Karl Walker explains how cognitive computing lets FMs use it to make informed decisions about how to optimise the experience of occupants, staff and management while minimising energy waste
09 October 2017 | Karl Walker
The main challenge facing FM teams is the number of job roles the uniforms must cover.
Buildings create a huge amount of unstructured data from many sources and pieces of equipment. Karl Walker explains how cognitive computing lets FMs use it to make informed decisions about how to optimise the experience of occupants, staff and management - and minimise energy waste
Although preventative maintenance, perhaps replacing a product based on a manufacturer's lifetime guidelines, is widely practised, research shows that only 18 per cent of assets have an age-related failure pattern, with 82 per cent of asset failures occurring randomly.
Research shows that 30 per cent of such maintenance is carried out too often and 45 per cent of all maintenance efforts are deemed ineffective. Add to this the fact that 40 per cent of preventative maintenance costs are spent on assets with negligible effect on uptime failure and 45 per cent of all such work is potentially disruptive.
Preventative maintenance isn't a bad idea, but condition-based monitoring and cognitive computing that drive predictive maintenance and better operational decisions can greatly help to optimise asset performance, inform better operational decisions and, of course, save money.
An intelligent building - a definition coined by the Intelligent Buildings Institute - is "one which provides a productive and cost-effective environment through optimisation of four basic elements: structure, systems, services and management, and the interrelationship between them."
Communication protocols, such as DALI for lighting, Modbus for motor and valve controllers and BACnet for HVAC equipment, are considered to be standard features now, often natively incorporated into products.
Taking a DALI luminaire or ballast as an example, about 10 items of data are constantly generated that might include lamp output level, hours run, voltage or temperature.
But because of the often disparate nature of building controls and the way in which contract packages are awarded, this data is rarely exploited as it remains within an isolated function or service that is generally unlikely to be connected to any IT enterprise infrastructure, let alone being integrated with all other building services controls.
'Intelligence' requires data and, with no one managing the interrelationship between these myriad disparate systems, the idea of a 'smart building', let alone any ideas of predictive performance and maintenance, falls by the wayside.
Cognitive computing forms the backbone of any predictive maintenance system, drawing on multiple sources of data, real-time and historic, predefined (e.g. manufacturers' data or BIM data) and measured (e.g. from sensors and equipment) and extracts, analyses and contextualises this data within self-learning algorithms that predict operational outcomes and impending equipment failures.
These systems already exist in such packages as IBM's Maximo, TRIRIGA and Watson platforms, and many other CAFM (computer-aided facility management) packages.
Recent standards such as the WELL Building Standard have driven a rapid expansion of green buildings and environmentally conscious building practices and are having a significant impact on design to enhance human health, comfort and well-being. WELL identifies 100 performance metrics, design strategies and policies that can be implemented by building owners, designers, engineers, contractors, users and operators. This also requires accurate data from, potentially, additional sensors and some disruption to install them.
Converged controls and data aggregation
The recurring theme for any intelligent building or predictive software system is that data must be captured from all sensors and controllers through the building or estate and consolidated on a single platform where the environment can be accurately controlled according to the needs of the occupiers and recommendations can be generated to turn into work orders for FMs. This is a tough task; many incumbent controls won't have existing connectivity to IT communications networks,let alone interconnectivity with each other.
Beckhoff's approach to this issue is to use a single PC- and Ethernet-based controller to connect all systems together, irrespective of physical network provision or protocol, and aggregate all building and asset data into one location. In a new-build or refurbishment, there's no reason why this same system couldn't replace all the traditional control functions too, saving the headache of retrospective interconnections.
This 'single system' approach has other benefits; for instance, the data from a PIR presence detector connected to a single input can be used by any aspect of the building function - lighting and air-conditioning - thus eliminating the need for multiple sensors in the same location, which we see all too often in modern buildings.
Using existing cabling infrastructure such as DALI or KNX can also help to minimise installation disruption. Where additional switches and sensors are required, EnOcean power- and battery-less energy-harvesting devices can be used.
At this point it is imperative that the 'power' of these intelligently networked devices is leveraged, extracting, storing and forwarding all data produced.
Edge analytics and cloud computing
Many people will assume that the whole idea of cognitive computing will be cloud-based and connected in some way with the Internet of Things. While IoT-enabled devices can assist with connections to the internet, it does not necessarily mean that they pass the required data upon which analytics can be performed or do it in a secure manner.
In many cases, analysis can be performed locally, at source, known as 'edge analytics'. By running data through analytics algorithms as it is produced - at the 'edge' of the local network - parameters can be defined on what material is worth sending to the cloud (or local data storage) and what isn't. Local control decisions can also be made and predictive outcomes sent to the relevant people.
As Beckhoff's controllers use an open PC-based platform running Microsoft Windows operating systems, analytical software can be run on the same system that connects and is controlling the building and its assets. Where data needs to be sent to a cloud-based storage and analysis platform, such as Microsoft's Azure or IBM's Bluemix, this can be done in real-time using secure MQTT messaging with pre-written software function blocks to simplify the process.
Value engineering - a false economy
So to generate usable information, data is required and until all data can be 'tamed', collected centrally, processed and interpreted, the whole idea of an intelligent building cannot be realised.
Unfortunately, 'value engineering' often sees the removal of intelligent sensing devices and control systems at the early stages of construction, the wider implications of which are perhaps not clearly understood. The more significant ramifications of this are that buildings fail to perform to predicted levels. A recent BRE report estimated this performance gap at between 200 and 450 per cent and, when you consider that the operational energy or the energy used in using a building is up to 50 per cent of the operation costs (before being potentially multiplied 2-4.5 times), the initial cost of implementing intelligent controls in buildings pales into insignificance.
A whole-building, intelligent control strategy, implemented in accordance with the BS EN15232 directive - 'Energy performance of buildings. Impact of Building Automation, Controls and Building Management' - can not only halve energy bills, but also provide the platform for a cognitive building that can reduce operational costs.