Research projects

Project director: Prof. Dr.-Ing. Peter Zeiler

Project worker: Fabian Mauthe

Number of industrial cooperation partners: 3

Brief description of the research project:

The research project aims to enable users, especially SMEs, to realise the latest maintenance strategies, primarily predictive maintenance. To this end, it will develop a method of assessment and selection which takes account of application-specific and industry-related constraints. This categorisation will form the basis for the generation of sets of real degradation data, which will then be made freely available. This helps users with suitable projects to select appropriate methods, and also to assess the implementability and the risks. The method employed here can be data driven, e.g. an artificial neuronal network or Support Vector Machines. The prognostic method can also be model based and use a physical model of the degradation process, e.g. a crack growth equation or a wear model. The possibilities described are illustrated below

Most of the PHM research being undertaken at present relates not to selection and assessment, but to the development or further improvement of prognostic methods or their exemplary application to an individual problem. It is essential that users in particular have a methodology available to assess and select suitable methods as they start to implement predictive maintenance. The research project aims to close the gaps in the current state of the research and the state of the art, and thus play a significant role in making predictive maintenance more widespread. The actual innovation consists particularly in linking up basic research results obtained during the development of prognostic methods with the actual practical demands of industry. Esslingen University of Applied Sciences is working with three different-sized industrial partners on this research project. The project is funded by the State of Baden-Württemberg.

Project director: Prof. Dr.-Ing. Peter Zeiler

Project worker: Simon Hagmeyer

Brief description of the research project:

Extension of data-driven diagnostic and prognostic methods used in Prognostics and Health Management to hybrid methods through the incorporation of knowledge about the degradation process

This research project explores approaches which can be used in the diagnosis and prognosis of the degradation state of technical systems. The various approaches here can be divided into model-based, hybrid and data-driven approaches. In the model-based methods, the degradation behaviour is simulated in a physical model. A widely used example of this physical modelling are crack progression models such as the Paris-Erdogan equation. They describe the fatigue cracking under cyclic load. However, the degradation process of many technical systems is highly complex, and hence a great deal of effort is required for the detailed implementation of the model-based approach. In contrast, the methods known as data-driven diagnostic and prognostic methods have the advantage that no or very little knowledge of the particular degradation behaviour is required to use them. These methods are categorised under the basic fields of statistics and machine learning. They simply replicate the mathematical relationship between the variables measured and the degradation. As the name suggests, the hybrid approach represents a mixture of the two aforementioned approaches.

The research project is looking at how knowledge of the physics of the degradation process taking place can be used to improve data-driven PHM applications. When sufficient training data is available, data-driven methods such as artificial neuronal networks can help achieve very good performance. However, one of the major drawbacks of PHM is that generating service life data for technical systems is generally very expensive and time consuming. An efficient alternative is to incorporate existing knowledge about the degradation process. This begins with individual physical laws which the data-driven model must obey as boundary conditions. On the other hand, when detailed knowledge is available, its incorporation also includes the construction of physical models and merging them with data-driven models to give a hybrid ensemble model. The research project thus investigates an approach which supplements the Similar Systems research project to solve the fundamental problem of the low availability of data with PHM.

Project director: Prof. Dr.-Ing. Peter Zeiler

Project worker: Marcel Braig

Number of industrial cooperation partners: 3

Brief description of the research project:

The diagnosis, prognosis and monitoring of the state of health and the remaining useful life of technical systems requires methods which can simulate the system behaviour, possible failure states and the degradation which occurs. These methods are usually categorised into those based on physical models, and data-driven methods. The former are based on a detailed understanding of the system being monitored and its degradation behaviour. The modelling is very involved, however, and can only be used to a limited extent for complex systems. Data-driven methods are thus an important alternative. They extract information about a system on the basis of data alone. The success of data-based approaches depends crucially on there being a sufficiently large quantity of training data, however. The problem is that generating and processing this data is very time consuming and expensive. Moreover, it is often the case that very little data on run time and run-to-failure is available for novel systems (e. g. a new generation of machines) and systems which are produced only in low numbers.

The research project adopts the approach of using information (data or models) from Similar Systems to train data-driven methods. The aim is to reduce the quantity of data required from the system under investigation itself. For example, knowledge of the degradation behaviour of an open deep groove ball bearing can be used for a closed version with similar dimensions. The approaches can be applied to individual components (e.g. bearings, axles, shafts), to machines (e. g. in production) or to products (e. g. gears, engines).

Project director: Prof. Dr.-Ing. Peter Zeiler

Project worker: Fabian Mauthe

Collaboration partner: University of Stuttgart, Prof. Dr.-Ing. Marco Huber

Brief description of the research project:

The use of data-driven methods for condition diagnosis and prognosis for technical systems has a broad spread in Prognostics and Health Management (PHM). These methods are based on statistical modeling of interrelations within the training data provided. Therefore, they are not suitable for extrapolation to domains without training data. Furthermore, their prediction may take implausible values that are not consistent with governing physical laws and other constraints.

In PHM, these drawbacks are of considerable relevance due to the high cost and time involved in generating comprehensive training data. To mitigate these drawbacks, there are approaches in machine learning that address the incorporation of various forms of knowledge about the system under consideration. These are referred to, among other things, as Theory-Guided Data Science (TGDS).

Although it is usually not possible to model degradation processes with sufficient physical detail, basic knowledge about the system under consideration and the inherent regularities of its degradation process is usually available. Therefore, the research project aims to reduce the drawbacks of data-driven methods in PHM by integrating knowledge on regularities occurring across applications. Within the project, the regularities that frequently apply in PHM are first identified and their scope analyzed. Subsequently, a selection of the most relevant regularities is made. The main focus of the project is to investigate how these regularities can be used in the context of TGDS to improve data-driven methods and how the improvement evolves in relation to the amount of available data. The research results are validated using experimental data from two mutually heterogeneous PHM use cases.

In PHM, various studies exist on combining data-driven and physical models using hybrid methods. The present project, however, considers the integration of cross-application regularities occurring in PHM, which are not sufficient for a physical modeling of the degradation process. This sets it significantly apart from the state of research. On the one hand, the methods of TGDS have not yet been specifically addressed in PHM. On the other hand, the literature on PHM so far lacks any cross-application consideration of the subject.

In this research project, the Esslingen University of Applied Sciences is cooperating with the Institute of Industrial Manufacturing and Management (IFF) at the University of Stuttgart. The German Research Foundation (DFG) funds the project (project number 514247199).

Project director: Prof. Dr.-Ing. Peter Zeiler

Project worker: Moritz Müllersch?n, Luca Steinmann

Cooperation partners: Cross-faculty project at Esslingen University of Applied Sciences

Brief description of the research project:

A Smart Factory Grid is the vision for an innovative, service-based and dynamically distributed manufacturing system. It is characterized by high flexibility and networking and represents a central component of Industry 4.0. Modular production units, so-called micro manufacturing units, interact within a smart factory grid. The aim of the System Health research unit is to develop methods for the intelligent diagnosis and prediction of the status of micro manufacturing units and the smart factory grid, as well as the optimization of maintenance processes.

Research question 1: Adaptive condition diagnosis and prognosis under variable loads (Luca Steinmann)

Due to its high adaptability, flexibility and resilience, matrix production places high demands on condition diagnosis and prognosis. Varying and unknown load profiles make it difficult to develop suitable methods. Initial approaches exist for forecasting future load profiles based on historical and current data, but condition forecasting under variable conditions remains a key challenge. Limited data quality, incomplete or censored data and uncertainty about future loads add to the problem.

The research approach aims to better understand and precisely predict the relationships between stress and degradation. One focus is on analyzing and integrating varying load profiles and overcoming data-related challenges. Approaches such as data augmentation and data generation are intended to expand the database and increase the robustness of the models in order to develop a holistic methodology for prediction considering variable load.

Research question 2: Adaptive condition diagnosis and prognosis with heterogeneous information (Moritz Müllersch?n)

The development of a smart factory grid poses specific challenges. In particular, the heterogeneity of information sources poses a central problem. The data is often highly dimensional, formatted differently and varies in terms of availability and quality. Examples of this are sensor and operating data, simulation results and physical models. This makes it considerably more difficult to combine and use this data for condition diagnosis and forecasting. Another obstacle is the lack of a comprehensive methodology for information fusion.

One research approach is to develop a hybrid model for information fusion that combines data-driven methods with physical and rule-based models. Methods are to be developed to compensate for the limited availability of data and to effectively integrate existing domain knowledge. By developing a standardized data framework for pre-processing and integrating heterogeneous information sources, the basis for a robust condition diagnosis and prognosis can be created.

Research question 3: Health management in a smart factory grid (Moritz Müllersch?n)

Health management deals with the continuous monitoring, analysis and optimization of the condition of production plants and processes and aims to ensure the availability and reliability of the system. In the context of the flexible and modular environment of a smart factory grid, this requires a high degree of adaptability.

The research approach aims to address health management in a smart factory grid as a complex optimization problem, as the complexity of production no longer allows health management using conventional methods. Important metrics here are costs, sustainability, productivity and many more. Predictive maintenance is intended to enable the automated generation of maintenance proposals in order to make precise and efficient maintenance decisions.

Further information on the entire Smart Factory Grids | DFG Research Impulse project can be found here:

Smart Factory Grids | DFG Research Impulse

(Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) Project-ID 528745080 – FIP 68)

Current doctoral projects

Extension of data-driven diagnostic and prognostic methods used in Prognostics and Health Management to hybrid methods through the incorporation of knowledge about the degradation process

This doctoral project in the field of Prognostics and Health Management is being undertaken as a collaboration between the Institute for Technical Reliability and Prognostics (IZP) and the Institute of Industrial Manufacturing and Management (IFF) at the University of Stuttgart. The doctoral student is a member of the PROMISE 4.0 Collaborative Doctoral Programme. More information on this doctoral project can be found in the description of the DATA-DRIVEN TO HYBRID research project.

Collaborating university: University of Stuttgart
First supervisor: Prof. Dr.-Ing. Thomas Bauernhansl
Second supervisor: Prof. Dr.-Ing. Peter Zeiler
Doctoral student: Simon Hagmeyer

Selection and implementation methodology for prognostic methods to predict the remaining useful life of a technical system taking account of practically relevant and application-specific constraints

This is a doctoral project in the field of Prognostics and Health Managements (PHM) and is being undertaken as a collaboration between the Institute for Technical Reliability and Prognostics (IZP) at Esslingen University of Applied Sciences and the Institute of Industrial Manufacturing and Management (IFF) at the University of Stuttgart.

A variety of possible prognostic methods are available to predict the remaining useful life of a technical system. They can be divided into data-driven, model-based and hybrid approaches.

The aim of this research project is to investigate a selection and implementation methodology which takes account of practically relevant and application-specific data. The focus is particularly on linking fundamental research results with the actual practical requirements of industry when developing prognostic methods.

Collaborating university: University of Stuttgart
First supervisor: Prof. Dr.-Ing. Marco Huber
Second supervisor: Prof. Dr.-Ing. Peter Zeiler
Doctoral student: Fabian Mauthe

Use of knowledge of similar systems for data-driven condition diagnosis and prognosis in industrial environments

This doctoral project in the field of Prognostics and Health Management is being undertaken as a collaboration between the Institute for Technical Reliability and Prognostics (IZP) and the University of Stuttgart. More information on this doctoral project can be found in the description of the REDAT research project.

Collaborating university: University of Stuttgart
First supervisor: Prof. Dr.-Ing. Peter Zeiler
Second supervisor: Prof. Dr.-Ing. Marco Huber
Doctoral student: Marcel Braig