Today, the Industrial Internet of Things produces data non-stop for machine and plant manufacturers of all sizes. Public cloud offers are often used to evaluate them with the help of artificial intelligence. Because AI approaches require data as a learning volume, which is naturally more significant in a public cloud solution than on systems installed on-premises, you can also access out-of-the-box solutions in the cloud. Nevertheless, many companies hesitate to take this step. There are two important reasons for this, of a legal and technical nature.
On the one hand, the problem in many business processes is that data transfer to the cloud for governance reasons is prohibited. Because there is a high degree of intellectual property in the IIoT data, i.e., secret, valuable knowledge about products and production processes that you do not want or are not allowed to publish or share. If one thinks, for example, of the production of armaments or parts and workpieces on behalf of a customer, the protective nature of such data becomes obvious. Another problem is the increasing volume. If a machine today generates several thousand data points in the millisecond range, real-time processing in the cloud is hardly possible for latency reasons.
Scalable Architectures On The Edge-Cloud Continuum
One solution is edge computing, combining the advantages of local and cloud computing as required. Edge computing can be understood as a preparation for the cloud – an on-premises solution at the Edge with functions for analytics, pre-processing and distributed intelligence, which, together with the cloud, forms a kind of continuum. Near Edge refers to edge computing close to the cloud, far Edge close to the machine. This distinction allows for scalable architectures in the edge-cloud continuum if necessary. IoT data can be aggregated and anonymized at the Edge for use in the cloud—that’s the legal side.
The prerequisites are connectors between the Edge and cloud for data transmission and data models that integrate data on both sides’ support. As far as the technical advantages are concerned, thanks to new and scalable technologies, even high data volumes can be processed at the Edge without latency problems. This is what makes real-time analysis possible. A warning that the machine could break down in a few minutes due to exceeding critical limits comes at the right time, as does an immediate production stop, as part of data-based quality control that checks each workpiece immediately after it has been manufactured. In this way, reject quantities are reduced, and material and energy costs are saved at the same time.
In such cases, edge computing is unavoidable, with large amounts of data to be analyzed and answers required quickly. AI is not necessary for solely processing the data with aggregation, anonymization, and analysis. In the Data in Motion edge solution approach developed by X-Integrate, machine sensor data can be recorded, pre-processed, and analyzed in almost real-time. On the other hand, with the data-at-rest approach, additional raw data or already pre-processed data for process documentation for downstream or complex analyzes in connection with data collected elsewhere are kept on a long-term basis. The latter also served as input for unsupervised machine learning for pattern recognition and supervised learning to derive predictions for unknown data.
Now, algorithms allow reinforcement learning at the Edge and contribute to model optimization through interaction with the environment. Largely isolated edge environments benefit from this, which must also function independently of the cloud in the long term and still function reliably. X-Integrate edge solutions are scalable; Producer components for data acquisition, analytics components, and consumer components for exporting the data are available in containers and can thus be put together application-specifically and precisely for the individual use case. An alternative to the IoT frameworks of well-known providers was created based on open-source applications, which enables SMEs, in particular, to quickly and inexpensively enter the IoT world and, thus, new digital business models.
Advantages Of The Edge Over Pure Cloud Computing:
- Security, privacy, and data sovereignty for the owner of the data.
- Large amounts of data are processed promptly and directly at the point of origin.
- Reduction of data volumes in the direction of private/public cloud.
- Independence from an available cloud, e.g., B. for environments in which systems should work largely independently.
- Optional use of AI or ML and thus inference at the Edge.
- Trained models from a central cloud are applied at the Edge.
- Optimization of publicly available models through edge-trained models.
- Mesh networking approaches: Meshed edges learn from each other.
New Business Models Through Data Analysis At The Edge
Of course, data is not collected for its own sake. Instead, their analysis should serve as a basis for digital business models. The system manufacturer supplies machines with integrated sensors, which the customer, the operator of the machine, uses to his advantage: Through the data analysis, he achieves cost advantages that he can either keep for himself or pass on to his customers – i.e., those who produce with these machines purchase items from him. Competitive advantage for the machine operator: He can optimize production costs and offer workpieces at lower prices than the competition. This increases the competitiveness of all stakeholders involved in the value-creation process.
They are providing or expanding machines with integrated edge devices and edge applications that are pre-installed or expandable, standardized, and specifically adapted analysis tools that can enable quality assurance to be tracked during production. Transparent pay-per-use invoicing models are only possible if sensors measure wear and tear and utilization, and the machine consumption is calculated and invoiced based on this. It is irrelevant whether the machine is located at the plant manufacturer’s site and produced directly for his customer, the machine operator, or at his plant. In the same way, as for maintenance, the manufacturer then bills the provision as a variable cost component based on usage, analogous to mileage leasing for motor vehicles.
An example: It is measured whether the machine is used in 1-, 2- or 3-shift operation or how high its load is. This can be done, for example, by measuring the vibrations during operation. So if you handle it with care, you save maintenance costs and improve the machine’s value retention. The system is optimally utilized in the same breath by temporarily renting out overcapacities. Many other business scenarios are conceivable, for example, in building management: By integrating data from room booking systems with sensor data from the individual floors and rooms consolidated at the Edge, the heat output on individual floors and rooms can be controlled in a targeted manner and also optimized for the entire building -the. Pilot projects promise savings of around 25 percent in annual heating costs.
Data In Motion
This approach is also referred to as data in motion: The data is not stored somewhere and later analyzed as with data at rest but processed in almost real-time immediately after its generation, which makes monitoring or alerting possible in the first place. This makes sense insofar as many data are only of short-term interest for processing. Users can therefore select precisely what is deleted after the analysis and what is kept permanently because they will need it later – in this case, the result is stored in a database or a data lake, depending on the application in the company or the cloud.
Integration Of Shop And Office Floor
Only a tiny part of the shop floor data is then moved to the cloud to integrate it with office floor data from ERP / CRM systems, for example. This works, for example, via manufacturer-neutral open-source frameworks such as the Open Integration Hub of the Cologne Cloud Ecosystem, e. V. includes technology, standard data models, rules, and a community of connectors. As part of the OIH plus project funded by the Federal Ministry of Economics and Climate Protection, the framework was expanded to include components for industrial interoperability. This enables SMEs to digitize their system processes and IoT data in a simplified manner and integrate them with data from the office floor.