Online platform for vibration analysis #Post2

Dear Openadptronik blog readers,

I am back with a new blog post on the topic ‘Online Platform for Vibration Analysis’. On my last post I mentioned that I am conducting interviews to get a better idea on how the dataset could be modelled and how a necessary data models could be created in order to store/query measurement data efficiently. During my interviewing process, I came to realize that a big part of my thesis work would involve requirement engineering. As easy or difficult as it may sound, the goal of that task is to essentially define and describe the What, How and Whys of the system, hence, it is a process of defining and describing customer requirements (What), system requirements (How) and the business requirements (Why). Ultimately, the outcome of those interviews should be able to answer one main question ‘What should the future system do?’ Based on that, I have to come up with a user interface and eventually a system architecture, that means, the ‘How’ part of the system.   

To get a better understanding how I can proceed, I started asking some simple questions to some of the very qualified mechanical engineers. First one being, ‘Imagine that there is already an existing online platform for vibration analysis with all the necessary data, you land on that platform and you have search field, what would you type on that search?’

I got various answers from different people and they are all so interesting. Based on that I am starting to form a structure for the platform as well as for the dataset. However, most of the search query came as a full sentence answer. Unfortunately, that falls short on the feasibility test, as this thesis work is only for 6 months, natural language processing won’t be a part of that schedule. However, the best part was to coming up with categorization of the problems from the search queries. I think, for now, we could settle for Keyword Search, such as ‘Car Noise’.

I will combine all the interviews I am performing and will write on my next post the main categories that came out of those interviews.

Alongside requirement engineering, I am looking into the uff58 file format. More specifically, I am looking into the pyuff library. The goal is to able to let users upload uff58 files to the platform, since this is the most general and most complete data format for vibration data.

I am also looking into technicalities of the data storage options. To be more specific, I am looking through different research papers that have done some tests using relational vs non-relational database by storing and querying sensor data.

Data analysis using machine learning is only meaningful when the data amount is huge and therefore the main reason of  the platform is to allow any user to upload their measurement data. The field of engineering is looking for ways to go for predictive maintenance, solution of vibration related problems by combining already solved problems. This platform would be one of its firsts to allow any user to upload their measured data from anywhere as well as provide them with the opportunity to use or add services that can be used to analyse the data and eventually store those results. Storing all these vibration data and analysed data can be used to make prediction for similar problem and probably come up with the best solution for such problems. But before all that can happen we need to think about Statistical classification. I know, it is a big word, but the main concept is to classify or categorize problems. In order to create sensible training data we need to first come up with the understanding of which data will belong to which data sets. Therefore, the categories that are coming out as the part of the interview process are going to be the very basic building blocks of categorization of vibration problems.

If you are aware of vibration problems and want to contribute, then feel free to write on the comment what do you think of as a search query given a platform which has all kind of vibration related problems and solutions.

Till my next post!