June 14, 2022
By: Daniel Voelp
Why your plant needs both top-down and bottom-up analytics insights?
Process manufacturing plants are incredibly complex, with an immense number of moving parts that interact with each other in complicated ways. Nested systems mean that even small errors in one part of the plant can have a knock-on effect on the entire plant.
As a result, it’s a challenge to access real visibility into most of the processes, systems, and equipment, but technological advances such as predictive analytics, artificial intelligence and machine learning (AI/ML), industrial Internet of Things (IIoT), and big data are changing that and opening up windows into plant operations.
The many different analytics systems out there can be classified into 2 main types:
- Top-down analytics
- Bottom-up analytics
Here’s a brief overview of the differences between these categories, and how they can work together to help your plant.
What is top-down analytics?
Top-down analytics can be thought of as a broad yet sometimes shallow way of looking at plant activity. With top-down analytics, the solution covers the entire plant, tracking activity across all its devices, processes, systems, and equipment, regardless of their manufacturer or origin.
Top-down analytics solutions draw on process data that’s in the data historian, gathered by the sensors and IoT-connected devices that are already in place in every process plant. In theory, any human operator can watch the data to see patterns and trends, and spot anomalies within them, but it’s not very realistic in practice.
This kind of data analytics needs to be carried out 24/7, because significant changes can blow up from tiny anomalies into something serious – and expensive – extremely quickly.
Meanwhile, the datasets are immense, and usually look something like this:
There are thousands of tags across all the equipment, resulting in a complex network of the behavior of each item and different relationships between them. As you can see, humans can’t easily analyze this data. It requires AI.
The value of top-down analytics lies in its ability to identify small changes in minor processes or in small and inexpensive items, which aren’t typically covered by dedicated bottom-up monitoring but can cause serious damage with significant economic impact. Those changes can be too small for the human eye to notice until it is too late.
Top-down analytics reveals anomalies anywhere in the plant while they are still small, so they can be remedied quickly and relatively cheaply before they snowball and become expensive, complex, and/or necessitate unplanned shutdowns.
What is bottom-up analytics?
In contrast to top-down analytics, bottom-up analytics take a deep yet narrow approach to plant monitoring. A bottom-up solution examines specific devices or items of equipment, often manufacturer-specific, tracking just one type of equipment or device. It draws on diagnostics data from that device or item. For example, valve management analytics uses data stored in the positioner and derived from the DCS.
Although bottom-up analytics has a narrow focus, the part of the plant that it covers could be extensive. In the case of valve management, valves make up as much as 30% of the plant, comprising not just control valves but on/off valves, hand valves, small valves, and more. Valves can be critical for the process and controlling the flow; if the valve is blocked or not operating correctly, that affects process quality and could result in a plant losing batches.
While bottom-up diagnostics like SAMSON Valve Management can use AI, it’s not a necessity. Bottom-up analytics relies on deep human or AI knowledge of the equipment and process, including how it works and what nascent damage looks like. The solution analyzes data about the part, formats a report, and enables the plant manager to assess the health of the part and when it will need to be maintained or replaced.
Because bottom-up diagnostics focuses on just one type of equipment, it can provide much more precise information about what may have gone wrong and what to do now than top-down analytics solutions.
How bottom-up and top-down can be more effective together
Each of these approaches has its pros and cons, which is why you’ll get maximum value when you use them both together. Top-down diagnostics requires process data, but not every plant tracks and stores this data in their historian. Bottom-up diagnostics uses equipment-specific data which is stored about that specific item, but it leaves big gaps in plant coverage.
A combination of both methods delivers a broad coverage of the entire plant, including those parts that don’t have their own dedicated solution, together with an in-depth analysis of the most critical elements.
The two approaches work symbiotically. For example, a top-down solution may trigger an alert about an anomaly, but plant personnel still need to investigate the alert, discover what caused it, and inspect its source. After receiving the alert, they could check the bottom-up data to see if there’s more information that guides them to the origin of the problem.
Alternatively, there might be an anomaly in the diagnostic data that shows there is something wrong with the valve operation, but nothing wrong with the valve itself. This would indicate a process issue which can then be identified and investigated using top-down analytics.
Top-down and bottom-up diagnostics work to different timelines. Top-down diagnostics track changes that appear fast and must be addressed fast. Bottom-up issues move much slower, with data stored over a lifetime and checked only about 3-4 times a year. Bottom-up analytics deal with long-term concerns, similar to when maintaining a car that will last 30 years if you look after it.
| Top-down analytics | Bottom-up analytics |
|---|---|
| Uses process data from the historian | Uses diagnostics data |
| Covers the entire plant | Covers specific, selected items of equipment |
| Is manufacturer-agnostic | Usually only looks at equipment from a particular manufacturer |
| Changes occur quickly | Changes can take a while to appear |
| Data is broad yet shallow | Data is deep yet narrow |
How do SAMSON Valve Management and SAM GUARD complement each other?
SAMSON offers two types of analytics solutions: SAMSON valve management, and SAM GUARD.
SAMSON valve management is a bottom-up diagnostics solution based on SAMSON’s expertise in valves, with over 100 years of experience. The service draws on just 3-5 data points which reveal a massive amount about valve health, behavior, and maintenance needs if you have the expertise to interpret it correctly.
These data points include:
- Valve open range, i.e. if the valve does not open as needed, it causes more wear and tear on other components
- Valve closed position i.e. if the valve is sealing correctly
- Valve size i.e. if the valve is too big, it requires more energy to power it, placing more pressure on the valve, which in turn requires more maintenance
The solution also provides detailed views of the valve affected, histograms, trend graphs, and reports, as well as detailed recommendations for what to do once the solution identifies an issue and specific guidance on how to address it. There’s also an option of hiring SAMSON experts to deal with it for you.
On the other hand, the SAM GUARD top-down analytics solution utilizes a combination of machine learning and human intelligence to map the entire plant, and guide the solution to understand the systems and networks that comprise it.
SAM GUARD monitors everything in the entire plant, enabling a connected view into emerging issues so the solution can deliver fast alerts about growing issues. SAM GUARD also provides smart alerts that help prevent alert fatigue, so you’ll only receive notifications about the most relevant and crucial anomalies.
Diagnostics aren’t one-size-fits-all
As you can see, each type of diagnostics solution has its pros and cons. You can improve operational efficiency at your plant and stay on top of emerging issues by applying the two types of diagnostics – top-down and bottom-up – so that they work in harmony, each filling in the gaps that the other leaves to give you complete and effective visibility into your systems and networks.
