Browsing by Author "Karhela, Tommi"

Now showing 1 - 20 of 24

Application of reinforcement learning for energy consumption optimization of district heating system
(2023-08-31) Deng, Jifei; Eklund, Miro; Sierla, Seppo; Savolainen, Jouni; Niemistö, Hannu; Karhela, Tommi; Vyatkin, Valeriy
A4 Artikkeli konferenssijulkaisussa
Heating residential spaces consumed 64 percent of total household energy consumption in Finland. Considering the heat transfer and time delay in the district heating system, the calculation of setpoints of supply temperature requires a comprehensive understanding of the real system, and experienced operators need to manually determine the setpoints. To save energy, a more effective and accurate method is needed for setpoints calculation. In this paper, a reinforcement learning based method is proposed. Through interacting with an Apros-based simulation model, the agents learn to calculate supply temperature parallelly for lowering energy costs. Simulation results show that the proposed method outperforms the existing method and has the potential to address the problem in real factories.
Applying graph matching techniques to enhance reuse of plant design information
(2019-05-01) Rantala, Miia; Niemistö, Hannu; Karhela, Tommi; Sierla, Seppo; Vyatkin, Valeriy
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
This article investigates how graph matching can be applied to process plant design data in order to support the reuse of previous designs. A literature review of existing graph matching algorithms is performed, and a group of algorithms is chosen for further testing. A use case from early phase plant design is presented. A methodology for addressing the use case is proposed, including graph simplification algorithms and node similarity measures, so that existing graph matching algorithms can be applied in the process plant domain. The proposed methodology is evaluated empirically on an industrial case consisting of design data from several pulp and paper plants.
Automated Process Modelling in Safety-Critical Systems
(2020-01-21) Rimppi, Alpi
Perustieteiden korkeakoulu | Master's thesis
Process industry utilises various simulation tools to test and validate system performance and functionalities. Throughout testing is especially crucial for safety-critical systems, such as nuclear power plants. Nuclear energy production is heavily regulated and extensive safety analysis is requires. However, modelling complex processes by hand is a laborous task that introduces human errors. By automating the modelling workflow model quality is increased and modelling workload reduced. Functional requirements for an automated modelling tool were identified by a literature review on nuclear engineering and process simulation. Based on the requirements an automated model generation tool for Apros simulation platform was implemented. The tool uses piping and instrumentation diagrams and equipment parameter spreadsheets to construct process simulation models based on predefined transformation rules. Models generated using the tool were successfully imported into Apros. The results show that automation solutions can increase modelling consistency and ease the modelling workflow. Further development is required to expand source material support and add tools for change management.
Automatic Generation of a High-Fidelity Dynamic Thermal-hydraulic Process Simulation Model from a 3D Plant Model
(2018-08-13) Santillán Martínez, Gerardo; Sierla, Seppo; Karhela, Tommi; Lappalainen, Jari; Vyatkin, Valeriy
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
Dynamic thermal-hydraulic simulation models have been extensively used by process industry for decision support in sectors such as power generation, mineral processing, pulp and paper, and oil and gas. Ever-growing competitiveness in the process industry forces experts to rely even more on dynamic simulation results to take decisions across the process plant lifecycle. However, time-consuming development of simulation models increases model generation costs, limiting their use in a wider number of applications. Detailed 3D plant models, developed during early plant engineering for process design, could potentially be used as a source of information to enable rapid development of high-fidelity simulation models. This paper presents a method for automatic generation of a thermal-hydraulic process simulation model from a 3D plant model. Process structure, dimensioning and component connection information included in the 3D plant model is extracted from the machine-readable export of the 3D design tool and used to automatically generate and configure a dynamic thermal-hydraulic simulation model. In particular, information about the piping dimensions and elevations is retrieved from the 3D plant model and used to calculate head loss coefficients of the pipelines and to configure the piping network model. This step, not considered in previous studies, is crucial for obtaining high-fidelity industrial process models. The proposed method is tested using a laboratory process and the results of the automatically generated model are compared with experimental data from the physical system as well as with a simulation model developed using design data utilized by existing methods on the state-of-the-art. Results show that the proposed method is able to generate high-fidelity models which are able to accurately predict the targeted system, even during operational transients.
Automatic Generation of a Simulation-based Digital Twin of an Industrial Process Plant
(2018) Santillán Martínez, Gerardo; Sierla, Seppo; Karhela, Tommi; Vyatkin, Valeriy
A4 Artikkeli konferenssijulkaisussa
A Digital Twin (DT) of a production plant is a digital replica of the plant’s physical assets which contains the structure and the dynamics of how the devices and process operate. Simulation-based DTs (SBDTs) are those based on online first principle simulation models. In these systems, model parameter estimation techniques keep an online plant simulator in the same state as the targeted device or process. As a result, non-measured information of the current state of the plant can be obtained from the model. SBDTs can be used for a number of important applications and they have various advantages compared to DTs based on data-driven models. However, wider industrial adoption of SBDTs is hindered by laborious development of their underlying first principle simulation model as well as by a lack of integrated lifecycle-wide implementation methods and simulation architectures. This paper focuses on applying previously presented methods for reducing implementation effort of SBDTs. Firstly, laborious simulation model development is tackled by applying an automatic model generation method. Secondly, an integrated implementation methodology of a lifecycle-wide online simulation architecture is followed for developing the SBDT. The results show a higher level of fidelity compares to previous publications. A SBDT of a laboratory-scale process is implemented to demonstrate the proposed method.
Coordinated Execution of Multiple Process and Automation Simulators
(2020-10-19) Milan, Henrik
Sähkötekniikan korkeakoulu | Master's thesis
Industries around the globe are taking the leap into a digital future. In the field of simulation, the digital future encourages simulator vendors to provide online services for their customers. These services encourage interoperability between multiple simulators. Multiple simulators co-operating jointly is called co-simulation, which has been an active area of research in the recent years. This thesis will research modern co-simulation solutions and build a prototypical co-simulation coordinator. The goal of the co-simulation coordinator is to coordinate the execution of multiple process and automation simulators. The research method used in this thesis focuses on an OPC UA based solution for data exchange between simulators during the co-simulation scenario. The developed co-simulation coordinator is a software component that is built in the Java programming language. The developed software component is tested with Apros process simulator and ABB System 800xA by utilizing OPC UA in co-simulation. The obtained results and the wide adoption of OPC UA in industrial simulators encourages to continue the usage of OPC UA technologies in co-simulation. The results of the thesis will be used for interconnecting Apros with other industrial simulators in the future.
Deep reinforcement learning for fuel cost optimization in district heating
(2023-12) Deng, Jifei; Eklund, Miro; Sierla, Seppo; Savolainen, Jouni; Niemistö, Hannu; Karhela, Tommi; Vyatkin, Valeriy
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
This study delves into the application of deep reinforcement learning (DRL) frameworks for optimizing setpoints in district heating systems, which experience hourly fluctuations in air temperature, customer demand, and fuel prices. The potential for energy conservation and cost reduction through setpoint optimization, involving adjustments to supply temperature and thermal energy storage utilization, is significant. However, the inherent nonlinear complexities of the system render conventional manual methods ineffective. To address these challenges, we introduce a novel learning framework with an expert knowledge module tailored for DRL techniques. The framework leverages system status information to facilitate learning. The training is performed by employing model-free DRL methods and a refined digital twin of the Espoo district heating system. The expert module, accounting for power plant capacities, ensures actionable directives aligned with operational feasibility. Empirical validation through comprehensive simulations demonstrates the efficacy of the proposed approach. Comparative analyses against manual methods and evolutionary techniques highlight the approach’s superior ability to curtail fuel costs. This study advances the understanding of DRL in district heating optimization, offering a promising avenue for enhanced energy efficiency and cost savings.
Defining test scenarios for simulation aided control system testing
(2016-10-31) Savolainen, Roope
Sähkötekniikan korkeakoulu | Master's thesis
Automated process control increases safety and enables better environmental sustainability, productivity and quality. Process industry relies increasingly on more automation intensive processes. Furthermore connecting advanced control solutions to upper level automation applications has eased along with the rise of industrial internet. This brings new possibilities to the industry, but is also making the systems more complex. Increased automation has made the processes vulnerable to malfunctions in process instrumentation and actuators. Furthermore verifying the functionality and quality of controls applications developed for these complex cyber-physical systems is difficult as control application requires the physical subsystems for functioning. The physical subsystems are however typically developed apart from the development of control application and the control application is connected to the physical system only in the final part of the development project. The solution is to simulate physical system in a virtual environment and connect the control system to the virtual instances of the physical subsystems. In this thesis a test automation framework prototype for simulation aided testing of control systems is designed. Finally the implemented prototype is used to test a process controls systems of an industrial process. The testing is done by exploiting modern testing techniques like keyword-driven testing and scenario testing.
Design challenges of an ontology-based modelling and simulation environment
(2007) Villberg, Antti
Helsinki University of Technology | Master's thesis
Improvement of the software process for renewing legacy simulation software on top of the Simantics platform: A case study
(2013) Kallio, Arto
School of Science | Master's thesis
Simantics is an open-source integration platform for modelling and simulation tools. It was originally developed by VIT Technical Research Centre of Finland. KCL-ECO is a legacy life-cycle assessment tool currently being renewed on top of the Simantics platform. In this thesis the software development processes of Simantics development team at VIT were studied and analysed based on observations and interviews. The KCL-ECO project was chosen as the target for a more detailed case study. Improvements for the current process are suggested based on literature study of state of the art agile software development methodologies. Implementation of serum practices had started before this study was conducted. Still a number of conflicts were identified between the current process and the requirements of scrum. Application developers are highly dependent on external platform support even though scrum teams should be independent. Quality of many of Simantics components is poor; components are not generic and lack documentation. Application developers are unable to implement product backlogs in order since they often have to wait for platform support. All developers must switch between tasks and/or projects often and have a huge amount of work in progress. The most significant suggestion is to prefer static, larger, co-located and more independent project teams to facilitate high-bandwidth face-to-face co-operation and learning. Larger teams have a wider set of skills and are less dependent on external support. Small projects should be implemented by a multi-project team constructing their sprint from several product backlogs. Project teams should set a realistic definition of done for themselves and make it more demanding over time. Teams should estimate their progress based on past velocity to enable better decision making. Simantics knowledge should be shared through pair programming especially for new developers. Platform developers should only add new requests to the platform backlog that they commit to implement in the near future. This would avoid filling the backlog with tasks that are never implemented and reduce unnecessary waiting. Improving the platform documentation should be started by commenting the source code. Some of the platform developers' work should be outsourced to others to make them more responsive. The platform codebase should be moved to Godthab to have more sandboxes for platform development, a review process and to gain more visibility for the Simantics.
Integrated simulation-based process and automation engineering
(2015-10-22) Paganus, Niklas
Sähkötekniikan korkeakoulu | Master's thesis
Process and automation design of process plants are performed using computer aided design tools in a managed engineering workflow. Process simulation has traditionally not been used as an integrated part of the workflow in early design phases. Dynamic process simulation can be utilised in plant design to mitigate risks and reduce costs of design, commissioning and operation and improve safety. Dynamic process simulation shall be integrated into the workflow by reducing the effort needed for exploiting simulation. This requires that process simulation tools are integrated with process and automation design tools. In this thesis the characteristics of integrated simulation-based process and automation engineering in process plant design was studied and an integration based on a standard engineering data format was designed, implemented and tested. Dynamic process simulation software Apros was integrated with automation design tools. Automation basic design is performed in Apros and the Apros automation data is transferred into detailed automation design in standard format. The IEC 61131-3-based PLCopen XML was chosen as the data format. Apros automation data was mapped into the standard format and transformed using model transformation. The generated automation data was imported into automation design tools for validation. The results confirm that process simulation software can be integrated with design tools using standard data formats to enable an integrated simulation-based engineering workflow.
Machine Recognition of Engineering Diagrams in Process Industry
(2020-10-20) Qu, Rui
Perustieteiden korkeakoulu | Master's thesis
Engineering diagrams are widely used in process industry as a standard graphic language to represent engineering schematics and convey information. Over the years, a large amount of legacy engineering diagrams has been accumulated in companies so that there is an increasing demand on digitizing the diagrams to improve productivity. With the progress of computer vision, especially deep learning-based object detection, we take advantage of the latest deep learning models and algorithms to process and recognize the legacy diagrams, which fuel the networks. In this thesis, an end-to-end digitizing model is proposed to recognize engineering diagrams as machine encoded format. Due to the complexity of diagrams, the recognition task is divided into three sub-targets: symbols, connectivity and characters. We experiment on multiple state-of-the-art deep learning-based approaches to recognize symbols, such You Only Look Once (YOLO). Line recognition algorithm is proposed based on Hough transformation and Skeletonization. For the characters, we conduct the recognition by two steps, first locating, second recognizing. The model is evaluated on real industry engineering diagrams with quantitative and visual results provided. Firstly, the results demonstrate that YOLO works well for symbol recognition, reaching over 90% mAP@0.75 of all symbols. YOLO can also be used for character locating, where the characters are regarded as a symbol. Secondly, connection lines can be recognized effectively by the proposed algorithm combining Hough Transformation with region of interests. The result of recognizing the three sub-targets are integrated to generate a DXF format diagram. There is also some discussion on a universal model which can expand the usage of our model to different types of diagrams. One of the most essential steps is the analysis of source diagrams and data preparation, which is time consuming dirty work but can significantly improve the recognition performance.
Modularisation Issues in Dynamic Equation-Based Process Simulation Solvers - Comparison of Two Different Approaches
(2003) Friman, Mika
Helsinki University of Technology | Master's thesis
Object detection in engineering diagrams with scarce training data
(2024-09-27) Niemistö, Jaakko
School of Electrical Engineering | Master's thesis
Engineering diagrams, such as piping and instrumentation diagrams (P&ID), are these days intended to be maintained as intelligent diagrams and linked to applications such as operation and maintenance systems, as well as computer-aided engineering (CAE) systems. However, a significant portion of existing diagrams are still in the form of raster images. Converting raster images to more easily processed and modified vector graphics can be performed utilizing neural networks as the diagrams largely consist of repeating symbols. The drawbacks of machine learning methods include erroneous detection and large time investment of both the training data composing and the training itself. Furthermore, disparate diagram sets apply slightly different symbols even if the general form of a symbol is universal. Symbol classes also differ. As a result, current solutions are limited to individual diagram sets, requiring a significant portion of the material to be applied to training the detection model. This is time consuming and isn’t viable for small diagram sets. This thesis aims to explore solutions to improve the performance of machine learning based object detection in engineering diagrams. Advanced machine learning techniques such as few-shot learning could minimize the time investment and the required training data size in individual cases. Additionally, generating synthetic training data by augmenting existing diagrams can aid with the lack of material. Few-shot algorithms that include a base training and a fine-tuning stage are tested. The results from those are compared to a YOLO-based approach that only consists of base training without data from the target diagram set. The datasets applied to base training and few-shot fine-tuning are constructed from available P&ID engineering diagrams, and the base training dataset is augmented via scaling to increase symbol variety and volume.
Ontology-Based Diagram methods in Process Modelling and Simulation
(2007) Lehtonen, Tuukka
Helsinki University of Technology | Master's thesis
Prosessimalleja konfiguroidaan usein piirtämällä virtauskaavioita tai PI- kaavioita, joita voidaan pitää alla olevaa tehdastietomallia kuvaavina kaavamaisina piirroksina. Kaavioita, jotka on liitetty tehdastietomalliin, kutsutaan prosessiteollisuudessa älykkäiksi kaavioiksi. Integroimalla simulaatiopalveluita älykkääseen kaavionpiirtoympäristöön voidaan tarjota suunnittelijalle mahdollisuus mallien jatkuvaan testaukseen ja täten suunnitteluvirheiden nopeampaan havaitsemiseen. Simulaatiotuloksia voidaan myös visualisoida kaavioilla, jolloin mallintaja saa paremman käsityksen prosessin käyttäytymisestä. Kyseisellä tavalla simulaatiota ja visualisointia yhdistävistä kaavioista on käytetty teollisuudessa nimitystä käyttäytyvät kaaviot. Tässä työssä esitellään grafikkasovelluskehys ja ontologiapohjainen vuokaavionpiirtotyökalu, jotka tulevat toimimaan pohjana käyttäytyvien kaavioiden ympäristön kehitykselle. Määrittelemällä ontologioiden välisiä kuvauksia, voidaan kyseisellä työkalulla pitää simulaatiomalli yhtenäisenä kaaviomallin kanssa. Ontologioita määritellään simulaatiomallille, 2D graflikalle, kaavioille sekä ontologioiden välisille kuvauksille. Mallinnus pohjautuu yhtenäiseen graafitietomalliin, joka mahdollistaa tiedon vapaan yhdistelyn. Työssä esitetään määritellyt ontologiat, ohjelmistototeutukset, sekä simuloitava esimerkkiprosessimalli ja peilataan tuloksia määritelty ja vaatimuksia vasten.
OPC-pohjainen tiedonsiirto laajan mittakaavan dynaamisessa prosessisimuloinnissa
(2001) Peltoniemi, Jyrki
Helsinki University of Technology | Master's thesis
Kehitys dynaamisen prosessisimuloinnin alueella on luonut tarpeen tehokkaalle ja helppokäyttöiselle tiedonsiirtomenetelmälle simulaattorin ja muiden sovellusten välille. Yksi tärkeä esimerkki tällaisesta tarpeesta on tiedon siirtäminen hajautetun automaatiojärjestelmän ja dynaamisen prosessisimulaattorin välillä. Integroitua järjestelmää, jossa automaatiojärjestelmä ja simulaattori ovat kytkettynä yhteen, voidaan käyttää operaattorien koulutukseen sekä automaation suunnitteluun ja testaukseen. Tällainen kytkentä täysimittaisen automaatiosovelluksen ja simulointimallin välillä vaatii suurta kapasiteettia tiedonsiirtojärjestelmältä. OPC (OLE for Process Control) on laajalti käytetty komponenttipohjainen määrittely automaatioalan järjestelmien väliseen tiedonsiirtoon. Tämä työ esittelee arkkitehtuuritason asioita, jotka vaikuttavat OPC -pohjainen tiedonsiirron tehokkuuteen. Myös yksityiskohtaisia ohjelmistoteknisiä suunnittelupiirteitä esitellään. Dynaamiselle prosessisimulaattorille Aprosille on kehitetty uusi OPC -palvelin ja mitattu toteutuksen tehokkuus. Tuloksia vertaillaan vanhaan toteutukseen, joka osoittaa selvästi uuteen arkkitehtuuriin perustuvan järjestelmän edut. Tulokset osoittavat, että OPC-pohjainen tiedonsiirto tarjoaa riittävän suorituskyvyn suurimpaan osaan sovelluskohteista. Lopuksi annetaan ohjeita tehokkaan OPC -asiakkaan ohjelmointiin ja luodaan katsaus mahdollisuuksiin, joita tehokas ja joustava tiedonsiirto tarjoaa.
Process Simulation Models as Part of a Process Plant Life-Cycle Information Environment
(2005) Karlsson, Erik
Helsinki University of Technology | Master's thesis
Integration of a simulation model to a life-cycle information environment is studied in this thesis. The aim is to find a solution for automatic simulation model creation from the process design information stored inside the life-cycle information environment. First the architecture of the simulator is studied. Modularised simulation architecture is shown to be a good selection for an integrated simulator. This makes it possible to extend the capability to simulate models. This is important for an integrated simulator because the simulated process objects and the requested level of simulation details of the model are not known in advance. The CAPE-OPEN specification for simulator modularisation is studied but it is shown not to be suitable as given when the simulation model is built from large number of smaller models. This is due to its usage of the component technology to modularise the software that causes significant efficiency penalty. The other part of the integration problem is the information management inside the life-cycle information environment. It is studied by integrating simulator and design information specific data models with each other. It is shown that in a life-cycle information environment a global data model between the specific data models makes the integration problem more manageable. The information transformation between the data models can be problematic requiring logic to derive the information from the target model to the source model. During the life-cycle of the project changes to the central data model can be needed. It is shown that this can be handled by using a data model framework that allows the change of the data model or by manually constructing new data model and copying the old information to it. The study shows that a simulator can be integrated with a life-cycle information environment. This requires that the simulation ability of the simulator can be extended and the information integration problems inside the life-cycle information environment are handled.
Requirement verification in simulation-based automation testing
(2016-02-15) Siivola, Eero
Sähkötekniikan korkeakoulu | Master's thesis
An important question in planning automation of industrial processes is whether or not it satisfies requirements imposed on it. The emergence of industrial internet combined with advanced control solutions has increased the number of cyber-physical connections between the system to be controlled and its automation. This has introduced new, complex dependencies to the system and thus made it harder to check whether the system satisfies its requirements. New advanced simulation-based design methodologies have made it possible to find find requirement violations before the actual physical system exists. Requirement verification consists of two problems: how to formalize the requirements in an unambiguous way, and how to check whether or not the system satisfies all requirements. An algorithm used in conventional software testing, which monitors a single simulation run and outputs whether or not a requirement formalized with an unambiguous language called Metric Temporal Logic is modified and implemented in this master's thesis. A case study where the algorithm implementation is used on the requirements imposed on a automation of a real industrial process is then conducted to test the algorithm and tool build around it. The results of the case study show that the algorithm and the tool built around it work well and that Metric Temporal Logic can be used to formalize requirements typical for at least some kind of automation systems. However, the case study also reveals a problem related to specific kind of requirements. At the end of the thesis, a possible solution to this problem is presented and extending of the algorithm with this solution is proposed as a topic for further research.
A software architecture for configuration and usage of process simulation models : software component technology and XML-based approach
(2002-12-11) Karhela, Tommi
Doctoral dissertation (monograph)
Increased use of process simulation in different phases of the process and automation life cycle makes the information management related to model configuration and usage more important. Information management increases the requirements for more efficient model customisation and reuse, improved configurational co-use between different simulators, more generic extensibility of the simulation tools and more flexible run-time connectivity between the simulators and other applications. In this thesis, the emphasis is on large-scale dynamic process simulation of continuous processes in the power, pulp and paper industries. The main research problem is how to apply current information technologies, such as software component technology and XML, to facilitate the use of process simulation and to enhance the benefits gained from using it. As a development task this means developing a new software architecture that takes into account the requirements of improved information management in process simulation. As a research objective it means analysing whether it is possible to meet the new requirements in one software architecture using open specifications developed in information and automation technologies. Process simulation is analysed from the points of view of standardisation, current process simulation systems and simulation research. A new architectural solution is designed and implemented. The degree of meeting the new requirements is experimentally verified by testing the alleged features using examples and industrial cases. The main result of this thesis is the design, description and implementation of a new integration architecture for the configuration and usage of process simulation models. The original features of the proposed architecture are its openness, general distribution concept and distributed extensibility features.
Synchronized Cooperative Simulation: OPC UA Based Approach
(2012) Miettinen, Tuomas
Master's thesis
Most simulation tools excel at only one technical domain. For efficient simulation of multi-domain systems, cooperative simulation (co-simulation) can be used. In co-simulation, a simulation model is divided into smaller submodels to allow each of the submodels to be simulated with a purpose-made simulator. The connectivity between the multiple simulators is a key factor in the performance of a co-simulation. In this work, the OPC UA standard was chosen as the communication interface between the different simulators. OPC UA is considered an effective communication interface and, moreover, the versatility of OPC UA allows the same interface to be utilized by the user to control and configure the co-simulation. In this thesis, the core functionalities of an effective and scalable synchronized co-simulation environment were designed and implemented. As an important part of the work, a novel solution for OPC UA based synchronization in continuous dynamic co-simulation is proposed. The evaluation conducted on the implementation confirms that both the synchronization solution and the OPC UA interface are suitable for being used in co-simulation of real-world systems.