Motivation:

In the new software generation STIWA is moving towards event stream processing based on Apache Kafka and Apache Flink. These event streams contain valuable information which is very helpful for the operation of the STIWA software as well as for the operation of the connected production lines. In particular, temporal sequence patterns in the event stream are to be monitored in "real time" in order to detect problems at an early stage. An example would be the detection of x defective parts in a series produced on a production line.

Tasks:

• Technology Evaluation: Compare Apache Flink CEP with alternatives based on a set of criteria.
• Design: Concept development for a platform-based solution that can be used to implement both standardized and user-defined rules for anomaly detection.
• Implementation: Platform solution and implementation of initial detection algorithms using Apache Kafka, Apache Flink CEP, and a graphical rules DSL such as Drools.

Targets:

• Develop an executable platform solution for event stream anomaly detection
• Implement initial detection algorithms that address real-world and common sequence patterns
• Validate the viability of the platform to generate initial customer value

Motivation:

The state of the art is to be assisted by chat AI in solving tasks. In combination, technical specialists without deep knowledge of the details of the programming language and APIs can quickly achieve remarkable results.
The STIWA production data analysis tool is used to display production data. The raw data is transformed into analysis views (cubes). Various facts, e.g. KPIs and dimensions, are provided in these cubes, which can then be displayed using the flexible analysis front-end. The semantics of the data available in the standard (raw data, KPIs) are well defined. If the information provided is not sufficient, new customer or project specific cubes can be created from the existing raw data and cubes. For this purpose, the product provides a programming interface that can be used to prepare more complex data in Java or Python (Compute).

Tasks:

• For this purpose, an existing open source model will be used that has the ability to generate Python code for general problems.
• The semantics of the analysis API will also be trained on this model to support the generation of application code.

Targets:

• Support the creator of the compute scripts:
  • Formulation of corresponding algorithms in Python/Java
  • Usage of external APIs and in particular usage of the API to access existing raw and aggregated data in the STIWA product.

Motivation:

It is often difficult to find the right information in large amounts of product documentation. In many cases, the information itself is available, but it is difficult or impossible to find the relevant articles using traditional search engine technologies. Furthermore, the information you are looking for is often spread over several systems (e.g. support ticket system, FAQ, etc.). In addition to the STIWA product documentation, other sources of information such as our support tickets and FAQs should help you to find the right information quickly.

Tasks:

• Evaluation of the right technologies for the implementation of such a chatbot, taking into account framework conditions such as data confidentiality, etc.
• Creation/implementation of a prototype
• Evaluation the results with questions such as "How good were the answers?", "Were answers found?", " Which questions did not have a meaningful answer?"

Targets:

• The first step is to create a chatbot that finds the right articles in our documentation systems based on specific questions.
• A possible second step is an automatic system analysis (e.g. based on log files) of our software products and the automatic derivation of recommendations for action.

Motivation:

The state of the art is to be assisted by chat AI in solving tasks. This is also the case in software development. The question is how and in which areas of software development such systems can usefully assist us.

Tasks:

• Consideration of the different AI systems (ChatGPT,...) taking into account the framework conditions such as confidentiality of data, etc.
• Consider/evaluate the potential of AI in the following areas of software development (using concrete practical examples):
  • Writing code
  • Testing
  • Debugging
  • Code optimization
  • Uncovering security issues
  • Code analysis and translation
• Figuring out how to set the tasks for AI systems in different areas of software development to get the best result

Target:

• Concrete practical examples show how and in which areas of software development AI systems can be used efficiently.

Motivation:

Apache Kafka is agnostic about the content of the messages being sent. Therefore, Kafka can be used to send messages with any serialization format. Currently, STIWA uses the text-based JSON as serialization format. Binary protocols such as MessagePack, Avro, ... seem to have some advantages over text-based JSON, such as small data volumes, faster serialization/deserialization, general resource saving. For this reason, JSON will be compared in detail with binary protocols in the context of Apache Kafka and Apache Flink.

Tasks:

• Comparison of different serialization formats based on a list of criteria: theoretical consideration of the alternatives, practical consideration based on test setups
• Concept development for the test setups (comparison of alternatives under realistic conditions, Apache Kafka and Apache Flink as important system components, impact of the serialization formats on these technologies in terms of programming, performance, data volume, efficiency and robustness from analysis of the current development process).

Targets:

• The aim of the work is to have a solid basis for the decision of a serialization format. The selected serialization format will be used as a standard at STIWA for many years, which is why this detailed comparison is necessary.

Motivation:

A main focus of STIWA Software is the modeling of products, processes and resources (PPR) in a data model for the production of complex, mechatronic components with many variants and short cycle times (e.g. non-trivial components for the automotive industry). Since modern components are becoming more and more complex and consist of a large number of components, which can also differ depending on the design variant and are connected via different process steps, a suitable description language (DSL = Domain Specific Language) is required. In this context, a tool based on GoJS is used, in which the assembly sequence is modeled and mapped, and the objects can be reused in other tools.

Tasks:

• The modeling of the process sequence (Bill of Process) will be presented on the basis of specific requirements. A rudimentary description of the underlying DSL is available and will be made accessible in this thesis.
• Based on existing objects in a model (e.g. the products to be manufactured), sequences (linear, parallel, ...) can be defined with respect to the processes.
• The connection to various systems such as the plant modeling tool and the equipment list should also be considered.

Target:

• Development of a domain specific application as a proof of concept for the modeling of the Bill of Process (BoP). This application is based on the GoJS software framework. Using different real use cases, the BoP can be modeled, edited and modified in this tool.

Initial situation
As Innovation Department of STIWA Advanced Products, we are responsible for the first product development phase, from the idea to the functional prototype. One of the topics, that are processed in the team, is the technology of magnetorheological media, i.e. media whose viscosity can be adjusted by means of magnetic field strength. With this technology, several product solutions are currently implemented and produced on site. At the same time, this range is to be supplemented by further products in this technology segment.

Requirements

• Completed Bachelor’s degree in the field of Mechatronics or Technical Physics
• Interest in physical basics
• Knowledge in the draft of smaller circuits and programming of microcontroller
• Interest in the conception of new solutions

Example for a task

• Draft the HMI control element on the basis of magnetorheological media
• Conception of mechanical components
• Draft electronics and software necessary for control
• Assembling, testing and optimization of the developed solution

Motivation:

STIWA Austria manufactures approximately 50 million milling and turning parts, from individual components/small series to large-scale and mass production. The machine park includes about 100 CNC processing machines which are operated with approximately 10,000 different tools. Tools as well as offset data must be available on the machines during setup. This big quantity of offset data must be transferred to the machines in a manageable, organized and reliable manner.

Tasks:

• Research work for:
  • Which technologies are available (Balluf, datamatrix, RFID, Excel, etc.)
  • Advantages/disadvantages of the technologies, availability, costs/risks, scalability, benefits, data safety/data quality
• At what company size/number of spindles/number of tool inserts is which technology profitable?

Objectives:

• Evaluation and presentation of a solution for subsequent changeover of the entire machinery
• System proposal for new installation of a new manufacturing site (new building onto the green field)
• Worldwide operation in the STIWA Group Business area Manufacturing

Motivation

Ultra-wideband (UWB) technology is one of the most promising techniques for indoor localization offering high accuracy, low power consumption, and resilience against interference. However, there is a lack of research on how it can be integrated into challenging industrial environments. 
In this thesis, UWB localization shall be used to locate and identify workpiece carriers within a transport system. A localization system based on custom sensor nodes shall be adapted in terms of firmware (e.g. communication protocol) and set up at a partner’s facility. Mounting positions for both static nodes on the transport system and moving nodes on the workpiece carrier have to be defined in order to minimize non-line-of-sight conditions.
The localization performance shall be assessed with a focus on the correct detection of the workpiece carrier order. At discrete positions, the transport system is capable of providing location information, which can be used as reference. The main goal is to evaluate the localization accuracy that can be expected from an UWB system in harsh environments.

Tasks

• Literature research suitable commercial UWB systems
• Adapt and set up a localization system at the company partner’s facility
• Acquire measurements on a real transport system
• Refine localization algorithm e.g. by assessing additional sensory
• Evaluate the system in terms of localization accuracy

Contact and supervision

Univ.-Prof. Dr. Andreas Springer (MT 0351, andreas.springer@jku.at)
DI Philipp Peterseil (MT 0347, philipp.peterseil@jku.at)

Download Details

Motivation
STIWA Austria produces around 50 million milling and turning parts, ranging from individual components/small series to large-scale and mass production. Among others, optical inspection technologies are used for quality assurance and documentation. In these optical processes, the surface of the components is scanned using stripe light projection. The reflections of the projection generate a volume model that can be compared with the 3-D model. It is therefore necessary to measure and compare components always with the same surface properties which aresurface pre-treated. This pre-treatment is currently carried out manually. The aim is to achieve an automated, consistent pre-treatment process.

Tasks

• Evaluate mecessary component cleaning concepts for KSM, oil, chipping etc.
• Conception of an automated surface preparation for optical geometry measurement
• Creation of an automation concept for automatic parts loading and further transport including positioning in the measuring machine
• Ensure component marking and data processing
• Research and compilation of safety-related instructions/standards/regulations
• Concept development for different components with
  batch size 1

Objectives

• Concept of an automated, component-independent surface preparation process
• Automated transfer of the components to the measuring machine

Motivation

Our production with high-performance, assembly, laser welding, punching systems, etc. only takes place in exceptional cases using flexible manpower. In order to be able to produce effectively, the complete production team for each system (number, qualifications) must be available. The tension between adherence to the production plan/specifications and dynamic changes (loss of availability) must be compensated for by means of a shift plan. On a day-to-day basis, this represents a high daily capacity commitment of employees with planning tasks. The existing shift planning software was programmed in-house a few years ago and contains / takes into account a large part of the relevant framework conditions. 
 

Extract from framework conditions

• Plants with different role distribution and different products (different number of employees and qualifications)
• Different shift models
• Operation of the systems only possible with full staffing
• Some systems with multi-machine operation
• Interface to the ERP system
• Mapping of employees who are undergoing training and are not yet allowed to produce on their own
• Shift plan (employee view) should provide a communication interface
• Detailed planning horizon approx. 4-6 weeks
• Mapping of the skills matrix

 
Project objective

• Comparison and selection of possible shift planning software
• Testing of this software, taking into account all relevant framework conditions
• Preparation / presentation of the potential benefits and savings of the selected software compared to the previously used software (including amortization)