TU München - Fakultät für
Es sprechen Studenten über ihre abgeschlossenen Diplomarbeiten und Systementwicklungsprojekte.
Am Montag, 19.12.16, ab 15:00 Uhr, im Raum „John von Neumann“ (00.11.038):
Hazards from High System Entropy: An Explorative Analysis of Case Report
Classical industries such as aviation and automotive are facing another industrial evolution. Software controllers are now widely deployed in safety critical control systems all over the world. As time shifts, system requirements may change. Systems become more and more complex as the demand of change raises. Therefore, there is a theory, that the distortion (such as architectural degradation) - which is called System Entropy - of such closed system never reduces over time. In this work, we did a qualitative and explorative research on case reports, which is capable of analyzing how the system entropy and complexity in safety-critical control system affects system safety. We developed heuristics and hypothesis which may answer the question: Does and how does high system entropy affect system safety throughout the system life cycle and safety engineering process?
Precise Method-Level Defect Localization through Analysis of Version Histories
Information about past defects in a software system has several important uses in software engineering. In research, defect data is used to develop and evaluate defect prediction models. In practice, it can be used to guide testing efforts, since areas that were especially error-prone in the past are more likely to contain errors in the future. The development of relative defect densities in a system can also serve as an indicator of quality or process problems. Unfortunately, accurate defect information is typically unavailable. In principle, version control systems store the entire evolution history of a software system, including its bug-fixes. In practice, however, bug-fix information is buried in the sequence of all changes, including extensions, refactorings and adaptations that do not concern defects. This lack of accurate defect data makes empirical validation of defect prediction models hard and effectively renders them useless in practice. The goal of this thesis is to create and validate an approach that can produce accurate defect information from the version history of both open source and industrial software systems. It comprises the implementation of a dedicated program analysis and its empirial evaluation.
Advanced experiment data collection and analysis for the MaCon approach
In modern manufacturing engineering software tools are widely used on different stages of the engineering process. The tools influence the engineering process and have an impact on the overall efficiency. As the quality and efficiency of the engineering processes have a strong impact on the economic success of production systems [Sch+13], engineering process analysis and evaluation is an important issue. Different methodologies for engineering process analysis have been developed. Some of them focus on evaluation of engineering tools, some on consistent and lossless migration of information and engineering artifacts along the engineering activity and tool chain, some on engineering networks crossing company borders, and others have a strong information oriented focus [Sch+13]. Anyway software is an important factor in the engineering approach analysis, as it influences its speed, cost, quality and efficiency. This thesis focuses on tool-based engineering approaches and evaluation of them on the example of the MaCon approach and the tool it offers. In this work we offer to evaluate the engineering approach by holding an experiment, where the approach and the tool would be used, capture the experiment and collect additional data like participants' feedback and perform the analysis and evaluation of the approach based of the information retrieved from the experiment. This work offers a methodology for preparing and conducting the experiment as well as a tool for further analysis of the collected data. We offer to integrate data collection into the engineering tool in order to capture the connection between inner software events and user feedback and collect the feedback according to the inner states of the system. In this thesis we design and develop a tool for the experiment data analysis, which is focused on the collected feedback and media files. The experiment that is described in the thesis shows an example of integrated data collection usage in a number of user studies and further analysis of the received data in the Analysis Tool. The approach is evaluated on 16 participants.
Current Challenges of System Safety Practitioners: Qualitative Analysis of On-line Discussions
This thesis aims to analyze various types of on-line discussions in order to find out the current challenges facing safety practitioners. Data sources of various types such as mailing lists, group discussions, research websites and forums will be examined and relevant data will be collected from them. The data will be collected, stored and analyzed and later techniques of qualitative content analysis and grounded theory will be applied to the data in order to develop theories and identify research questions that address the challenges currently facing safety practitioners. In addition, an interview will be conducted with a safety practitioner from the industry in order to validate the results of the research, take their opinions on certain identified issues and their recommendations for solving some of the problems being faced.