Industrial Data Intelligence: On-Site Streamed Insight

How much potential resides in your production line?
We can help you find it out,

Production analytics methods and technologies can optimize your production. Do you have a specific use-case? We happily carry out a proof of concept and evaluate its technical and commercial feasibility.

Should you need support with defining and assessing a specific use-case, we can organize an on-site industrial data analytics workshop at which we collectively determine the commercial value of the data at hand, define project objectives and requirements and derive the specific task and approach.

After a successful proof of concept we can help you gather and analyze production data in an industrial data analytics project. The goal would be optimizing your own or your customer’s production.

Our dataTHINK industrial analytics solution processes your data in real-time, in your plant, and enables plug and play analytics through OPC UA, with the goal of optimizing your production.


Your Industrial Analytics Partner: We Equip You for the 21ˢᵗ Century

For specific use-cases in which data represent a commercial value, we develop an individual production analytics concept for data-based production optimization. At first production data are gathered and analyzed with the goal to recognize patterns as a basis for improving the Overall Equipment Efficiency (OEE). In addition to technical architecture, security concept, data handling analysis, visualization and integration, we enable measuring success relative to the initially set goals, with our dataTHINK solutions.

The Industrial Data Intelligence group is an interdisciplinary team at Softing Industrial that deals exclusively with data-based production optimization. We build on 35 years of experience with industrial data communication and data acquisition.

If you are a production company or a plant or machine manufacturer, we can be your industrial data analytics service provider and partner with the goal of improving your or your customer’s production.


Project Management according to CRISP-DM

Our approach running industrial data analytics projects is based on the Cross-Industry Standard Process for Data Mining (CRISP-DM) process model, which was developed since 1996 within the scope of a European Union R&D project and which has become the cross-industry standard for data mining and analytics projects, also for production analytics by means of our dataTHINK solutions.

CRISP-DM divides between 6 phases, amongst which one switches backwards and forwards oftentimes:

Business Understanding: The project goals and requirements as well as the resulting specific task and approach are determined from the perspective of the business goals.

Data Understanding: The available data are gathered and examined, possible data quality problems investigated and first hypotheses for embodied information formulated. 

Data Preparation: The final dataset is constructed in multiple consecutive selection, transformation and cleaning activities.

Modelling: Appropriate modelling procedures are applied, parameters optimized. Depending on format requirements of specific algorithms it may be necessary to go one step back to the Data Preparation phase.

Evaluation: The model which best fulfills project goals and requirements from phase one is selected.

Deployment: Depending on project requirements, deployment may vary from a one-off preparation and presentation to a permanent integration in the customer’s decision process.


Real-Time Industrial Data Acquisition

In our production analytics projects we gather data – raw or processed – from automation components and field devices like PLC’s, sensors, actors and databases as well as additional sources like for example production flow or weather data by means of our dataTHINK solutions. Next, data are processed in an industrial Extract-Transform-Load (ETL) process. Outliers are deleted, faulty entries eliminated, timestamps aligned, metadata added and the clean-up data formatted. This process takes – as far as possible – place in real-time. Depending on the specific task data are processed in a second industrial ETL process or transferred in some kind of persistent storage (files, NoSQL, SQL, cloud, cluster) for later industrial data analytics. For acquiring the data either Softing products (e.g. dataFEED OPC Suite, echocollect, OPC server and OPC UA technology) or third party products are used.


On-Site Stream Production Analytics: Real-Time Processing in the Production Line

Basically we divide between two kinds of approaches for industrial data analytics:

One-off approach: a problem is analysed and understood and can be solved by a production modification. In this case the problem is solved by means of engineering.

Permanent approach: A problem cannot be solved permanently, but shall be recognized predictively. In this case a solution is looked for by means of a dataTHINK streaming analytics implementation (e.g. anomaly detection).

Independent of the approach data analytics is foremost about producing a model. This represents relevant patterns, relationships and regularities from the situation at hand and is possibly extended by facts from given circumstances and tested on its capabilities. Next the implementation with real life data takes place with the goal of predicting a specific - normally unwanted - situation (anomaly detection).


Industrial Analytics Model Assessment on the basis of Objectives / Requirements

In the implementation phase the industrial data analytics results are examined by use-case domain experts. Only the production engineering experts can judge if encountered mathematical correlations actually exist. Furthermore the production analytics results are validated with help of additional independent samples in order to guarantee a high level of reliability. In case ambiguities or contradictions occur after bringing in expert knowledge and further samples, the data analytics shall be repeated or the parameters altered. Does a high consensus exist between industrial data analytics results and expert opinion, also business, technical, organizational and legal constraints shall be regarded in preparation of the implementation phase. These highly determine the implementation feasibility of the analytics results, by means of our dataTHINK solutions.


Industrial Analytics Model Integration with dataTHINK Portfolio

After a decision regarding the production analytics results implementation was made one of two routes may be chosen:

Conversion in the design process: The problem was analyzed and understood and can be solved by means of a production modification.

Permanent Industrial Data Intelligence implementation: The problem or the task can only be solved by means of a permanent approach und shall be recognized predictively. By means of a dataTHINK streaming analytics implementation the permanent solution, e.g. anomaly detection is installed and the production optimized.