Data is becoming more and more essential to support decisions in the industry. Performant companies are making decisions based on facts (numbers, KPIs) and no more on feelings.

There are different tools in the industrial IT market for data collection in production to improve the decisions of operators, engineers, and managers.

ERP was the first system to bring data management to support financial and planning decisions. MES (Manufacturing Execution Systems) is deployed to collect data to support activities in production, product tracking, asset health, performance monitoring, quality, management - SPC, and traceability.

Process historian is another tool used to collect process/production data to support decisions and analysis by providing access to historical data for analysis and real-time data for life decisions. 

IoT (Internet of Things) and Cloud solutions (Cloud hosted databases) have recently entered the industrial space to deliver data from connected devices in situations where automation systems are not present or too old to provide accurate data.

Those tools provide different services (a process historian is a pure monitoring solution whereas an MES can also execute production requests, meaning impacting the process) and use other technologies, so an efficient solution for a plant would combine a set of tools, depending on needs and priorities (high payback first). 

The integrated solution can come from the same provider. Still, it is possible to combine solutions from different suppliers like Automation from A, SCADA/HMI from B, MES from C, Historian from D, IoT/Cloud from E, and ERP from F with no problem, knowing that each provider will try to draw you to an exclusive relationship (and dependency). Today, there is no evidence that sourcing from a single supplier will be cheaper.

 

This article was contributed by our expert Didier Collas

 

Frequently Asked Questions Answered by Didier Collas

Q1. Which tool is best for data collection?

• The best tool to use depends on the data you want to collect. If you look for process data (including hybrid or discrete processes like volume counts or energy consumption) with some context (production order, shift, etc.), a process historian will make it very well (there are several options on the market). 

If you need high volume (many variables or many samples over time), you need to look for a real process historian. Be careful that a single recording in a relational DB could be OK for a few parameters to monitor and for a short period of time.

• MES would be the right solution if you want to record events (transactional data). You can then record downtime events, production volumes, and consumptions of material or energy per operation/work order and put it in context (line, team, shift, etc.). 

We know that implementing an MES requires more effort than a process historian, as you need to define the production model as a reference over which data is recorded and checked.

• The final solution is the IoT approach for equipment with limited information (lack of sensors or limited communication capacity of the automation PLC). The solution can be based on a Cloud hosted DB or a local (called Edge) one. In that case, you can add IoT sensors and collect the data to a local DB, and you are close to the situation of a process historian.

Be careful when implementing Cloud solutions with cyber security as your data will move out of your plant with an open door, to the external world, so you need to protect your local network against potential trials of external control.

 

Q2. How do you collect industrial data?

At the bottom of the chain, the data must come from sensors (more reliable than a value entered by a human being). In most cases, the sensors will be connected to an automation layer (PLC or DCS), and the automation layer will provide the data either to an HMI SCADA (in the case of MES) or directly to a process historian. 

The information can also be generated by the automation layer (like a downtime event) and then pushed to the storage tool. As said before, it is also possible to use IoT technology to extend the automation layer data capabilities. 

This can be the case for validated systems (like pharma production were modifying the automation layer to provide data would require a new expensive system validation), for old systems with limited capacities (number of I/O, network capabilities or bandwidth) for cost reasons (IoT points are cheaper than automation ones as they don’t need cables to transport the data).

 

Q3. What are big data tools and technologies?

Big data is more a concept than a technology. As it is easier to generate data, people speak of big data, and some suppliers are pushing for a unique data store (data lake), where to store industrial data as well as marketing, product, and assets quality data to be able to perform analysis across domains and machine learning.

This requires storage in the cloud as this is a costly solution in resources. The cost is huge to collect (multi-sources and multi types of data), store and retrieve the data (a few dollars per transaction). Cloud providers like IBM, Microsoft, Google, and Amazon push IT managers to use this solution, which makes the company fully linked to the supplier.

Imagine you put all your company data on a data lake based on one supplier like Amazon or Google (petabytes of data per year), and after several years wanting to switch suppliers, it would cost a lot of money and time.

 

Q4. What types of data are typically collected by manufacturing systems?

Manufacturing systems are not only tools to collect data (logging manufacturing activities), they also manage production by starting/stopping work orders, downloading recipes to the automation (how the product should be produced) or a set of set points, requesting operators to take samples for quality control, etc.

But on the data collection side, the MES records information related to:

• Asset performance (downtime events with reasons, time used, production volume, scrap or bad quality products volume) from which it calculated OEE (Overall Equipment Efficiency).
• Order execution-related information: where is the work order and how it progresses through the process, how much local storage is available on machines or inventories.
• Quality/traceability-related information: Key parameters over the process (per work order and operation), which materials have been consumed, which machine has been used, and which recipe was.

 

Q5. How do you analyze manufacturing data?

Once the data is stored, analysis can be performed via reports produced on demand (per shift, line, work order, etc.). These reports are defined per project, and the tools provided by MES suppliers help you define your own reports. 

Very few products offer out-of-the-shelf reports per domain. They allow comparing the production of the same product over time over different plants and lines and correct wrong practices.

The information can also be integrated into the HMI/SCADA screens making the life of operators easier as they don’t need multiple systems to manage the production.

For process historians, the primary analysis tool is a page of trends where users can select which parameters they want to analyze in a user-friendly way. Data can be extracted and pushed to Excel or BI analysis tools like Power BI from Microsoft or Tableau.

Cloud-based solutions now offer more self-service capabilities to create your own dashboards (mix of maps, graphics, trends, and events). They usually provide easy-to-usekr query tools to find the data needed for the analysis and a library of objects to present the data in a meaningful way.

 

Q6. What are analytics tools?

You can use business intelligence tools to combine data from multiple sources like a process historian, an MES, an ERP, an Alarm DB, and Excel sheets and create a data store (or data cube) where the data is prepared for analysis. 

This permits multi-dimensional analysis with good data extraction performance. You can then compare assets performance or product quality with a graphical tool (Power BI or Tableau, for example) in which you create your own dashboards supporting your analysis.