Skip to main content
Strategy
29/03/2019

Enable access to data

Artificial intelligence needs access to data
Dr. Jan Regtmeier
Dr. Jan Regtmeier
Director Product Management, Market, HARTING IT Software Development
Edge Cloud

With the "Artificial Intelligence" master plan of the Federal Government, Germany seeks to overcome the massive disparity vis-a-vis other countries. The goal is to make the "Artificial Intelligence made (AI) in Germany" seal of quality a recognised label. The collective term “AI” is understood to refer to computer systems that are capable of "intelligently" detecting problems and solving them on their own. Better access to data comprises key elements of the strategy.

In reality, the topic of "Artificial Intelligence" is nothing new. AI got its start in the 1930s with Alan Turing. Turing dealt with what constitutes an intelligent machine and when a machine can be termed “intelligent”. In 1950, the first algorithms were created as a basis for artificial intelligence. The first hype regarding "Artificial Intelligence" came in 1970. However, the hoped-for results failed to materialise, and enthusiasm quickly faded. The rapid development of chip technologies, capacity and performance that was seen after 2010 meant the necessary data could be made available in sufficient quantity and quickly processed with a high degree of complexity. In 2013, startup DeepMind built a new AI that taught itself to play Atari games, and a short time later a Deep Neural Network from Microsoft achieved similar image recognition error rates as a human being. Today, the topic of "AI" is really taking off, and the "Artificial Intelligence" strategy that was jointly developed by BMWi, BMBF and BMAS is intended to ensure that Germany is cemented in place and developed as a research and industrial location and achieve the same goals for the competitiveness of the German economy. Better access to data will form the basis for this. The MICA creates the necessary foundation and is already able to implement learned neural networks. Thanks to its data collection and pre-processing capabilities, the HARTING MICA® as an edge device becomes an enabler of artificial intelligence.

Digital retrofit

In order to specifically drive forward the topic of Industry 4.0, HARTING's customers and partners see great potential in the "digital retrofit" of existing systems and machines. Many systems are depreciated, technically fully functional, but provide no interfaces for networking and recording relevant parameters and data. Simple retrofit solutions are needed here. The modular industrial-grade mini-computer MICA® (Modular Industry Computing Architecture) offers a wide range of interfaces, from Modbus, RFID to IO-Link in the future. In line with the accepted Industry 4.0 maturity model, the MICA® helps to create connectivity and visibility.

But data collection alone produces no added value and no return on investment (ROI). Meaningful conclusions and actions must be derived from the data. Above all, as the volume of data grows, there are two issues that need to be overcome. For one, it is impossible to collect all data. Therefore, one of the core features of an edge device is the pre-filtering and reduction of data. Here, people decide which parameters and data will be collected. In addition, it is difficult to decide what data are actually relevant. How do I know what is relevant and how do I deal with large amounts of data? This is where the Cloud comes in, as it provides storage space and tools for Analysis.

Need for a hybrid Edge-Cloud architecture

Thus the need arises for a hybrid Edge-Cloud architecture. Data is captured and pre-processed on the edge, for example the HARTING MICA®. Only the relevant data is then sent to the Cloud for analysis. If one still has no idea at all which data are relevant, for a limited period of time a large amount of data can be sent to the Cloud and analysed there to answer the question of “which data are actually relevant?”

AI really helps here. "Machine learning" makes it possible to analyse very large amounts of data and to determine which parameters are relevant for the respective question at hand. Following this analysis, the mini-computers can be reconfigured to send only this "Smart Data" to the Cloud, as opposed to "Big Data".

Big Data and Smart Data are terms that are hotly debated in public. But what exactly do they mean, and wherein lies the difference? Big Data refers to enormous amounts of data that cannot be analysed or processed using existing methods. Smart Data, on the other hand, goes beyond this notion, as this simple formula shows:

Smart Data = Big Data + Benefits + Semantics + Data Quality + Security + Privacy = Useful, High-Quality and Secure Data.

So Big Data is the raw material that needs to be prepared so that it can be refined into Smart Data. Once again, a hybrid Edge-Cloud architecture with the MICA® can help.

“Machine learning” can in turn now be applied to the relevant data in the Cloud. For example, neural networks can be learned in the Cloud. The learned neural networks are then so compact that they run on the MICA® in the Edge. As a result, decisions can be made directly on site on the machine. This is especially interesting if there is no 24x7 access to the Cloud or short reaction times are required. In addition, it saves costs.

At the same time, more data in the Cloud enables the neural network to continue to be improved. If there is significant improvement here, the latest version of the neural network is brought back onto the MICA®.

Pragmatic Approach

We recommend a gradual and pragmatic approach. The most important question is about the benefit: which problem is relevant and needs to be solved? After that, the MICA® architecture offers a future-proof solution. Existing machines can be networked, data collected and aggregated. Frequently, initial insights are already created via visualisation and acquisition. In the next step, the data can be sent to the Cloud. The neural networks learned here can then be diverted to the MICA® and executed there. The result is a powerful architecture in which AI and its tools help answer business-related questions and create Solutions.

Artificial Intelligence (AI)

Basically, there is no common definition of the term "artificial intelligence" because it is very context-dependent. In general, however, it can be defined as follows: "Artificial intelligence denotes machines that operate based on algorithms, perform tasks and thereby react autonomously and adaptively to unknown situations. This means they demonstrate similar behaviour as humans: they not only perform repetitive tasks automatically, but can also learn from success and failure and enhance their behaviour in ways that resemble human creativity."

Machine Learning

Machine learning is a sub-area of artificial intelligence. It enables IT systems to identify patterns and laws and develop solutions, based on existing databases and algorithms. The insights gained from the data can be generalised and used for new solutions or for the analysis of previously unknown data. In order for the software to learn independently and to find solutions, rules must be set up for the analysis of the data base and the recognition of patterns. So humans have to lay the groundwork in advance. For example, the systems must be supplied with the data and algorithms relevant for learning. Once the appropriate data is available and rules are defined, the systems can then begin with machine learning and e.g. make predictions based on the analysed data, calculate probabilities for particular events, adapt to developments independently, and optimise processes based on recognised Patterns.

Recommend article

Eingeschränktes HTML

  • Allowed HTML tags: <a href hreflang> <em> <strong> <cite> <blockquote cite> <code> <ul type> <ol start type> <li> <dl> <dt> <dd> <h2 id> <h3 id> <h4 id> <h5 id> <h6 id>
  • Lines and paragraphs break automatically.
  • Web page addresses and email addresses turn into links automatically.