The ATM (Accurate Temperature Measurement) project is concerned with determining the exact temperature of a medium at low immersion depths of a temperature sensor in the medium. The aim is to connect the sensor in the probe thermally to the medium as well as possible and to decouple it from the differently tempered environment. In addition to design modifications, plastics in particular can be used for this purpose, the properties of which can be manipulated by adding additives. For example, it has been possible for some time to produce plastics with good thermal conductivity. These can then be used to optimise the transport of thermal energy to the sensor. FEM simulations and optimisation algorithms are also increasingly being used, which can reduce the time and cost-intensive construction of prototypes and improve the understanding of physical processes.

The exact determination of temperature is becoming increasingly important in heat consumption measurement, among other things. Here, the heat energy extracted is determined by determining a temperature difference at a flow and a return in combination with a flow rate to be determined. To determine an exact temperature difference, the individual sensor does not necessarily have to measure exactly. If both sensors measure under the same installation conditions, for example, the installation-related deviations are largely compensated for when the difference is calculated. Due to their design, there is little space at the temperature sensor installation point of the return flow in standard heat meters. Here, the sensor is often immersed tangentially in the medium, which causes a non-negligible measuring error. This can be largely compensated for by using immersion sleeves at the second installation point, for example. In order to avoid this unsatisfactory solution in the future, the deviation of the determined temperature difference from the true temperature difference could be minimised by using directly and vertically immersed temperature sensors, which measure accurately even at low immersion depths.

This project is a joint venture between JUMO GmbH & Co KG, the Ilmenau University of Technology and the Fulda University of Applied Sciences.

Research processing
The project was completed on 10 April 2019 with the successful doctoral degree of Dr Andreas Brethauer .

Supervisor
Univ.-Prof. Dr.-Ing. habil. Thomas Fröhlich (TU Ilmenau)
Prof. Dr Elmar Engels

Control engineering comprises all technical tasks that influence a time-varying system from the outside in such a way that it follows a predefined sequence. From irons and ovens to motor vehicles: control engineering is part of our everyday lives. In the course of digitisation and the increasing automation of our everyday lives, control technology has become indispensable.

Prof Dr Steven Lambeck, who is in charge of the teaching area of measurement and control engineering, and his Control Engineering Team (CET for short) conduct research into applied methods and procedures for advanced control engineering in a wide range of subject areas. These include, for example, the predictive control of room temperature and humidity, the design of control algorithms for e-bike chargers, the expansion of control strategies with artificial intelligence procedures, the development of intelligent and adaptive robotics and much more.

The CET is constantly offering new and exciting student projects and theses. Students of all semesters with a thirst for knowledge and a thirst for action are invited to join the CET! Open project positions can be found on our ET Compass course.

At around 38%, the German building sector accounted for a very large share of total final energy consumption of 2,403 TWh in 2021. According to the "dena Building Report 2023", residential buildings accounted for 577 TWh and non-residential buildings for around 330 TWh, with space heating being the dominant final energy consumer in both sectors at around 75 %.

These figures motivate the development of energy-efficient indoor climate control strategies, which is why some well-known enterprises in the building automation industry and many experts from academia have been increasingly researching and developing the implementation of corresponding algorithms in recent years.

Ongoing doctoral degrees at the CET are focussing on similar projects, with an additional focus on controlling indoor humidity in order to not only save expensive energy but also protect valuable cultural assets from damage. The Fulda region has a number of old buildings used as museums and historical buildings, such as the Episcopal Seminary (see photo), St Michael's Church in Fulda and Schloss Fasanerie in Eichenzell, which fall within this requirement profile.

Contact person: Alessio Cavaterra

The market for e-bikes is in the "fast lane", according to the German Bicycle Industry Association (ZIV) in its press release of 15 March 2023. This circumstance motivates the development of lightweight and compact battery chargers for e-bikes.

As part of the "SCharger" research and development project, the CET is working together with Prof Dr Ulf Schwalbe on chargers based on the LLC resonant converter topology, which dispense with the DC link capacitor. This component takes up a lot of space and therefore offers the potential to enable small and compact chargers. However, dispensing with the DC link capacitor requires advanced battery charging strategies that not only reliably regulate the graded charging current, but also ensure a high active power factor.

Publications:

• Alessio Cavaterra, Steven Lambeck and Ulf Schwalbe: "Approximative
  modelling of an LLC resonant converter with Takagi-Sugeno models",
  at - Automatisierungstechnik, vol. 71, no. 10, 2023, pp. 853-866. https://doi.org/10.1515/auto-2023-0015
• Alessio Cavaterra, Martin Wattenberg, Ulf Schwalbe and Steven Lambeck: Software-based Power Factor Correction by using Iterative Learning Control for Battery Chargers with LLC Resonant Converter Topology. PCIM Europe. May 2022. DOI: 10.30420/56582219
• Cavaterra, Wattenberg, Schwalbe and Lambeck. Approximate Modelling of an LLC Resonant Converter with Takagi-Sugeno Models. Proceedings of the 31st Workshop Computational Intelligence, GMA Technical Committee 5.14 Computational Intelligence. Berlin, November 2021. DOI: 10.5445/ksp/1000138532 - awarded with the Young Author Award 2021 of the GMA Technical Committee 5.14 Computational Intelligence

Contact person: Alessio Cavaterra

The "Dezent" project

The "Dezent" research and development project aims to develop innovative and decentralised air humidification and dehumidification devices to improve the preventive conservation of cultural assets in museums and old museum buildings. Among other things, thermoelectric components and adaptive control algorithms are being used for this purpose.

Overview

Cultural artefacts exhibited in museums are permanently subject to climate-induced deterioration. Particularly extreme values and strong fluctuations in humidity often cause irreparable damage to valuable paintings, sculptures and documents. However, controlling humidity is a major challenge, especially in listed buildings. The engineers working with Prof Dr Steven Lambeck are now tackling this challenge. Together with Michael Kirner, an expert in museum air conditioning, and HKE Heinrich & Kloss Electronic GmbH, which is in charge of device construction, they are developing innovative decentralised air conditioning modules that are set to revolutionise museum air conditioning with the help of a wireless sensor network and intelligent control. By using modern semiconductor heat pumps without moving parts, this will be possible for the first time with both low noise and low maintenance. Museum operators and exhibition visitors can look forward to the preservation of cultural treasures and the unclouded enjoyment of art in the future.

Motivation

In museums and old museum buildings (over 6,000 in Germany alone), centralised air conditioning systems cannot be used in most cases. In the vast majority of cases, their installation is prohibited due to monument protection or is not feasible due to the high costs. This challenge is currently met with stand-alone air conditioning units, which only allow very imprecise control of the room climate. This results in fluctuations in relative humidity in particular, which can sometimes manage to cause irreparable damage to works of art. The project partners' idea for solving these challenges is the new development of climate modules that combine both air humidification and dehumidification in one device. Instead of the usual, usually noisy, compressor technology, thermoelectric cooling using so-called Peltier elements is to be used for dehumidification. These semiconductor components have no moving parts and are therefore noiseless and particularly low-maintenance. The very good controllability of the Peltier elements offers an additional advantage. In contrast to the compressor, their output can be adjusted continuously and very precisely. For the first time, this enables exact control and precise regulation of the air humidity. For their part, the new air conditioning modules use an adaptive control system developed by Fulda University of Applied Sciences (consortium leader).

Project participants

• Preventive Climate Management working group of the ET department (consortium leader)
• HKE Heinrich & Kloss Electronic GmbH(www.hke-gmbh.de)
• Michael Kirner Luftbe- und Entfeuchtung(www.kirner-shop.com)
• Städel Museum (www.staedelmuseum.de)
• Cultural Foundation of the House of Hesse (www.schloss-fasanerie.de/stiftung.html)

The "HumFlow" project

The temperature and humidity of a room are not only influenced by ventilation, heating and air conditioning systems, but also by the climatic conditions outside the room in question. The outside temperature, outside humidity, solar radiation and many other factors influence the indoor climate of the room through the outside walls of the building.

In the "HumFlow" project, the Control Engineering team is developing a minimally invasive measurement method that measures the surface temperatures and humidity of the exterior walls and estimates the temperature and humidity levels inside the walls. The project is divided into two fields of work, which will be explained shortly.

The first field of work deals with the development of an algorithm for estimating the temperature and moisture layers of an exterior wall. For this purpose, a non-linear system model of the exterior wall is used and analysed in a so-called state observer. In addition to measuring the current surface temperature and humidity of a wall, this approach also enables short-term predictions to be made on an hourly basis. The latter provide important information for energy-efficient room air conditioning or for building physics analyses.

The development of a cost-effective, space- and energy-saving hardware platform falls within the second field of work. The measurement method is based on two sensor units mounted opposite each other on an external wall. These can be attached to the wall with the simplest of means in such a way that the wall is hardly damaged (minimally invasive). The use of energy-saving radio technology should also make it easy to retrieve the recorded data.

Modern measuring devices are able to measure the moisture in a wall up to 80cm deep. However, these devices are expensive, cumbersome, difficult to transport and prone to measurement errors. The HumFlow project aims to counteract these disadvantages and make a contribution to the collection of important measurement data for everyone.

Measurements are continuously taken on a test stand with various model walls and different scenarios are tested.

Publications
Alessio Cavaterra, Andreas Böttcher and Steven Lambeck: The "HumFlow" Project - Developing a minimally invasive measurement system for estimating energy and humidity transfer processes through building walls. 13th REHVA World Congress CLIMA 2019 Bucharest Romania, 2019

The "Self-learning air hockey robotics" project

Teaching a robot to play air hockey is a complex and complicated, but by no means impossible, endeavour. With the help of algorithms from the field of machine learning, the Control Engineering team has succeeded in developing an air hockey robot that learns from the actions of its human opponent and, based on this experience, makes a self-employed decision on the best playing strategy at any given moment.

The functional principle of the robotics is based on the pre-programmed tactics of defence, attack, counter-attack and wait. With these modes and an overriding logic, the robot is already capable of competing against a human opponent, but if the game lasts long enough, a human opponent will quickly recognise the robot's weak points. The principle of reinforcement learning (RL) is used to make precisely this process of familiarisation accessible to the robot: in each scanning step, an action is chosen based on the current game state and the resulting game state is saved. If an action-state pair scores a goal, this action is rewarded with the help of a reward function. However, if an action-state pair manages to score a goal, the action is penalised. The algorithm takes the collected experience values in a table and averages them. The longer the robotics now plays, the more experience it collects and, theoretically, the better the resulting game strategy.

Another special feature of the self-learning air hockey robot is the implementation of the algorithm on an Arduino™ Mega 2560, which is only equipped with 8 kilobytes of RAM. An exact description of the "air hockey robot" system requires many process variables and is therefore a high-dimensional problem that requires a large amount of memory. Consequently, the problem had to be reduced. This is described in more detail in the publication below.

Publications
Maximilian Janßen, Alessio Cavaterra and Steven Lambeck: Self-learning air hockey robot - reinforcement learning in practice. Applied Automation Technology in Teaching and Development 16th AALE Conference Heilbronn, 2019

The research carried out as part of the doctoral degree is intended to develop concepts for the use of machine learning in engine diagnostics. The aim is to develop a strategy for transferable analysis models that enable the advantages of model-based data analysis without being tied to the engine systems. By applying the conceptualised models, the research results are to be used and evaluated in a prototypical manner related to practice. The focus of the component faults to be recognised is on bearing damage.

The sensor data available in modern industrial motors allows well-founded statements to be made about the current condition of the motor components. However, analysing the sensor data requires a high level of knowledge of the process as well as extensive general knowledge of the machine components. In addition, manually analysing the recorded sensor data is very time needed. This could be remedied by using model-based analysis methods through machine learning. However, machine learning-based status analyses require a certain amount of time and resources to train the analysis models - i.e. to adapt them to the problem at hand. However, this process is dependent on the specific motor types to which the model parameters have been adapted. It is therefore necessary to create transferable models, i.e. models that are decoupled from the actual engine system, in order to enable the business application of model-based analysis and engine diagnostics.

The project was carried out in the context of a doctoral procedure at Bosch Rexroth AG and supported by the departments of applied computer science and Electrical Engineering and Information Technology supervised by the HAW Fulda.

The project was completed in 2023 with the successful doctoral degree of Dr Tobias Wagner.

WAGNER, T., SOMMER, S.: Bearing fault detection using deep neural network and weighted ensemble learning for multiple motor phase current sources. In: International Conference on INnovations in Intelligent SysTems and Applications (INISTA), IEEE 2020, DOI: 10.1109/49547.2020.9194618194618

As part of this project, a decentralised and universally applicable, location-flexible, modular battery storage system (capacity ~1 MWh) with used traction batteries from electromobility is being developed. This intelligent second-life energy storage system (iSLE for short) transfers used traction batteries that are no longer suitable for electromobility into a second application scenario, the so-called second-life cycle. This increases their economic efficiency and significantly improves their ecological balance.
The iSLE system is planned as an adaptive and scalable battery storage solution for multiple applications, such as static voltage maintenance, feed-in buffering from renewable energy systems, power provision for high-power consumers such as charging parks or the provision of control power. Special attention is also being paid to the topic of modularisation of the storage capacity and the power electronic components for future "high-power charging" applications in the megawatt (MW) power range for electrically powered trucks.

An adaptive, universally applicable, location-variable and scalable fast-charging system for electric vehicles was developed on the basis of used batteries from the e-mobility sector. The prototype with a capacity of 180 kWh and a charging power of 150 kW is currently in research operation at the "Klinikum Fulda" location. The system design promises scalability up to 1 MWh capacity and up to 400 kW charging power.
The special feature of the energy storage system is the use of used traction batteries from e-vehicles, which are integrated into a second application scenario as part of a recycling process. The result is a sustainable solution that is characterised by intelligent energy management and innovative power electronics. This allows the system to be excellently integrated into electrical energy networks. This is necessary in order to favour a significant, flexible and efficient densification of charging points for the electrification of the transport sector with renewable energies, including in rural areas. The iWEnT system also provides valuable data on the ageing of batteries in real application scenarios and for the development of charging and utilisation forecasts.

The prototype of the multi-energy storage system passes through an electrolysis unit with a maximum electrical power consumption of 5 kW. This feeds a low-pressure hydrogen storage unit with a volume of 50 litres. A hydrogen compressor ensures compression up to 300 bar in a 600 litre high-pressure hydrogen storage tank (approx. 12 kg hydrogen), which corresponds to a chemical energy of approx. 400 kWh. The waste heat from the system is transferred to a 500 litre thermal buffer storage tank, which can store an additional energy of approx. 14 kWh at a temperature increase of 25 K. The built-in fuel cell with a maximum electrical output of 8.4 kW enables the stored hydrogen to be converted back into electricity. The waste heat is also transferred to the buffer storage tank.
The system was additionally equipped with a 26 kWh battery storage unit, which can be charged with surplus electricity from the grid, photovoltaics or via the fuel cell. With a maximum output of 15 kVA, the battery can supply the entire system and all connected consumers with energy in the event of a power failure. The built-in heat pump with a thermal output of 2 - 8 kW can also be used for additional cooling and heating. Heat energy can be extracted from the buffer storage tank via a heat exchanger to supply a local heating network or a detached house with heat.

The secure mobility of people and the reliable supply of energy are the cornerstones of stable social development and a high quality of life.
Alongside private transport, local public transport is an indispensable part of mobility and everyday life in the Fulda region in Germany. As a component of municipal services of general interest, it ensures spatial mobility and accessibility, which are key conditions for social participation, business exchange, employment and prosperity. The quality of life and urbanity of German cities therefore require attractive, efficient and powerful public transport: the challenge for future developments lies in the environmentally and climate-friendly transformation of public transport - sensible use of electromobility coupled with well thought-out measures to expand and support the electric energy network, in particular through the integration of energy storage systems.
The project focuses on research into technology for people.

The RTI (Real-Time Intelligence) project comprises the ability of automation systems to react to sensor data from industrial processes with actuator controls based on self-learned rules in order to achieve the respective processes with a given objective. The control intelligence must not manage any undesirable behaviour for components of the attachment or the products, even during the learning process. This can be achieved, for example, through conservative algorithms, a learning phase using historical data or a purely virtual learning phase. The focus here is on special approaches from the field of reinforcement learning and online learning.
Image sources: Bosch Rexroth AG

Growing digitisation in industry is managing to collect an increasing amount of actuator and sensor data that can support control and analysis functions. This creates new possibilities in intelligent automation, which are, however, associated with challenges. In addition, the increasing complexity of manufacturing processes makes it more and more difficult to understand how a process is influenced by physical conditions. This makes it particularly difficult to optimise process parameters using traditional methods. The project aims to use machine learning methods to enable so-called real-time intelligence to react to circumstances in the environment and thus actively influence industrial processes. Components of the attachment should not be exposed to increased risk as a result of this capability. Promising concepts are being implemented as examples in the control architectures of Bosch Rexroth AG.

The project was carried out in the context of a doctoral procedure at Bosch Rexroth AG and supported by the departments of applied computer science and Electrical Engineering and Information Technology supervised by the HAW Fulda.

The project was successfully completed in September 2022.

DE BLASI, S. | Active Learning Approach for Safe Process Parameter Tuning.
In: Machine Learning, Optimisation, and Data Science, pp. 689-699, Springer
International Publishing 2019, ISBN: 978-3-030-37599-7,
https://doi.org/10.1007/978-3-030-37599-7_57

DE BLASI, S., ENGELS, E. | Next generation control units simplifying industrial machine learning. In Proceedings to 2020 IEEE 29th International Symposium on Industrial Electronics (ISIE), 2020

DE BLASI, S. | Machine Learning for Industrial Process Optimisation.dissertation 2022

DE102019217859A1: Computer-implemented procedure and system for controlling a machine and/or an attachment

DE102020200173A1: Computer-implemented procedure and system for optimising a process parameterisation

DE102021209582A1: Procedure and control device for the automated optimisation of adjustable parameters of a controllable system for executing a process based on correspondingly set parameters