Machines and other infrastructure components often have a long service life that extends far beyond the tax depreciation period. This is especially true for buildings and facilities for energy supply. These “Legacy Devices” have so far been largely excluded from the Internet of Things, although applications such as predictive maintenance are of great importance, especially in old age. The reason is that digital technologies were not yet available at the time of production and changes to the control systems would often be expensive or would require a new approval of the system. However, the rapid development of sensor technology and electronics has led to high performance and low cost of IoT edge devices. Current gadgets are both cheap and capable of executing complex algorithms for signal processing and pattern recognition in real-time. It is now possible to equip existing Things with a digital heartbeat via minimally invasive sensor technology. We call this Artificial Intelligence of Things (AIoT). In this publication, we describe the idea and basics and show application examples.
Aging technical systems, commonly referred to as “Legacy Devices”, are an integral part of our daily lives. Who does not know the picture of an elevator from the seventies that tries to close its doors in an endless loop? The malfunction is simply caused by dirt on the reflector that triggers the door sensor. House inhabitants only shake their heads and nobody informs property management.
Elevators, finned radiators, kitchen appliances, and even the humans themselves share their fate as "Legacy Devices". This generally refers to "devices" that have no own digital identity on their own. The automatic acquisition and evaluation of data and the corresponding actions or recommendations are therefore not possible, sometimes with fatal consequences.
Challenges: Operation of Legacy Systems
Existing systems often fulfill their original purpose. They heat and cool, process workpieces, transport goods or people, and carry water and other raw materials.
Functioning infrastructure is only modified or even replaced if there is a significant change in the operational requirements. Examples of these conditions are significant increases in efficiency through the use of new equipment or the demand for improved environmental protection requirements. Modifications of existing installations represent a particular challenge. In many cases, the original manufacturer of the product no longer exists or its successors have not followed its further development. It is also possible that changes may require a renewed inspection of the installation by the authorities or even a complete re-approval, which fundamentally calls into question the economic viability of the project of renewal.
The aim is therefore to operate existing infrastructure as long as possible, possibly up to total failure. In doing so, one accepts the sometimes enormously high maintenance and repair costs, since a complete replacement, especially in the real estate sector, would involve a temporary loss of revenue and additional compensations and would be even more expensive than ongoing maintenance. There is a risk of discontinuation of services - the repair of the existing infrastructure is not covered by current revenues and there is no capital for complete renewal.
Even with proper functioning and acceptable maintenance costs, it would be unwise to remain inactive. Minimally invasive extensions of existing systems can increase the economic efficiency of the installations, improve comfort and user experience and thus extend the life of the products - a contribution of modern technology to sustainability.
Evolution into AIoT
Developments in recent years have created a great basis for meeting these challenges. Overall, milestones were necessary in all relevant areas, from basic technological developments for sensor technology and data processing, to the optimization of supply chains and the operational acceptance of technical systems. The right time to get started with AIoT is now, because:
- Prices for powerful sensors and microprocessors that meet all requirements for security, connectivity, and performance have fallen below the critical limit that allows total module costs of less than 10 USD [1]. As a result, hardware hardly plays a decisive role in the amortization of the total investment.
- Optimizations of low-level code and special coprocessors enable the implementation of AI applications based on neural networks even on small devices of the sub-10-USD class.
- Universal and easy-to-use communication networks, some of them even free of contract (examples: NBIoT, LoRaWAN)
- More and more companies have implemented a cloud strategy, enabling IT and security departments to introduce digital solutions quickly and manage and administrate them efficiently.
- Cloud services and software solutions for IoT and intelligent asset management have reached an industrial maturity level and usability that is suitable for non-technical users. The times when early adopters still had to act as beta testers are over.
The minimally invasive approach
With available sensors, we distinguish between invasive and minimally invasive implementations. Invasive implementations generally require intervention in closed-off technical systems and are excluded in many areas of application, as they can invalidate approvals. Minimally invasive sensor solutions do not have any such requirements but specify that sensors are only installed outside the legacy asset and thus cannot influence its integrity. As a result, sensor values may be less meaningful or require more complex signal processing.
There are often solutions to both approaches for measuring the characteristics of an asset. For example, for measuring the temperature of a component inside device one has several options:
- Housing temperature using a temperature sensor (minimally invasive): a PT100 temperature sensor is attached to the spot on the outer housing closest to the component. This solution provides only inaccurate, highly delayed measurements of the actual temperature of the component.
- Component temperature using a temperature sensor (invasive): the housing is opened and a PT100 temperature sensor is attached directly to the component. The integrity of the instrument may no longer be guaranteed.
- Component temperature using infrared sensor (minimally invasive): an infrared sensor is directed at the component through a cooling fin, which measures the temperature safely and reliably. However, the sensor technology used here is significantly more expensive than the PT100 temperature sensors from the previous solutions.
In the following, we show examples of AIoT technology in various application areas. Special attention is paid to minimally invasive implementations.
Use Cases
We present such minimally invasive implementations in the following use cases. Each of them features one legacy asset that requires recurring human supervision and care; otherwise, we have a risk of substantial damages or discontinuation of operation.
Monitoring a passenger elevator
The operation of a passenger lift in a residential building is to be monitored. This is to prevent critical damage and minimize failures.
A sensor kit is installed inside the elevator next to the door. It contains an acceleration sensor, which detects the movement of the elevator in terms of direction and acceleration. One of the two proximity sensors triggers when the door is fully closed, while the other detects full opening. A sound sensor registers operational noises. This results in the following digital representation of the elevator:
- Movement in [Standing, Up, Down]
- Acceleration in g
- Door status in [open, closed, in motion].
- Operation sound level in dB
With the help of the door status, unusual door behavior can be registered, which may be caused by jamming or contamination of the door sensors. Operating noises in combination with movement are used to monitor the mechanical condition of the lift.
A small investment enables the operator to prevent defect situations (e.g., passengers trapped in the lift) and oftentimes small fixes such as applying lubricant at the right time can prevent a catastrophic system failure.
Monitoring a compressor
The piston compressors of a mechanical workshop are to be monitored to detect typical defects. Such problems are leakage, blockages, and engine damage.
A sensor kit records the emitted noise and vibrations, as well as the temperature at several critical points. Furthermore, cameras record the analog barometer gauges at the compressor and the main tank. Using pattern recognition, the current pressure is derived from the scales and needle positions.
Thus, it is possible to create a Digital Twin of the piston compressor with the following attributes:
- Temperature at measuring points 1-3 in °C
- Volume in dB
- Vibration in g
- Pressure at compressor and main tank in bar
The temperature sensors are used to detect overheating of the engine. These can be indicators of a clogged intake filter. Volume and vibration provide information about other problems with the engine or the overall construction (e.g. loose bolts). A long delay between the pressure value changes at the compressor and the main tank can indicate blockages. A slower pressure build-up is also detected and can also be the first sign of impending problems.
Moisture and mold protection
A damp cellar is a common problem for a significant portion of real estates, even of more recent year of construction. As a result, brick walls and concrete suffer alike:
- Increased pH levels: integrity of concrete and adhesive or coating materials
- Decreased strength: lower compressive strength, lower durability
- Microbial growth: organisms affect the health of people in surrounding area
To monitor the situation a sensor kit is placed on the wall next to a drainage pipe, where the risk of moisture damages is the highest. The contact between the sensor kit and wall contains a surface thermometer that measures the wall’s temperature. A Bosch BME280 3-in-1 sensor [2] measures relative humidity, barometric pressure, and ambient temperature.
The sensor kit allows us to create a digital twin of the cellar room with the following properties:
- Wall temperature in °C
- Ambient temperature in °C
- Relative humidity in %
- Barometric pressure in hPa
It also calculates simple metrics that are useful in the context of humidity and mold protection:
- Dew point in °C
- Mold risk in Days to Mold
- Equilibrium moisture content in %
The system allows service personnel to monitor the values over time and dispatches notifications in case critical values are reached.
In its second iteration, the system should not only alert but also try and actively keep the cellar dry. For that purpose, a second kit is mounted on the window consisting of a drive capable of opening and closing the window and another 3-in-1 probe that is fixed outside of the window. As long as outside conditions are more beneficial (warmer and dryer) the window is automatically kept open to ventilate the room.
Overview
As demonstrated, rather simple sensor kits can be utilized to retrofit powerful monitoring capabilities with low investment costs and – most importantly – without any interference with the assets’ core operations and integrity.
Benefits
The retrofitting of existing systems with AIoT components can extend the life cycle of products - this is a tangible contribution to the sustainability of our economy! The concept makes the owner of a system more independent of the manufacturer of the existing equipment because the minimally invasive retrofitting of sensor technology and IoT edge devices makes expensive investments in completely new control systems unnecessary. AIoT thus increases the freedom to make decisions, reduces operating costs, and in many cases improves the user experience.
Outlook
The areas relevant to AIoT have been subject to both disruptive changes and steady trends that reach critical points regularly. For enterprises, it has become difficult to obtain an overarching picture across all relevant topics.
If you want to bring new life to your Legacy Assets Deloitte is the partner of choice to materialize your ideas and deliver them to your customers.
This article is the first iteration of our blog series on Artificial Intelligence of Things. You can look forward to reading more about:
- Retrofitting assets at the Deloitte Digital Factory
- How do things talk to us? Current trends in IoT communication
- How to design a custom sensor kit for AIoT retrofitting
- AIoT Edge architecture
Contact us
Andreas Staffen
Andreas Staffen verantwortet das Offering IoT and IT Architecture (Smart Manufacturing) für Deutschland und gestaltet die Digitalisierung der Supply Chain seit 2004. Dabei begleitet er deutsche, europäische und globale Unternehmen bei der erfolgreichen Umsetzung schlanker und integrierter IT Architekturen für die Entwicklung und Produktion. Durch die Umsetzung des Industrie 4.0 Gedanken in der Deloitte Digital Factory werden die Auswirkungen auf die Geschäftsmodelle unserer Kunden erlebbar und die weitere Gestaltung einfacher realisierbar.