17 Nov Profitable facility management? Avoid these 5 mistakes when implementing your anomaly detection strategy
17 November, 2021
Nowadays, most facilities or physical assets managers use some kind of a software management tool. It may be a stand-alone application or a cloud-based system running on several terminals, a BMS (building management system), an EMS (energy management system) or a multipurpose platform.
Quite a few of these solutions promise to the management teams an elaborate data analysis, which results to a realistic overview of the facility equipment health and the prediction of future malfunctions. These predictive analytics come usually under the name of “anomaly detection” or “outlier detection” and constitute a significant factor of how competitive a business can be.
However, the extent to which a business can be benefited by anomaly detection is heavily dependent on the implemented strategy. This includes both hardware and software, as well as the decisions on which data will be collected, how they will be categorized, which procedures will be followed for identifying anomalies and what alarms or actions will be triggered once an anomaly is detected. And since anomaly detection is a term widely used to describe different tools and implementation philosophies, it’s important to select the right system and strategy for the specific facility. Here are a few critical mistakes one must avoid:
Beware of what you label an “anomaly”
In the early days of an anomaly detection system implementation, there is usually enthusiasm around the different types of detected anomalies (outliers or novelties). Specific rules are introduced, together with alarms and less urgent reaction scenarios. However, as more and more anomalies are being identified, things start becoming overwhelming.
The system’s control screen gets full of alarm signals and abnormal operation indications, the O&M people run around in circles and the real meaning of “urgency” and “efficiency’” is lost. Furthermore, these can be very stressful conditions for inexperienced personnel, who are not familiar with the severity of every possible malfunction.
So, it’s important to think thoroughly since characterizing data or events as anomalies , incorrectly or with no reason, may lead to undesirable false alarms. Also, it’s equally important to revise this categorization periodically, using the accumulated knowledge from successful operation.
Data isolation or miscommunication?
Deciding what type of data you will collect and which sensors you will use to do so, is one thing. And it’s another thing to carefully analyzing all the facility’s data collection points, the existing infrastructure and the possible needed additions to complete a reliable anomaly detection system’ s input.
What plays important role in the effectiveness of such a system is the ability to intercorrelate data and make use of every potentially useful bit of information. To do so, your anomaly detection application needs to be integrated into the existing IT, ERP and other operational systems. This process should be both communication protocol and device agnostic, to ensure that no network or HMI specific characteristics will prohibit your system from providing you with valuable conclusions.
Don’t incorporate artificial intelligence, just for the shake of it
Both, artificial intelligence (AI) and machine learning (ML) are trendy terms. Any ambitious IT system is being marketed as having AI capabilities and promise a much more efficient operation. But can they deliver?
Unfortunately, there is usually not much talk around the anomaly detection algorithm itself, nor the specific ways that AI techniques are being used to guarantee autonomous decision making. And although not all asset managers are that technical savvy to know how to choose the right anomaly detection method for their facility, they should have enough insight and guidance from the system supplier, in order to come with an informed decision.
Additionally, the anomaly detection application should be relatively easy to set up and – most importantly – to re-configure. Best practices show that as users become wiser on how to best utilize the systems data and characteristics, or the operational conditions change, its imperative to customize the system accordingly for maximum reliability on the predictive analytics.
Single location data analysis
Centralized data analysis has been very common in the past. The topology of the network usually dictated all data to be fetched from the distributed points to a centrally installed application, which would then perform the necessary calculations, estimations and predictions. What this approach lacked, however, was speed.
Today, communication between IT systems, edge points and cloud-based platforms is much faster and it’s possible to come with an extra benefit. Given the capability, system moderators can decide for some data analysis to take place at the edge, so that local incidents can be promptly acted upon. This also saves time and money in communications and infrastructure. The rest of the analysis and high-level decision making is left for the cloud, which has now fewer tasks to accomplish.
Anomaly detection strategy without ESG metrics?
Energy efficiency is in the heart of a facility management application and its anomaly detection method, as it is a significant factor for the limitation of a building’s operational costs. Most frequently, they are directly connected with the facility manager’s KPIs. Therefore, the identification of inefficient energy consumption patterns is extremely critical for both people and businesses.
On the other hand, as the world has grown more sensitive around environmental and sustainability issues, the international organizations introduced the ESG (Environmental, Social and Governance) guidelines, which are expected to become increasingly more decisive for how future facilities will be managed. So, it makes sense for businesses to invest on advanced systems that can monitor ESG metrics, connect them to business KPIs and have built-in functions for the corresponding reporting.
If the above seem too much to be realized through a single platform, they really aren’t. In the market today, there are already a few available and reliable solutions. One of them is YodiFEM – the facility management IoT/AI platform from Yodiwo – which uses advanced anomaly detection techniques to help you save money and protect the environment at the same time. If you look for ways to manage your physical assets more profitably and identify OPEX leaks as soon as they appear, a 30-minute free consultation call with our Smart Facility Management team could give you all the information and confidence to take the right decisions.