Data & analytics in the digital era

Traditionally data has been gathered on the PLC or an Industrial PC. Today in the era of IIoT [Industrial Internet of Things] this scope is extended upward to the Cloud and downward to individual Smart devices. As Industry 4.0 progresses it is driving the standardisation of data formats and communication protocols making the transition faster and easier. The aim is to use the data to guide or make decisions related to the business KPIs such as increased quality, reduced energy consumption, OEE etc.

Connecting data to man & machine

Central to Industry 4.0 is connectivity – the link between human and machine, and machine-to-machine.

The first building block, as shown in figure 1, is the use of smart products, which gather data about their operation and performance. Smart products already exist with the capability of connecting and networking. For example, Festo’s CPX valve terminals have on-board intelligence and utilise the Industry 4.0 standard OPC-UA communication protocol to enable access to the data generated and stored on-board. Connecting a series of CPX valve terminals to, for example a Festo CPX-IOT Gateway enables manufacturers to gather the required data, aggregating, channelling and filtering it as required. This data then can be accessed for Industry 4.0 services either in the cloud or On-premise, so manufacturers can utilise the data.

Cloud hosting services enable data to be viewed securely and globally by manufacturers across the internet, transferred via APIs or exported.  On-premise devices are deployed to reduce the amount of data that is transferred to the cloud, saving data transmission and storage costs and fulfilling security requirements by further restricting access to confidential information. They are installed in the plant, with manufacturers or service providers storing and utilising the data locally. It remains contained within the confines of the factory and where permitted  limited information is allowed to be pushed or pulled to the cloud. Both have their advantages and disadvantages. Premise services are preferred by some users as more secure, although that is not always the case. However, even if data is not exported off-site it is frequently drawn down from the cloud to provide insight and added value. If data needs to be compared from different production lines and in different countries, then it makes sense to share this information via the cloud.

Once relevant data has been collected it can then be used in a hierarchy of services of increasing complexity and value such as data visualisation and condition monitoring – providing alerts when thresholds are met, to preventative and predictive maintenance. Data can be viewed either on pre-configured dashboards or custom generated dashboards depending on the application requirements. Pre-configured dashboards make it very quick and easy to access information on standard products, whereas generated dashboards can use standard templates and proprietary packages such as Node Red to create decision trees and determine notification outcomes tuned to the applications and viewer’s requirements. Progressing from this first step, Festo has demonstrated how manufacturers can then move forward to using machine learning and AI to gain greater insights from the data.

The ability to get greater value out of connectivity is also demonstrated in Figure 1. When manufacturers store data in the cloud, they have global access, knowing the status of their assets (including configuration, hardware, firmware and software versions), utilisation, how they’ve performed historically and are currently performing. This knowledge enables the manufacturers to operate more effectively: for example to speed up commissioning, increase overall equipment effectiveness (OEE) and save energy. One of the industry use-cases is energy monitoring as it provides a quick, proven payback and ROI. If you know your energy consumption at individual plant, production line and machine level, you can then take relevant steps to control, manage and reduce it.

Industry 4.0 is also defining a descriptor for products and systems called the Administration Shell, this is their standardised Digital Twin. It can be auto-recognised, within a machine-to-machine environment just as we expect with USB linked devices in the PC world.  This makes, and will make it increasingly easier to understand data, moving it from mysterious numbers stored in numerous registers into information that relates to the application, the process and the business.

Figure 1. Steps to realising the full benefits of Industry 4.0

Delivering data insights with AI

AI is also accelerating and being enhanced through the implementation of Industry 4.0. But what is AI in the context of industrial automation? AI can be defined as the concept of improvement and gaining insights through smart analytics and modelling, it is a collective term that incorporates the following steps:

• Artificial Intelligence – summarises all technologies and methods
• Machine Learning – recognises patterns and rules of existing data
• Deep Learning – machine improves independently
• Reinforcement Learning – machine uses a system of reward and punishment to make own decisions
• Transfer Learning – knowledge acquired completing a task contributes to the solution of a related problem

Fig 2. Where AI takes place

AI can take place in all three locations shown in figure 2: in the cloud, where large quantities of data can be evaluated easily, on-premise, which is on the system at production network level, or on-edge, which is on the component at field level.

Machinery will be increasingly autonomous and will use AI algorithms to organise cooperation among themselves, sharing data with the supply / delivery chains and with users: creating ad hoc networks as the need arises. The data produced from the manufacturing process is analysed and actioned through AI to create dynamic self-learning production environments that are able to provide increasingly higher levels of productivity, operating with higher quality in a safer working environment.

Some are concerned that AI will dramatically decrease or even eliminate the need for human interaction on the factory floor. However, AI frequently requires human intervention, to define the model and objective, and refine the output. AI usually provides a probability scored proposal that is rewarded or punished my the human-in-the-loop feedback. AI gradually improves the accuracy of the probability score based on the human feedback, refining its algorithm model.  

Conclusion

Creating business improvements within the industrial automation environment through data insights, is in common usage today. However, it will become dramatically easier with the progression of digitalisation and standards embedded within Industry 4.0.  Artificial Intelligence will increasingly play a part, providing causality insights, neural networks that self-learn and even documenting their internally developed decision making processes.  Without a doubt we are in an exciting era for data insights within the factory automation and machinery application field.

www.festo.co.uk/i40