance/stroke characteristics in swimming. The initial focus was to develop a device that automated the detection of a swimmer�s movements, specifically lap times, stroke rate and stroke count. The Traqua consists of a tri-axial accelerometer packaged with a microprocessor, which attaches to the swimmer at the pelvis to monitor their whole body movements while swimming. This study established the failure/error rate in the first generation algorithms developed to detect the swimming-specific movements of stroke identification, laps (start, turn and finish), and strokes (stroke count and stroke rate) in a cohort of 21 elite and sub-elite swimmers. Movements were analysed across a range of swimming speeds for both freestyle and breaststroke. These initial algorithms were reasonably successful in correctly identifying the markers representing specific segments of a swimming lap in a range of swimmers across a spectrum of swimming speeds. The first iteration of the freestyle algorithm produced error-rates of 13% in detection of lap times, 5% for stroke rate, and 11% for stroke count. Subsequent improvements of the software reduced the error rate in lap and stroke detection. This improved software was used in the following two studies. The next study (Chapter 5) evaluated the reliability and validity of the Traqua against contemporary methods used for timing, stroke rate and stroke count determination. The subjects were 14 elite and 10 sub-elite club-level swimmers. Each swimmer was required to swim seven evenly paced 200-m efforts on a 5-min cycle, graded from easy to maximal. Swimmers completed the test using their main competitive stroke (21 freestyle, 3 breaststroke). Timing was compared for each 50-m lap and total 200-m time by electronic touch pads, video coding, a hand-held manual stopwatch, and the Traqua. Stroke count was compared for video coding, self-reported counting, and the Traqua, while the stroke rate was compared via video coding, hand-held stopwatch, and the Traqua. Retest trials were conducted under the same conditions 7 d following the first test. All data from the Traqua presented in this and the subsequent studies were visually inspected for errors in the automated algorithms, where the algorithms had either failed to correctly identify the start, turn, finish or individual strokes and corrected prior to analysis. The standard error of the estimate for each of the timing methods for total 200 m was compared with the criterion electronic timing. These standard errors were as follows: Traqua (0.64 s; 90% confidence limits 0.60 � 0.69 s), Video (0.52 s; 0.49 � 0.55 s); Manual (0.63 s; 0.59 � 0.67 s). Broken down by 50-m laps, the standard error of the estimate for the Traqua compared with the electronic timing for freestyle only was: 1st 50-m 0.35 s; 2nd and 3rd 50-m 0.13 s; 4th 50-m 0.65 s. When compared with the criterion video-coding determination, the error for the stroke count was substantially lower for the Traqua (0.6 strokes.50 m-1; 0.5 � 0.6 strokes.50 m-1) compared to the self-reported measure (2.3 strokes.50 m-1; 2.5 � 2.9 strokes.50 m-1). However, the error for stroke rate was similar between the Traqua (1.5 strokes.min-1; 1.4 � 1.6 strokes.min-1) and the manual stopwatch (1.8 strokes.min-1; 1.7 � 1.9 strokes.min-1). The typical error of measurement of the Traqua was 1.99 s for 200-m time, 1.1 strokes.min-1 for stroke rate, and 1.1 strokes.50 m-1 for stroke count. In conclusion, the Traqua is comparable in accuracy to current methods for determining time and stroke rate, and better than current methods for stroke count. A substantial source of error in the Traqua timing was additional noise in the detection of the start and finish. The Traqua is probably useful for monitoring of routine training but electronic timing and video are preferred for racing and time trials. Having established the reliability and validity of the Traqua, Chapter 6 addressed the ability to discriminate the pattern of pacing between different levels of swimmers in the 7 x 200-m incremental step test. This study also sought to quantify the differences in pacing between senior and junior swimmers. Eleven senior elite swimmers (5 female, 6 male) and 10 competitive junior swimmers (3 female, 7 male) participated in this study. Each swimmer was required to swim seven evenly paced 200-m freestyle efforts on a 5-min cycle, graded from easy to maximal. The Traqua was used to measure time, stroke rate and stroke count. The senior swimmers were better able to descend in each of the 200-m efforts. Overall the senior swimmers were ~2-3 s per 50 m faster than the junior swimmers. Both groups were fastest in the first 50-m lap with the push start. The senior swimmers then descended the 50- m time for each of the subsequent laps, getting ~0.5 s faster per lap, with the final lap the fastest. In contrast, the junior swimmers swam a similar time for each of the subsequent laps. The junior swimmers were marginally more variable in their times (coefficient of variation: ~2%) compared with the senior swimmers (~1.8%). In comparison to junior swimmers, the senior swimmers in this study were faster, adopted a more uniform negative split strategy to pacing within a 200-m effort, and were more consistent in reproducing submaximal and maximal swimming speeds. The final study (Chapter 7) analysed the effect of 14-d of intensive training on the reproducibility of submaximal swimming performance in elite swimmers. Submaximal physiological and performance testing is widely used in swimming and other individual sports but the variability in test measures, and the effects of fatigue, during intensive training have surprisingly not been quantified systematically. Seven elite swimmers (3 male and 4 female) participated in an intensive 14-d training camp one month prior to the National championships. The aim of the study was to characterise the intra-session, daily and training block variability of submaximal swimming time, physiological and stroke characteristics in elite swimmers. The swimmers performed a specified submaximal 200-m effort in most sessions, after the warm-up and at the end of the session for both morning and afternoon sessions. During the efforts, swimming time and stroke mechanics were measured and physiological measures were recorded immediately on completion. The Traqua was worn by all swimmers in every training session. Mixed linear modeling was used to provide estimates of changes in the mean and individual responses (within-athlete variation as a coefficient of variation) for all measures. The swimmers were moderately slower (1.4%; �1.4%) over the 14-d training camp. The mean submaximal 200-m effort was very likely to be faster (0.7%; confidence limits �0.7%) in the afternoon compared with the morning session. The females were more variable in their submaximal performance times (CV=2.6%) than the male swimmers (1.7%). Blood lactate concentration was almost certainly lower (-23%; �10%) following higher volume in the previous session; however a higher intensity workout the previous session almost certainly leads to higher lactate (21%; �15%) in the current session. Considered together, these results indicate that the 200-m submaximal test is useful in monitoring submaximal physiological and performance measures and the negative effects of cumulative fatigue. In conclusion, changes in the physiological and performance measures derived from the poolbased progressive incremental step test are moderately correlated with changes in end-ofviii. season competition performance. The magnitudes of changes and differences in test measures between phases within a season, from season to season, and between males and females, established in this study can be applied to similar elite level swimmers preparing for major competition. The quantification of typical error of the same measures demonstrates that coaches and scientists can distinguish real and worthwhile improvements using the 7 x 200-m step test. Continual pool-based monitoring with the automated smart sensor Traqua device may provide more accurate and detailed information about a swimmer�s training adaptation than current fitness tests and monitoring methods. Finally, submaximal testing in trained swimmers is useful in monitoring progress in physiological and performance measures, and the impact of cumulative fatigue during an intensive period of training. Collectively, the outcomes of these studies indicate that routine physiological and performance testing can provide measurable benefits for elite swimmers and their coaches.