<p>There are many aspects to consider when designing system identification experiments in control applications. Input design is one important issue. This thesis considers input design both for identification of linear time-invariant models and for stability validation.</p><p>Models obtained from system identification experiments are uncertain due to noise present in measurements. The input spectrum can be used to shape the model quality. A key tool in input design is to introduce a linear parametrization of the spectrum. With this parametrization a number of optimal input design problems can be formulated as convex optimization programs. An Achilles' heel in input design is that the solution depends on the system itself, and this problem can be handled by iterative procedures where the input design is based on a model of the system. Benefits of optimal input design are quantified for typical industrial applications. The result shows that the experiment time can be substantially shortened and that the input power can be reduced.</p><p>Another contribution of the thesis is a procedure where input design is connected to robust control. For a certain system structure with uncertain parameters, it is shown that the existence of a feedback controller that guarantees a given performance specification can be formulated as a convex optimization program. Furthermore, a method for input design for multivariable systems is proposed. The constraint on the model quality is transformed to a linear matrix inequality using a separation of graphs theorem. The result indicates that in order to obtain a model suitable for control design, it is important to increase the power of the input in the low-gain direction of the system relative to the power in the high-gain direction.</p><p>A critical issue when validating closed-loop stability is to obtain an accurate estimate of the maximum gain of the system. This problem boils down to finding the input signal that maximizes the gain. Procedures for gain estimation of nonlinear systems are proposed and compared. One approach uses a model of the system to design the optimal input. In other approaches, no model is required, and the system itself determines the optimal input sequence in repeated experiments.</p>

There are many aspects to consider when designing system identification experiments in control applications. Input design is one important issue. This thesis considers input design both for identification of linear time-invariant models and for stability validation. Models obtained from system identification experiments are uncertain due to noise present in measurements. The input spectrum can be used to shape the model quality. A key tool in input design is to introduce a linear parametrization of the spectrum. With this parametrization a number of optimal input design problems can be formulated as convex optimization programs. An Achilles' heel in input design is that the solution depends on the system itself, and this problem can be handled by iterative procedures where the input design is based on a model of the system. Benefits of optimal input design are quantified for typical industrial applications. The result shows that the experiment time can be substantially shortened and that the input power can be reduced. Another contribution of the thesis is a procedure where input design is connected to robust control. For a certain system structure with uncertain parameters, it is shown that the existence of a feedback controller that guarantees a given performance specification can be formulated as a convex optimization program. Furthermore, a method for input design for multivariable systems is proposed. The constraint on the model quality is transformed to a linear matrix inequality using a separation of graphs theorem. The result indicates that in order to obtain a model suitable for control design, it is important to increase the power of the input in the low-gain direction of the system relative to the power in the high-gain direction. A critical issue when validating closed-loop stability is to obtain an accurate estimate of the maximum gain of the system. This problem boils down to finding the input signal that maximizes the gain. Procedures for gain estimation of nonlinear systems are proposed and compared. One approach uses a model of the system to design the optimal input. In other approaches, no model is required, and the system itself determines the optimal input sequence in repeated experiments. / QC 20101109

A wastewater treatment plant has the task to refine the wastewater from substances that should not be released into the environment. The decomposition process can, in a simple way, be described as follows: micro-organisms breathe oxygen and eat unwanted substances. The environment in which the micro-organism lives is known as sludge and the correct amount of oxygen available in the sludge is important for the decomposition to be effective. The oxygenation of the sludge alone stands for about 30% of the plant’s energy consumption. The purpose of this thesis is to propose a controller for a new technology which oxygenates the sludge. In this project the oxygenation process is modeled. The model is then used to validate the proposed controllers by simulations. The result from this thesis is a PI controller with anti-windup. Other results are a monitoring system that can be used to detect toxic substances in the plant and an approach to, through the control of the oxygen level, minimize the plant’s power consumption.

Då detta skrivs (Maj 2018) står industrier på tröskeln till att genomgå sitt fjärde paradigmskifte i och med Industri 4.0. I detta paradigmskifte, som har sin grund i digitalisering av såväl befintliga som nya industrianläggningar, kommer en ny typ av underhållsarbete att möjliggöras. Den nya typen av underhållsarbete är en form av prediktivt underhåll och innebär att smarta system kommer att kunna detektera och klassificera systemavvikelser, och genom detta sålunda prediktivt varna för att något håller på att fallera hos såväl hela maskiner som dess komponenter. Vid Högskolan i Gävle bedrevs det från hösten 2017 till våren 2018 ett forskningsprojekt vid namn Flexibla modeller för smart underhåll, i vilket modeller för prediktivt underhåll togs fram. Denna examensuppsats på grundnivå har som målsättning att beskriva hur en demonstrationsutrustning framtas åt Flexibla modeller för smart underhåll för att demonstrera hur den nya typen av prediktivt underhåll kan ske. De metoder som tillgrips och utvärderas för att skapa modeller är från området systemidentifiering. Det system som modelleras består av en pump, en vattenreservoar, två vattentankar och en ventil. Både black-boxidentifiering och grey-boxidentifiering utförs på systemet. Mer specifikt ARX- och OE-modeller testas i Matlab-applikationen System Identification och fysikalisk modellering med parameterskattning via minsta kvadratmetoden utvärderas.  För feldetektering och felklassificering granskas två olika typer av maskininlärningsalgoritmer från Matlab-applikationen Classification Learner. Dessa är Support Vector Machine och K-nearest neighbour. Det erhållna resultatet visar att fysikalisk modellering med parameterskattning via minsta kvadrat-metoden ger den bästa modellen, där FIT-måtten hos de två tankarna är 88,75 procent respektive 90,76 procent. Med denna modell som underlag framtas sedan detekterings- och klassificeringsalgoritmer, där den mest framgångsrika algoritmen för detektering av öppnad ventil är Fine Gaussian SVM och för öppnad ventil eller simulerad sprickbildning i botten av den undre tanken är algoritmen Fine KNN den bästa. Slutsatsen av detta examensarbete är att det är möjligt att via metoder från Systemidentifiering och maskininlärning bygga en demonstrationsutrustning som åskådliggör såväl Flexibla modeller för smart underhåll:s arbete som hur den nya typen av prediktivt underhåll kan ske. / When this is written (May 2018) industries are on the threshold to experience their fourth paradigm shift because of Industry 4.0. This paradigm shift, which has its foundation in digitalizing existing as well as new industries, will make a new sort of maintenance work possible. The new type of maintenance work is a sort of predictive maintenance and means that smart systems will be able to detect and classify system abnormalities, and thereby predictively warn that something is about to fail in machines as well as in their components. At University of Gävle there were in a period from autumn of 2017 to the spring of 2018 a research project named Flexible models for predictive maintenance, in which models which could be a part of predictive maintenance were developed. The objective in this bachelor thesis is to show how to construct an equipment for Flexible models for predictive maintenance that can be used to demonstrate how the new type of predictive maintenance can be done. The methods that are used and evaluated to generate models belong to the field of system identification. The system that modelled consists of a pump, one water reservoir, two tanks and a valve. Both black-box identification and grey-box identification are performed on the system. More specifically ARX- and OE-models are tested in the Matlab application System Identification and physical modelling with parameter estimation by the method of least squares is evaluated. To detect and classify system failures, two types of machine learning algorithms from the Matlab application Classification Learner is audited. These algorithms are Support Vector Machine and K-nearest neighbour. The results show that physical modelling with parameter estimation by the method of least squares gives the best model, where the measure of FIT for the two tanks are 88.75 percent respectively 90.76 percent. With the model as basis detection- and classification algorithms are developed, where Fine Gaussian SVM turns out to be the most successful in detecting an open valve and Fine KNN performs the best at detecting an open valve or a simulated crack in the bottom of the lower tank. The conclusion of this thesis is that it is possible to construct an equipment by using methods from system identification and machine learning that visualizes the work in Flexible models for predictive maintenance as well as how the new type of predictive maintenance can be performed.

<p>ABB Surge Arrestors in Ludvika have for a long time had a problem to keep the moisture at a steady state in the production of their ZnO-powder, that is used to produce varistors.</p><p>Some black-box models of the spray drying process have been designed and evaluated to find a solution for the problem. After evaluating the collected data it has been found that variations in the supply voltage causes control difficulties for the operator. A cascade control system was designed, consisting of three PI control loops designed with lambda tuning. The disturbance in the supply voltagewas used as a feed forward in the control system.</p><p>At the end of the project the control system was installed, and tests were made to verify the functionality of the regulator. It was shown that most of the variations in the moisture of the powder could be eliminated using small resources, through purchase of a process controller and four power meters. The</p><p>standard deviation in the moisture was decreased from a level of 0.32%, measured when the process was manually controlled, down to 0.07% measured when the control system was used. This also solved the given problem.</p>

Patients with idiopathic normal pressure hydrocephalus (INPH) have a disturbance in the cerebrospinal fluid (CSF) system. The treatment is neurosurgical – a shunt is placed in the CSF system. The infusion test is used to assess CSF system dynamics and to aid in the selection of patients that will benefit from shunt surgery. The infusion test can be divided into three parts: a mathematical model, an infusion protocol and a parameter estimation method. A non-linear differential equation is used to mathematically describe the CSF system, where two important parameters are the outflow conductance (Cout) and the Pressure Volume Index (PVI). These are used both for clinical and research purposes. The analysis methods for the non-linear CSF system have limited the infusion protocols of presently used infusion investigations. They come with disadvantages such as long investigation time, no estimation of PVI and no measure of the reliability of the estimates. The aim of this dissertation was to develop and evaluate novel methods for infusion protocols, mathematical modelling and parameter estimation methods for assessment of CSF system dynamics. The infusion protocols and parameter estimation methods in current use, constant pressure infusion (CPI), constant infusion and bolus infusion, were investigated. The estimates of Cout were compared, both on an experimental set-up and on 20 INPH patients. The results showed that the bolus method produced a significantly higher Cout than the other methods. The study suggested a method with continuous infusion for estimating Cout and emphasized that standardization of Cout measurement is necessary. The non-linear model of the CSF system was further developed. The ability to model physiological variations that affect the CSF system was incorporated into the model and it was transformed into a linear time-invariant system. This enabled the use of methods developed for identification of such systems. The underlying model for CSF absorption was discussed and the effect of baseline resting pressure (Pr) in the analysis on the estimation of Cout was explored using two different analyses, with and without Pr. A novel infusion protocol with an oscillating pressure pattern was introduced. This protocol was theoretically better suited for the CSF system characteristics. Three new parameter estimation methods were developed. The adaptive observer was developed from the original non-linear model of the CSF system and estimated Cout in real time. The prediction error method (PEM) and the robust simulation error (RSE) method were based on the transformed linear system, and they estimated both Cout and PVI with confidence intervals in real time. Both the oscillating pressure pattern and the reference CPI protocol were performed on an experimental set-up of the CSF system and on 47 hydrocephalus patients. The parameter estimation methods were applied to the data, and the RSE method produced estimates of Cout that were in good agreement with the reference method and allowed for an individualized and considerably reduced investigation time. In summary, current methods have been investigated and a novel approach for assessment of CSF system dynamics has been presented. The Oscillating Pressure Infusion method, which includes a new infusion protocol, a further developed mathematical model and new parameter estimation methods has resulted in an improved way to perform infusion investigations and should be used when assessing CSF system dynamics. The advantages of the new approach are the pressure-regulated infusion protocol, simultaneous estimation of Cout and PVI and estimates of reliability that allow for an individualized investigation time.

ABB Surge Arrestors in Ludvika have for a long time had a problem to keep the moisture at a steady state in the production of their ZnO-powder, that is used to produce varistors. Some black-box models of the spray drying process have been designed and evaluated to find a solution for the problem. After evaluating the collected data it has been found that variations in the supply voltage causes control difficulties for the operator. A cascade control system was designed, consisting of three PI control loops designed with lambda tuning. The disturbance in the supply voltagewas used as a feed forward in the control system. At the end of the project the control system was installed, and tests were made to verify the functionality of the regulator. It was shown that most of the variations in the moisture of the powder could be eliminated using small resources, through purchase of a process controller and four power meters. The standard deviation in the moisture was decreased from a level of 0.32%, measured when the process was manually controlled, down to 0.07% measured when the control system was used. This also solved the given problem.

<p>The jetengines of today are growing in complexity. Reliability for aircraft engines are of extreme importance, mainly due to safety reasons but also economical ones. This master thesis deals with faultdiagnosis in the turbine section of RM12, the engine used in Saab/BAe's Gripen. Three different faults which can occur in the turbine section was studied. These faults are: clogged fuel nozzle, hole in outlet guide vane and sensor fault. An analysis of the behaviour of the engine with these faults present was made. Based on this analysis an existing simulation model of RM12 was modified, so that these faults could be simulated. For the purpose of fault diagnosis two models were developed for two different engine parameters, one linear state space model and a neural network. These two models are then used to isolate the faults. The linear state space model is used to estimate the temperature right behind the engine turbines. This is a state space model with two states. This model estimates the temperature well at higher throttle levels, but has a temperature discrepancy of almost 100 K at lower throttle levels, the temperature right behind the turbines varies between 300 and 1200 K. A neural network was estimated to detect a decrease in turbine efficiency which is a phenomena which occurs when one or several of the engine's eighteen fuel nozzles are clogged. The neural network was able to detect this fault at some points. The diagnosis algorithm developed, based on the models mentioned above, is able to detect faults at most operating points, but fails to isolate the present fault at some points.</p>

<p>This master thesis covers the development of an automatic tuning process for the existing speed controller for drive systems. The drive systems are resonant two-mass systems where a motor is used to drive a load connected by a shaft. The developed method relies heavily on system identification and the construction of a complete mechanical model of the process. With this approach, the common problem with poor load speed control that derives from measuring only the motor speed can be addressed and solved for a majority of such processes.</p><p>The automatic tuning method has along with general test functions been implemented in a complete tool for automatic tuning, testing and performance evaluation and reporting for drive systems.</p>

The jetengines of today are growing in complexity. Reliability for aircraft engines are of extreme importance, mainly due to safety reasons but also economical ones. This master thesis deals with faultdiagnosis in the turbine section of RM12, the engine used in Saab/BAe's Gripen. Three different faults which can occur in the turbine section was studied. These faults are: clogged fuel nozzle, hole in outlet guide vane and sensor fault. An analysis of the behaviour of the engine with these faults present was made. Based on this analysis an existing simulation model of RM12 was modified, so that these faults could be simulated. For the purpose of fault diagnosis two models were developed for two different engine parameters, one linear state space model and a neural network. These two models are then used to isolate the faults. The linear state space model is used to estimate the temperature right behind the engine turbines. This is a state space model with two states. This model estimates the temperature well at higher throttle levels, but has a temperature discrepancy of almost 100 K at lower throttle levels, the temperature right behind the turbines varies between 300 and 1200 K. A neural network was estimated to detect a decrease in turbine efficiency which is a phenomena which occurs when one or several of the engine's eighteen fuel nozzles are clogged. The neural network was able to detect this fault at some points. The diagnosis algorithm developed, based on the models mentioned above, is able to detect faults at most operating points, but fails to isolate the present fault at some points.