The concept of Industry 4.0 has broadened the horizon of Business Intelligence.
New sources of data, which have never been taken into consideration up until now, allow us to obtain more information and bring value to the analytics system. New approaches and uses have also appeared, such as real-time monitoring and predictive maintenance, expanding the possibilities of analytics.
The evolution in the industrial model towards a digitalized environment has brought the rise of new actors in the world of analytics. One of the fundamental bases that sustain this evolution is the Internet of Things (IoT). The availability of increasingly cheaper sensors, the increase of communications infrastructures, the lowering price of storage systems, as well as the easy scalability of these solutions, have allowed the rapid implementation of this kind of systems within industrial environments - adding new layers of knowledge to information systems.
Still unfamiliar with Industry 4.0? head over to this blog that explains the basic concepts of the fourth industrial revolution.
In the current environment, the introduction of these new levels of knowledge allows us to gain new competitive advantage, provided that we are able to give a meaning to all of this information; it’s not about which devices we equip with sensors, nor which data we collect - it’s about what we want to obtain with this information. The success of these kind of projects depends on the power of the tools (integrating analysis and prediction), the speed of deployment, the capability to adapt to any need and the savings obtained.
Data processing and storage
An important element behind the Industry 4.0 boom is the ability to analyze the collected information immediately, enabling real-time decision making (edge data integration). The ability to distinguish between a normal or abnormal behavior implicates savings, both in costs and time, that we must take into account, and this new analytical scenario has changed the way of working at an operational level. For analysts as well, this means an evolution, shifting from working on a database to working with a data stream, from using information coming from the business activities to information collected from environments that are heavily influenced by hardware.
However, we’re not just talking about edge analytics, but also about predictive analysis. The ability to deploy models on the machine itself or on the hardware related to a sensor can enable full control on the future behavior of the production chain, preventing quality deviations, avoiding unplanned stops due to mechanical incidents or allowing the schedule of automated maintenance actions.
Another factor that comes into play in Industry 4.0 projects is data processing. The main challenges here are to distinguish between relevant and redundant data, define whether we need to keep the information or all possible meaning has already been extracted, and define the level of aggregation. We also need to consider the use we intend to do with the information, especially if we are talking about real-time analytics or predictive analysis projects, and the storage model, taking into account the multiple options of Cloud storage available today - although, in this case, we’ll need to overcome the fear of losing control over the involved data.
In this sense we always need to keep in mind legal restrictions (for instance, GDPR) and constraints on communications, that require us to grant the storage of data also in case of connection problems, ensuring a proper bandwidth to allow a fluid data streaming.
Once data are received and stored, we need to exploit the acquired information. There are multiple options and the capacity of analytical representation of information is practically unlimited - from dashboards for the real-time monitoring and alerts management through decision-making tools, to the presentation of the historical information and evolution of a sensor or of alert indicators.
This point is where one of the most prominent barriers usually appears in this kind of projects, since you need to know in depth the power of representation of each platform, as well as their ability to process, modify and represent data at the right speed. Given the multitude of available options, it’s crucial to analyze the technical constraints of each platform, the advantages they can bring by and the intrinsic limitations of the project (on-premise or cloud solution, refresh rate, etc.).
In any Industry 4.0 project, one of the key factors is the predictive analysis capabilities that can be integrated through the application of models on the edge layer, described above, expert systems, basic algorithms or supporting machine learning. Depending on the project needs, we may need a simple temporal projection, a trend analysis or, in more extreme cases, the application of neural network or deep learning algorithms. Here is where the knowledge of different platforms and the experience in project management comes to light.
We need to take into account that the information needed for constructing a model will be available after the set up of the sensor system. A correct programming of the different project phases is therefore necessary, to allow the acquisition of a test set of data and their validation according to the model that we need to develop.
The data scientist's ability to reflect business needs in robust predictive models, and with a high degree of success, will be key. The current trend is to assign the data scientist a double role: on the one hand, s/he’s the responsible for abstracting the business needs (citizen data scientist); on the other hand, a technical profile able to choose the best algorithm for each need and parameterize it in the right way.
There are a multitude of solutions that enable the development of Industry 4.0 projects, whatever their type is. The current trend is offering cloud-based tools for analytics, such as SAP Analytics Cloud, Microsoft Power BI and Qlik Sense Cloud. These platforms have very contained costs, do not require any infrastructure and are very flexible in terms of licensing.
In addition to this, they also allow a multi-platform access (web browser, mobile, tablet), offer a modern and pleasant look & feel and are truly user-friendly.
On the other hand, it is also important to take into account modular cloud services (Amazon Web Services, Microsoft Azure, Google Cloud), which allow to adapt to any specific case by activating the necessary modules and paying only for what we use. Moreover, we must consider specific tools for predictive maintenance, such as SAP Predictive Maintenance and Service or SAS Asset Performance Analytics.