Empiricai’s AI-based Industrial Analytics platform empowers industries to improve yield and reliability while reducing costs. Instrumentation Monthly recently caught up with Emipircai’s CEO Salman Chaudhary to talk about artificial intelligence (AI) and the value it brings to plant management
What does Empiricai’s Industrial Analytics do?
Plant data is being produced from tens of thousands of sensors in an industrial plant every second. This data helps industrial plant operators and engineers to evaluate whether the plant is running optimally, to troubleshoot when there is a problem that relates to quality or output, or to track the consumption of resources or energy. And, of course, plant operators and engineers are constantly tweaking all of the parameters they have under their control to maximise the yield of the product they are producing, to maximise the consistency of the product, and to do it at the lowest cost, consuming the least amount of resources. Doing this requires analysis of the data to look at patterns and then, based on the direction of these data values, evaluating if the plant is operating efficiently or if corrective action if required. However, industrial data analysis is a research intensive process which currently has a long life cycle due to inefficiencies in many steps of this process.
These limitations can lead to delays in obtaining valuable insights that can result in production losses and increased cost.
Our Industrial Analytics product is a software platform that equips users with advanced selfservice analytics in an industrial plant setting.
It allows process engineers and performance engineers to rapidly investigate and share insights from plant data and enables them to do it on their own without the need of an IT specialist or a data scientist. The platform contains features that will allow you to look at and manipulate data correlations, anomalies and trends to keep the plant running without disruption or to solve a specific problem you may be troubleshooting.
What are the key benefits of Industrial Analytics?
Industrial Analytics improves productivity for plant engineers and process engineers on some pattern matching with current problems.
Storing a lot of data (if you store ten years of subsecond data, for example) can lead to a slow down of some of the more complex calculations or analysis that you might be doing on the data. So there is a small trade off based on the amount of data that you want to look at. Of course you can store as much data as you
want but when you come to analysis you have an option to select data ranges and if you select a very broad date range it slows down the analysis in some cases. Typically, our customers would store data for a year; some customers store it for three years.
Is the data easy to read?
It is a customised product that has been designed by process engineers and plant operators for process engineers and plant operators. As such, we believe it is very easy for them to analyse data, view data and derive insights from that data. The user interface is designed to make all aspects of data analytics very quick and accessible. Industrial Analytics allows for different levels of users so you would have users like process engineers, who, for example, would interact with the data day-in day-out in a chemical plant, and then you would have a plant manager who is more interested in a subset of that data, not doing much analysis but looking at a dashboard that is providing them with a real-time view of certain KPIs. So we allow everything that they need to do ranging from completing daily tasks through to troubleshooting emerging problems. The platform helps improve plant performance, increase machinery reliability, reduce energy consumption, reduce environmental impact, and reduce raw material wastage. These are all things that plant engineers and operators aim to do and we make it easier for them to do so by making specific and targeted KPIs that can be measured, monitored and refined.
We have a base module that is the Industrial Analytics platform and on top of that we enable customers to add on any AI machine learning or physics-based models that help them to solve a particular problem or specific use case. So, for example, if the customer already has a model that helps them optimise the net heat exchange rate within a particular part of the plant, they can integrate that seamlessly within our analytics platform and leverage the dashboards and the user interface of our product to be able to utilise that model. Customers can do that for a number of different models, whether they are models they have created themselves, third party models, or a custom Empiricai model;they can all be integrated into the same product to help solve problems of efficiency, quality and wastage.
Can Industrial Analytics be used with any equipment?
Industrial Analytics integrates with the data that comes into a distributed control system (DCS) or a historian from sensors throughout the plant. As long as there is data that is available, that data is then leveraged by our product. It doesn’t matter what machinery or equipment there is, as long as there are sensors that are providing data, Industrial Analytics can operate. It is completely agnostic around specific types of machinery and equipment in terms of the manufacturer but we do rely upon data being supplied from some source. So pure analogue systems that may not have any sensors would not be suitable for the platform. different users to have different views so that the data is relevant to the audience and easy for them to navigate.
Can Industrial Analytics be set up to trigger alerts when essential maintenance is required?
Users can set alerts to generate early warnings when an anomaly is detected whether it is on the asset side or in the process data. Any anomalies can trigger an alert. Industrial Analytics essentially captures the customers’ operational knowledge digitally and then applies Empiricai’s proprietary machine learning algorithm to really simplify anomaly detection and troubleshooting. There are a couple of different types of alerts that can be set up. Value based alerts allow you to set thresholds for different values for any process in the plant.You can set, for example, an alert threshold for the number of rotations of a particular machine so whenever you reach the prescribed lifetime of a particular component or a machine, an alert will be sent. A more complex alert is a pattern based alert.
For example, when certain patterns lead to a nonconformist behaviour or a degradation in quality you can set an alert to be triggered when the 10, 20 or maybe 100 parameters that created that negative pattern start to emerge in combination again. This highly complex alert helps process engineers and plant operators make sure that problems are not repeated or are warned well in advance of that repeated pattern.
How easy is it to set up?
It is designed to be a self-service product so it is very easy to set up.All Industrial Analytics needs is to connect to a source of data - often that is a historian system or data that is derived from a control system that is on site. So the set up is very easy for the data to be imported into our product and that can be done on a real time basis, a batch basis or an overnight/ once-a-day basis.
There are two versions of the product - the first is a complete SaaS product based in the Cloud. It is simply installed, connected to the data sources, the data is populated and then you are ready to start doing some analysis.We also provide an on-premises version of the product - certain industries or industries in certain countries may have a specific requirement to be on-premises so we have a product for that and, again, that requires a very simple install of the software.
There is some configuration that needs to be done in terms of setting up users and tagging some of the data that may not be tagged but that is a process that we can take our customers through. A challenge that often appears in the setup stage revolves around the quality of all the data as well as the completeness of the data.
We often find that there are sensors that are faulty or that are not providing data, or that the data is getting corrupted somehow and it is not being caught by the control systems or the historians. Sometimes the sources need to be replaced or fixed as part of the implementation process. Of course, we’ll still take on that data but the data may not be as useful if it is not complete.
How long is the data stored for?
There are no limits from our side on how much customers can store and how far back they can go in their history to analyse problems or to do
Finally, does Industrial Analytics bring additional benefits for plant operators in the COVID era?
The pandemic has accelerated the digital transformation journey for many industrial plant managers. They have been forced to adopt digital tools and to implement processes that support digitilsation far earlier than they had planned. Empiricai’s Industrial Analytics helps organisations to be more productive and to do more with less. In addition to that, because Industrial Analytics is primarily a SaaS product, it means that remote access is much easier so you don’t have to be in
the control room of the plant to troubleshoot. You don’t even need to be on site to ensure the plant is operating at an optimal level and the KPIs that you are monitoring are all within the ranges that you want. Collaboration capabilities have also been built into the product. We have integrated chat, videocalling, white boarding, screen sharing, and saved analysis sharing into Industrial Analytics; if a process engineer is in the middle of doing some analysis and they need to hand this over to the next shift worker then that analysis can be saved and passed on with ease. You can also share alarms and cases. There is a lot of collaboration that is built into the product that allows multiple people offsite to be able to continue to do what they need to do remotely and collaboratively