Our solutions have been designed for operational staff in order to facilitate their access to information to improve plant performance.
As a result, they provide business solutions to users to meet their field uses. Some functionalities use techniques based on Machine Learning to meet data analysis needs by avoiding exposing the complexity of these techniques to users.
However, our solutions make it easier for professionals in charge of data processing (data scientists, data analysts, data engineers, etc.) to facilitate the task of professionals in charge of data processing (data scientists, data analysts, data engineers, etc.) who can thus more easily make the fruits of their work available to operational staff.
The observation we make is simple: the language that is emerging to work on data processing or Machine Learning is undeniably Python. Thanks to its data manipulation and cleaning functions, its very complete libraries and its versatility, Python has established itself on the market compared to languages like R, which have historically been widely used by data scientists.
It is a free language that benefits from the production of the international scientific community. In addition, many commercial Data Science solutions rely on this language (Dataiku, Rapideminer...).
With the Process Data Lake From OIAnalytics, we continuously collect, store, structure and process factory data. It is therefore possible to build a data set very quickly with data from multiple sources, already combined and with the desired aggregation (per time step, per production batch, etc.) according to the need for analysis or modeling.
In this way, we save the teams in charge a very important amount of time by eliminating the often underestimated complexity of this stage. The Process Data Lake has many export options (CSV, XLS files) or data query (Rest API).
Another challenge for model teams is their implementation in the field. It is necessary to be able to have the necessary data continuously, to be able to execute the models at the desired rate and to return the information resulting from the execution of the models to operational staff. This can become a headache and generate significant delays and costs.
To remedy this, we have integrated into our Process Data Lake the possibility of running models built in Python by taking advantage of the entire data collection, structuring and processing layer. The information from the models is then available like any other data and thus usable in all the functionalities of our solutions (visualization, reporting, operational control, analysis, etc.).
Optimistik takes care of the infrastructure, performance, availability and security of data and the execution of models, allowing teams to stay focused on their business.
Our solutions thus make it possible to streamline collaboration between operational staff and data scientists who have data expertise with the common objective of improving plant performance.
They are therefore positioned as a complementary approach to the data-science tools on the market.
Author: Mathieu Cura