https://www.mdu.se/

The energy generation landscape is changing, pushed by stricter regulations for emissions control and green energy generation. The limitations of renewable energy sources, however, require flexible energy production sources to supplement them. Micro gas turbine based combined heat and power plants, which are used for domestic applications, can fill this gap if they become more reliable. This can be achieved with the use of an engine monitoring and diagnostics system: real-time engine condition monitoring and fault diagnostics results in reduced operating and maintenance costs and increased component and engine life. In order to allow the step change in the connection of small engines to the grid, a fleet monitoring system for micro gas turbines is required. A proposed framework combines a physics-based model and a data-driven model with machine learning capabilities for predicting system behavior, and includes a purpose-developed diagnostic tool for anomaly detection and classification for a multitude of engines. The framework has been implemented on a fleet of micro gas turbines and some of the lessons learned from the demonstration of the concept as well as key takeaways from the general literature are presented in this paper. The extension of fleet monitoring to optimal operation and production planning in relation to the needs of the grid will allow the micro gas turbines to fit in the future green energy system, connect to the grid, and trade in the energy market. The requirements on the system level for the widespread use of micro gas turbines in the energy system are addressed in the paper. A review of the current solutions in fleet monitoring and diagnostics, generally developed for larger engines, is included, with an outlook into a sustainable future.

Real-time engine condition monitoring and fault diagnostics results in reduced operating and maintenance costs and increased component and engine life. Prediction of faults can change the maintenance model of a system from a fixed maintenance interval to a condition based maintenance interval, further decreasing the total cost of ownership of a system. Technologies developed for engine health monitoring and advanced diagnostic capabilities are generally developed for larger gas turbines, and generally focus on a single system; no solutions are publicly available for engine fleets. This paper presents a concept for fleet monitoring finely tuned to the specific needs of micro gas turbines. The proposed framework includes a physics-based model and a data-driven model with machine learning capabilities for predicting system behaviour, combined with a diagnostic tool for anomaly detection and classification. The integrated system will develop advanced diagnostics and condition monitoring for gas turbines with a power output under 100 kW.

This paper presents an overview of different methods used in what is normally called AI-methods today. The methods have been there for many years, but now have built a platform of methods complementing each other and forming a cluster of tools to be used to build “learning systems”. Physical and statistical models are used together and complemented with data cleaning and sorting. Models are then used for many different applications like output prediction, soft sensors, fault detection, diagnostics, decision support, classifications, process optimization, model predictive control, maintenance on demand and production planning. In this chapter we try to give an overview of a number of methods, and how they can be utilized in process industry applications.

Data-driven algorithms require large and comprehensive training samples in order to provide reliable diagnostic solutions. However, in many gas turbine applications, it is hard to find fault data due to proprietary and liability issues. Operational data samples obtained from end-users through collaboration projects do not represent fault conditions sufficiently and are not labeled either. Conversely, model-based methods have some accuracy deficiencies due to measurement uncertainty and model smearing effects when the number of gas path components to be assessed is large. The present paper integrates physics-based and data-driven approaches aiming to overcome this limitation. In the proposed method, an adaptive gas path analysis (AGPA) is used to correct measurement data against the ambient condition variations and normalize. Fault signatures drawn from the AGPA are used to assess the health status of the case engine through a Bayesian network (BN) based fault diagnostic algorithm. The performance of the proposed technique is evaluated based on five different gas path component faults of a three-shaft turbofan engine, namely intermediate-pressure compressor fouling (IPCF), high-pressure compressor fouling (HPCF), high-pressure turbine erosion (HPTE), intermediate-pressure turbine erosion (IPTE), and low-pressure turbine erosion (LPTE). Robustness of the method under measurement uncertainty has also been tested using noise-contaminated data. Moreover, the fault diagnostic effectiveness of the BN algorithm on different number and type of measurements is also examined based on three different sensor groups. The test results verify the effectiveness of the proposed method to diagnose single gas path component faults correctly even under a significant noise level and different instrumentation suites. This enables to accommodate measurement suite inconsistencies between engines of the same type. The proposed method can further be used to support the gas turbine maintenance decision-making process when coupled with overall Engine Health Management (EHM) systems.

Predictive health monitoring of micro gas turbines can significantly increase the availability and reduce the operating and maintenance costs. Methods for predictive health monitoring are typically developed for large-scale gas turbines and have often focused on single systems. In an effort to enable fleet-level health monitoring of micro gas turbines, this work presents a novel data-driven approach for predicting system degradation over time. The approach utilises operational data from real installations and is not dependent on data from a reference system. The problem was solved in two steps by: 1) estimating the degradation from time-dependent variables and 2) forecasting into the future using only running hours. Linear regression technique is employed both for the estimation and forecasting of degradation. The method was evaluated on five different systems and it is shown that the result is consistent (r>0.8) with an existing method that computes corrected values based on data from a reference system, and the forecasting had a similar performance as the estimation model using only running hours as an input

Access to energy is fundamental to economic and technological advancement. Hence, the more the world develops, the greater the demand for energy becomes. Evidently, the production and consumption of energy alone account for more than 80% of global anthropogenic greenhouse gas (GHG) emissions. There is broad scientific consensus that efficiency improvements in energy production and consumption must come first on the path to reducing global GHG emissions. As the largest producer and consumer of energy, the industrial sector faces tremendous challenges due to stringent environmental regulations, intense price-based global competition, rising operating costs and rapidly changing economic conditions. Therefore, increasing energy and resource efficiency while improving throughput and asset reliability is a matter of utmost importance. Satisfying such demanding objectives requires an optimal operation, control and monitoring of plant assets and processes. This is one of the main driving forces behind developing digital solutions, methods, and frameworks that can be integrated with old and new industrial automation platforms. The main focus of this dissertation is to investigate frameworks, process models, soft sensors, control optimization, and diagnostic techniques to improve the operation, control, and monitoring of industrial plants and processes. In this thesis, a generic architecture for control optimization, diagnostics, and decision support system, referred to here as a learning system, is proposed. The research is centred around an investigation of different components of the proposed learning system. Two very different case studies, one representing large-scale assets and another representing a fleet of small-scale assets, are considered to demonstrate the genericness of the proposed system architecture. In this thesis, a very energy-intensive chemical pulping process represents the case study of large-scale assets, and a micro gas turbine (MGT) fleet for distributed heat and power generation represent the case study of a fleet of small-scale assets. One of the main challenges in this research arises from the marked differences between the case studies in terms of size, functions, quantity, and structure of the existing automation systems. Typically, only a few pulp digesters are found in a Kraft pulping mill, but there may be hundreds of units in a MGT fleet. The main argument behind the selection of these two case studies is that, if the proposed learning system architecture can be adapted for these significantly different cases, then it can be adapted for many other industrial applications as well. Within the scope of this thesis, mathematical modelling, model adaptation, model predictive control, and diagnostics methods are studied for continuous pulp digesters, whereas mathematical modelling, model adaptation, and diagnostics techniques are explored for the MGT fleet. Due to the naturally varying wood quality, significant residence time, insufficient measurements, and complexity of pulping reactions, modelling and controlling a continuous pulp digester is a challenging task. Moreover, process abnormalities due to non-ideal flow in the digester often occur that considerably affect the pulp quality. Within this dissertation, variation of wood-chip quality is identified as one of the main process disturbances. Thereafter, a feedforward model predictive control (MPC) approach is explored by feedforwarding the lignin content of the wood chips to the controller. The result shows that the disturbance rejection and tracking performance of the feedforward MPC are superior to other alternatives, like Proportional–integral–derivative (PID), MPC, and current industrial control. When it comes to diagnostics, a literature gap is identified in the area of modelling digester faults. Hence, the well-known Purdue model, a widely used dynamic model of the digester, is extended to simulate process faults like screen-clogging, hangups, and channelling. The findings suggest that both hangups and channelling considerably affect the pulp quality at the blowline. The impact of channelling is prominent on reaction temperature compared to hangups, while hangups change the residence time of the wood chips significantly. Subsequently, a hybrid diagnostics scheme for pulp digester, combining a physical model and a Bayesian network (BN), is demonstrated. Overall, the results show that fault type and severity can be estimated with acceptable accuracy even in presence of noise. Enabling remote fleet diagnostics is expected to foster the commercialization of distributed micro-combined heat and power (micro-CHP) generators, i.e., MGTs. Even though the modelling and diagnostics of large-scale gas turbines are well researched, studies targeting MGT are limited. In this thesis, a physical model of a commercial MGT system is developed. Subsequently, a hybrid scheme by combining a physics-based gas path analysis with a data-driven approach is used to enable MGT diagnostics. The proposed scheme was tested by simulating case studies corresponding to single and multiple faults. Furthermore, sensitivity studies are performed for different measurement uncertainties (i.e., sensor noise and bias) to evaluate the robustness of the scheme against measurement uncertainties. The findings show that the proposed diagnostics approach performs satisfactorily even under measurement uncertainties. To sum up, the increased availability of data and higher computing power is fostering the development of accurate process models and algorithms necessary for optimal operation, control, and monitoring of industrial processes. With the emergence of new measurement techniques, it is possible to leverage productivity and quality with tighter control of key process parameters. Additionally, studying the underlying mechanism of process degradation and developing diagnostics methods by incorporating these can lead to significant economic benefits. Having said that, to tap the full potential of these digital solutions, an integrated framework like that presented in this thesis, i.e., a learning system is essential.

Driven by intense competition, increasing operational cost and strict environmental regulations, the modern process and energy industry needs to find the best possible way to adapt to maintain profitability. Optimization of control and operation of the industrial systems is essential to satisfy the contradicting objectives of improving product quality and process efficiency while reducing production cost and plant downtime. Use of optimization not only improves the control and monitoring of assets but also offers better coordination among different assets. Thus, it can lead to considerable savings in energy and resource consumption, and consequently offer a reduction in operational costs, by offering better control, diagnostics and decision support. This is one of the main driving forces behind developing new methods, tools and frameworks that can be integrated with the existing industrial automation platforms to benefit from optimal control and operation. The main focus of this dissertation is the use of different process models, soft sensors and optimization techniques to improve the control, diagnostics and decision support for the process and energy industry. A generic architecture for an optimal control, diagnostics and decision support system, referred to here as a learning system, is proposed. The research is centred around an investigation of different components of the proposed learning system. Two very different case studies within the energy-intensive pulp and paper industry and the promising micro-combined heat and power (CHP) industry are selected to demonstrate the learning system. One of the main challenges in this research arises from the marked differences between the case studies in terms of size, functions, quantity and structure of the existing automation systems. Typically, only a few pulp digesters are found in a Kraft pulping mill, but there may be hundreds of units in a micro-CHP fleet. The main argument behind the selection of these two case studies is that if the proposed learning system architecture can be adapted for these significantly different cases, it can be adapted for many other energy and process industrial cases. Within the scope of this thesis, mathematical modelling, model adaptation, model predictive control and diagnostics methods are studied for continuous pulp digesters, whereas mathematical modelling, model adaptation and diagnostics techniques are explored for the micro-CHP fleet.

Being at the heart of modern pulp mills, continuous pulp digesters have attracted much attention from the research community. In this article, a comprehensive review in the area of modeling, control and diagnostics of continuous pulp digesters is conducted. The evolution of research focus within these areas is followed and discussed. Particular effort has been devoted to identifying the state-of-the-art and the research gap in a summarized way. Finally, the current and future research directions in the areas have been analyzed and discussed. To date, digester modeling following the Purdue approach, Kappa number control using model predictive controllers and health index-based diagnostic approaches by utilizing different statistical methods have dominated the field. While the rising research interest within the field is evident, we anticipate further developments in advanced sensors and integration of these sensors for improving model prediction and controller performance; and the exploration of different AI-based approaches will be at the core of future research.

Kappa number variability at the continuous digester outlet is a major concern for pulp and paper mills. It is evident that the aforementioned variability is strongly linked to the feedstock wood properties, particularly lignin content. Online measurement of lignin content utilizing near-infrared spectroscopy at the inlet of the digester is paving the way for tighter control of the blow-line Kappa number. In this paper, an innovative approach of feedforwarding the lignin content to a model predictive controller was investigated with the help of modeling and simulation studies. For this purpose, a physics-based modeling library for continuous pulp digesters was developed and validated. Finally, model predictive control approaches with and without feedforwarding the lignin measurement were evaluated against current industrial control and proportional-integral-derivative (PID) schemes. 

Due to the naturally varying feedstock, significant residence time, insufficient measurements and complex nature of the delignification process, producing pulp with consistent quality i.e. stable kappa number with sufficiently high yield is a challenging task that requires multi-variable process control. A wide variety of control structures, ranging from classical concepts like cascade control, feedforward, ratio control, and parallel control to more modern concepts like model-based predictive control, is used in pulp and paper industries all over the world. In this paper, a survey of model-based predictive control will be presented along with the control challenges that lie within the chemical pulping process. The potential of this control concept for overcoming the aforementioned technical challenges will also be discussed in the second part of the paper. Particular focus will be given on the use of near-infrared spectroscopy based soft-sensors coupled with dynamic process models as an enabler for feedforward model-based predictive control. Overall, the proposed control concept is expected to significantly improve process performance, in the presence of measurement noise and various complex chemical process uncertainties common in pulp and paper applications.

Kappa number, which essentially indicates the amount of lignin left in the pulp after cooking, is the most important physical quantity linked to the quality and economics of a Kraft-pulp mill. Controlling the Kappa number is a difficult task mainly due to the naturally varying feedstock, significant residence time, insufficient measurements and complex nature of the delignification process. Moreover, faults such as screen clogging, hang-ups and channeling in the process often occur and increase the operational costs considerably. In this work, the possibility of feedforwarding the lignin content of incoming wood chips, by a near-infrared spectroscopic measurement of one of the major process disturbances, to a model predictive controller, is investigated by means of modeling and simulation studies. Additionally, a simple Bayesian network based diagnostics approach is proposed to detect the continuous digester faults.

The operation of continuous pulp digesters plays a pivotal role in the profitability of a pulp mill. Particularly, process faults due to the non-ideal flow behaviour in the digester have a direct impact on the pulp quality, production rate and chemical utilization of the plant. The study begins with an introduction to process faults in continuous reactors by explicitly focusing on continuous pulp digesters. Operational data from a commercial pulp digester is analysed to examine process abnormalities. Subsequently, the extended Purdue model is modified to simulate non-ideal flows related to continuous digester faults. The effect of different process faults on important process variables is analysed with the help of simulation studies. In the second part of the paper, a hybrid fault diagnostics approach based on the first-principle digester model and Bayesian network is proposed and validated. Additionally, the methodology of developing the diagnostics approach are also described.

Due to the intense price-based global competition, rising operating cost, rapidly changing economic conditions and stringent environmental regulations, modern process and energy industries are confronting unprecedented challenges to maintain profitability. Therefore, improving the product quality and process efficiency while reducing the production cost and plant downtime are matters of utmost importance. These objectives are somewhat counteracting, and to satisfy them, optimal operation and control of the plant components are essential. Use of optimization not only improves the control and monitoring of assets, but also offers better coordination among different assets. Thus, it can lead to extensive savings in the energy and resource consumption, and consequently offer reduction in operational costs, by offering better control, diagnostics and decision support. This is one of the main driving forces behind developing new methods, tools and frameworks. In this chapter, a generic learning system architecture is presented that can be retrofitted to existing automation platforms of different industrial plants. The architecture offers flexibility and modularity, so that relevant functionalities can be selected for a specific plant on an as-needed basis. Various functionalities such as soft-sensors, outputs prediction, model adaptation, control optimization, anomaly detection, diagnostics and decision supports are discussed in detail.

The market for the small-scale micro gas turbine is expected to grow rapidly in the coming years. Especially, utilization of commercial off-the-shelf components is rapidly reducing the cost of ownership and maintenance, which is paving the way for vast adoption of such units. However, to meet the high-reliability requirements of power generators, there is an acute need of a real-time monitoring system that will be able to detect faults and performance degradation, and thus allow preventive maintenance of these units to decrease downtime. In this paper, a micro gas turbine based combined heat and power system is modelled and used for development of physics-based diagnostic approaches. Different diagnostic schemes for performance monitoring of micro gas turbines are investigated.

The correct and early detection of incipient faults or severe degradation phenomena in gas turbine systems is essential for safe and cost-effective operations. A multitude of monitoring and diagnostic systems were developed and tested in the last few decades. The current computational capability of modern digital systems was exploited for both accurate physics-based methods and artificial intelligence or machine learning methods. However, progress is rather limited and none of the methods explored so far seem to be superior to others. One solution to enhance diagnostic systems exploiting the advantages of various techniques is to fuse the information coming from different tools, for example, through statistical methods. Information fusion techniques such as Bayesian networks, fuzzy logic, or probabilistic neural networks can be used to implement a decision support system. This paper presents a comprehensive review of information and decision fusion methods applied to gas turbine diagnostics and the use of probabilistic reasoning to enhance diagnostic accuracy. The different solutions presented in the literature are compared, and major challenges for practical implementation on an industrial gas turbine are discussed. Detecting and isolating faults in a system is a complex problem with many uncertainties, including the integrity of available information. The capability of different information fusion techniques to deal with uncertainty are also compared and discussed. Based on the lessons learned, new perspectives for diagnostics and a decision support system are proposed.