Adobe Stock © 胜 张
In the first step, it is necessary to make a clear distinction between highly automated and truly autonomous systems. Because even if a high degree of automation can already be achieved by automated systems, this is still basically a form of time-based predefined remote control. If a sensor reports a predefined value back to the system, an actor will perform a suitable but equally predefined action. Even highly automated systems are ultimately always limited to this "if, then ..." logic. Autonomous production is also coming into focus due to the fact that automated systems today have in part become so large and complex that it is becoming impossible to anticipate all relevant parameters and dependencies in advance. Compared to automation, autonomous systems go a decisive step further: Here, unforeseen events should also be detected and evaluated. The system should ultimately be able to make a sensible decision on its own. In this way, self-learning systems are created that can make previously time-consuming and cost-intensive changes without the intervention of engineers or computer scientists.
The development of such autonomous production systems is linked to a number of core issues and resulting prerequisites, which are briefly presented in the following paragraphs.
stockunlimited.com © 1289243
Artificial intelligence (AI) is an important core technology for developing self-learning systems. In the context of machine learning, for example, the aim is to gain artificial knowledge from existing experiences. To this end, neural networks are trained to solve specific tasks using sample data and to make autonomous decisions in the process. In industrial production, corresponding systems are already used in process monitoring, process control and preventive maintenance.
stockunlimited.com © 1434348
In order for a system based on AI and machine learning to make autonomous decisions, a sufficient amount of data must be provided to the system. The acquisition of sensor data, as well as highly developed image processing, is therefore of particular importance in production operations. The more high-quality data or data supplemented by structured additional information from production operations is available for further processing, the better neural networks can be trained and subsequently make good decisions.
From order entry, through manufacturing, to delivery of a product, all systems involved must be networked with each other so that information from technical and commercial systems is available for the self-learning process. Equally, production decisions can only be made in real time if the robots and machines involved in the manufacturing process can be controlled in their own decentralized computing capacities.
If you are wondering how autonomous your own production system already is, the Wissenschaftliche Gesellschaft für Produktionstechnik e.V. (Scientific Society for Production Engineering) can provide some initial assistance.
The association of leading German professors of production technology has developed a model - based on the stage model of autonomous driving - with which the maturity of production systems can be assessed as a first approximation.
The production line is adjusted exclusively by machine operators to all occurring disturbances.
Controlled drives reduce inaccuracies of the production line.
The production parameters are adjusted by machine operators to the given product specifications.
Individual selected process parameters are controlled by the production plant itself according to defined specifications. The remaining production parameters are adjusted to the given product specifications by machine operators.
The production plant regulates individual selected product properties independently. Necessary adjustments to changed production conditions are made automatically.
All relevant product properties are controlled independently by the production system. It can eliminate all deviations from defined properties and monitor its own system limits.
The production plant controls all relevant product properties, recognizes and corrects explicitly and implicitly specified errors and deviations. It is able to recognize and extend its own limits.
This question is difficult to answer concretely and unambiguously from today's standpoint. Because one thing is certain: the varieties and individual designs of the smart factory will be as varied as the business cases of the companies that set out to implement the concept of autonomous production.
However, it is also clear that the PLM system will often take on a central role in all the versatility of the emerging solutions. Today, a large part of the important product data is already stored in the PLM system in the form of a digital product file. As part of the implementation of the "digital twin", further data, such as sensor and imaging data, as well as data from the real-time operation of the products, will have to be integrated into this structure in the coming steps in order to make the data available in a structured manner for the training of AI applications.
The implementation of autonomous manufacturing will not least depend on the seamless interaction of PLM, ERP and MES. With regard to the PLM system, this point also results in requirements for flexible solutions for the integration of these and other systems. In addition to standard integrations, the most important thing here will be flexible interfaces that also enable individual solutions.
And finally, the fact that a number of intermediate steps will be required on the way to autonomous manufacturing should not be ignored. In this process, the PLM system must have the ability to automate even complex processes and workflows. In these intermediate steps, it will prove to be an indispensable advantage if processes can be designed on the basis of an established standard, such as BPMN 2.0, and implemented in the system.
The road to autonomous production therefore remains exciting - not least in the environment of the PLM system.
Get an overview of the important components already available today for developing your PLM strategy for the road to the Smart Factory. You can already start your first PLM homework today...
With keytech PLM you already have the possibility to combine all documents and data relevant for your product in a digital product file. Find out more about the possibilities of a future-proof PLM system...
With the keytech api we already provide a REST web service interface with which versatile integrations to other systems based on modern web standards can be developed and implemented in the company.
With the keytech workflow engine we already provide the possibility to design processes based on the BPMN 2.0 standard and to execute them easily and without much programming effort in keytech and to automate process steps.