Möchtest du unsere Inhalte auf Deutsch sehen?

x
x

AI Facilitates Detection of Leaks in Compressed Air Systems

IoT Use Case - Agile IT Management (AIM) + Emerson
4 minutes Reading time
4 minutes Reading time

Determining the real cost of compressed air is an inseparable part of pneumatic applications. And closely related to this is determining the costs caused by leaks. One thing is certain: if air escapes where it shouldn’t, it can become expensive due to damage to the product, reduced performance with increasing energy consumption, and increased CO2 emissions. It is therefore in the interest of both providers and users to identify and address unwanted air loss. For Emerson, this is accomplished with an AI-powered predictive maintenance solution from AIM.

The challenge: Locating leaks in pneumatic systems in a targeted manner and improving the energy balance

While the machine manufacturer themselves have access to the control system, this is not necessarily the case for the user. Therefore, there is a desire to make the individual components smarter so that they can directly feed their information into IoT analysis models. Since pneumatic components do not inherently provide data, specially equipped cylinders and valves are needed for automatic predictive maintenance. Energy consumption and energy efficiency are playing an increasingly important role in manufacturing companies, both due to rising energy costs and the potential tax benefits and funding opportunities associated with energy efficiency and CO2 reduction. This interest is also extending to leak detection in pneumatic systems. It is still common today for leaks to be accepted as a necessary evil. One, albeit labor-intensive, step further are companies that manually check their pneumatic systems at regular intervals – for example, every six months. An automatic solution must work with the limited data available. The interplay between component expertise and industrially applicable AI is a crucial success factor.

The ideal scenario is to locate leaks early without extensive additional sensors, using the existing signals from the flow sensor and the control system. Classical machine learning and AI models learn from a wide variety of data, but here we are dealing with the other end of the spectrum. Even with little data, it should be possible to continuously monitor a plant and identify the right time to intervene.

The solution: Targeted selection of data and automation of predictive maintenance

Before any successful implementation, regardless of the industry or technology used, it is necessary to understand the system. Only then will the characteristics of the signals become clear and the data provided can be sensibly sorted. This creates a data model that maps the plant structure. This is where AIM, with its in-depth knowledge of AI logic and systematics, supports component providers such as Emerson. In doing so, AIM builds closely on the data Emerson already captures. One key aspect is that instead of extracting the required information from an almost overwhelming amount of data that can be easily captured in any system with the corresponding number of sensors, Emerson adopts the opposite strategy. Capturing the important data – and only that – for monitoring with as few additional sensors as possible. In the case of pneumatics, monitoring is synonymous with leak detection. Through AI, the disruption noticed in a system can be specifically attributed to individual components. This shortens troubleshooting, regardless of whether it takes place in a regular maintenance cycle or on the basis of a fault message.

The result: Optimizing maintenance means reducing costs and emissions

The greatest effects can be achieved with specific solutions. The basic principle of obtaining results that ideally significantly improve everyday production life with as few but the right data and manageable effort remains the same everywhere. For the best results in each case, a comprehensive understanding of the technology’s specificities is necessary, in this case, pneumatics. In reality, there is neither the pure machine learning scenario, in which a training case is generated from fault messages, nor the counterpart, in which all necessary threshold values are stored purely on the basis of operating data.

AIM and Emerson jointly implement a realistic solution based on the current state. The smart approach is to immediately respond to detected leaks to save duplicate localization efforts. Whether immediate repairs or planned maintenance are carried out depends on the extent of the damage. This can be calculated by entering the corresponding costs in the indicators for repair measures. Savings in energy costs at the shopfloor level can be as high as 50%. Added to this are the advantages of increased plant availability and better planning of maintenance intervals.

Get our IoT Use Case Update now

Get exclusive monthly insights into our use cases, activities and news from the network - Register now for free.