Awareness of the value of data has become widespread in various markets in recent years. While working with data in the financial sector (stock market, fraud detection) is now a common discipline, the use of data in the industrial sector, specifically in production, is still in its infancy. In most cases, it is limited to KPIs such as scrap, quality, performance and other metrics that are consumed by the higher-level MES or ERP system.

Figure 1: Changing the sequence from "Algorithms -> Data -> Decisions" to "Data -> Algorithms -> Decisions".

© Softing

Only recently has interest expanded towards the use of machine data at production level for optimizing processes, understanding influencing factors as well as influences and interactions of system variables or finding the causes of machine failures. This growing interest is being driven by global trends such as Industry 4.0 and IIoT. These developments explain the automatic and efficient use of data, which is continuously produced at every step of the manufacturing process in order to pass on knowledge to systems further up the automation pyramid or back to the production process - with the aim of improving, automating and individualizing or monitoring the entire production chain with all its machines.

Alongside land, capital and labor, data is increasingly becoming a production factor. It enables cost savings and new business models. The increasing dominance of data and the resulting change in the sequence from "algorithms -> data -> decisions" to "data -> algorithms -> decisions"(Figure 1) is the basis of the revolution that is currently taking place. Since the first programmable chip - the Intel 4004, which came onto the market in 1971 - companies have followed the same pattern when developing their software: They define the problem, determine the goals to be achieved and specify the necessary steps. Finally, they write the application as a sequence of algorithms. In practice, data is fed into the algorithms and users make decisions on the basis of this data. This procedure is currently changing structurally: in the first step, the data is collected and analyzed in the second step using generally valid algorithms. On the basis of the resulting causalities, a person makes decisions on production optimization today; tomorrow, this will be done by algorithms.

Machine learning in production

Figure 2: A fault is detected using algorithms.

© Softing

Without machine learning algorithms, important parts of our business world or consumer world would already be at a standstill. Algorithms decide within milliseconds whether a customer will be paid out money at a cash machine. They recognize the faces of our friends on social media or let us set tasks on our smartphones and will soon be responsible for autonomous driving. The same algorithms tell us how and where we can optimize our production, as long as we provide them with the relevant data. Production optimization is about predictively identifying a malfunction based on data from a functioning machine or system(see Figure 2).

As early as 1959, the American computer scientist and computer pioneer Arthur Samuel defined 'machine learning' as a field of study that "gives computers the ability to learn without being explicitly programmed to do so beforehand". Unlike classic applications, machine learning does not take its solution directly from the software code written by humans. The essence of machine learning is that a pattern exists which we cannot capture mathematically, but which can be found by algorithms based on data. You can specify solution categories to the algorithms and instruct the algorithm to decide which category should represent future data (supervised learning). Alternatively, it is left to the algorithm to find patterns or clusters that were previously unknown to humans (unsupervised learning).

On site instead of in the cloud

The volume, speed and variety of data generated today exceeds the capabilities of the operating personnel and calls for new, data-based approaches. Predictive maintenance aims to reduce the majority of non-age-related failures and thus increase system performance. Machine learning algorithms predict the failure of specific system components. This enables needs-based maintenance of specific parts at non-production times before a failure occurs. In the classic sequence of "algorithms -> data -> decisions", the overall equipment effectiveness (OEE) cannot be better than the human who programmed it. Machine learning algorithms applied to large amounts of production data, on the other hand, can find causalities that improve the OEE and were previously hidden from the plant operator.

Many decision-makers have a queasy feeling about the idea of putting their production data outside into the cloud. Alternatively, the security issue can be minimized with the help of an edge solution - i.e. in the plant or on the machine where the data is generated - on a standard IPC. The goal of improving availability, performance and quality across the board is nothing new. What is new is the data-based approach using machine learning algorithms - if desired in the plant.

Automated work of the data scientist

Many companies and start-ups have recently developed platforms and analytics tools to implement analytics for industrial data. However, there is still a gap between data acquisition and the tools and platforms for data storage and processing. This gap is usually bridged by a professional data scientist who uses tools to extract knowledge from the data and develop IIoT business value.

While the majority of IIoT projects are still characterized by a more manual and exploratory data science approach, we envision a system that bridges the gap between the sea of data and data analytics tools by automating most of the steps that a data scientist performs.

This area of research has recently emerged as 'Autonomous Analytics' and is closely related to the field of Guided Analytics. Both disciplines attempt to minimize the interaction and necessity of involving a data scientist. In the case of Guided Analytics, the execution of data analysis is initiated but otherwise automated by a domain expert, whereas Autonomous Analytics is about the complete automation of the entire data analysis process, from input to presentation of results.

Author:
Peter Seeberg is Business Development Manager Industrial Data Intelligence at Softing.