How to get started with industrial data analytics
Knowing the key components of industrial data analytics can help streamline industrial processes for more effective data analytics.
Industrial Data Analysis Insights
- Understanding four types of industrial data analysis can mean the difference between reacting to something that already happened and knowing what will happening and proactively making changes to optimize a response.
- Choosing the best next step for industrial data analytics requires understanding the stage of data analysis now and the path to prescriptive data analytics benefits.
Understand industrial data analysis types.
See what industrial information help operations and other areas.
Learn about an example of prescriptive data analytics that provided manufacturing benefits.
Industrial data analytics can streamline industrial processes. First, what are the building blocks of industrial data analytics? Matt Ruth, president of Avanceon, explains more about industrial analytics in this partial transcript from the April 21, 2022, RCEP PDH webcast (archived for 1 year), “Just enough industrial data analysis?“ This has been edited for clarity.
Related data analytics articles
Relate articles in this data analytics series are:
– Webcast preview: What is just enough industrial data analysis?
– Webcast Q&A: Questions answered: What is just enough industrial data analysis?
– Data analysis disruptors in the industrial space
– How to overcome obstacles to industrial data analysis
– Industrial data analysis case studies, effectiveness.
Key components of industrial data analytics
The key components of industrial data analytics are simple because it’s the things that you use every day. It’s all of the back-end systems and in the front-end systems inside an operational technology (OT) stack.
Inside a facility, the different data sources are included in the things that enable analytics. That includes different sensors, programmable logic controllers (PLCs), controllers and input/output (IO) devices go into making up the way that a company controls a process or production, as well as in the data storage components. Different business data silos and operational technology data silos that exist inside the facility also are key components.
Tie that into front-end systems such as PCs, web clients, mobile apps, and tablets. Important from an analytic perspective is how information from these are shared with existing user interfaces. Visualization reporting and navigation methods are key components to leveraging data analytics and getting the information to the right people at the right time to make a data-driven decision.
Another big component is cybersecurity and cyberattacks. Security and risk mitigation are key components to data analytics when opening up the ports and allowing data to flow. Also important is where that data’s flowing.
The repository for information often is a cloud-based architecture. Next, examine how to coordinate components for data analytics. The place to start is in the manufacturing facility, specifically in the operational technology (OT) stack.
Inside the facility is multiple different data sources with a wealth of information that analytics needs to get and understand how to predict and support recommendations or show the issues requiring attention to improve. After identifying issues, put them in one spot. That can be in the preferred cloud platform or a private cloud, but the idea is to get data analytics information away from the existing databases so you can improve performance.
Industrial data collection, integration, analysis
Once you have the data in a database and cleaned up, math is performed in software or cloud. Using a model for the way processes run brings understanding to exceptions and issues that occur. From those exceptions and issues, dive in and use analytics to develop the idea or the event requiring action.
Once the analytics model understands, starts to learn and be able to predict that action, it can present and produce reports back to the same user interfaces, enabling people to make needed process changes. That loop can happen and circle a number of times as the analytics get smarter and the things get more refined.
Enabling people to modify processes based on analytics provides confidence and documentation need for process optimization in real time. The system should notify appropriately and provide visibility into what happens, why and when. Those involved learn to leverage data use and gain experience about what it can provide in real time.
Types of analytics: Descriptive, diagnostic, predictive, prescriptive
What types of analytics are there, how are they segmented and how are they portioned? Types of analytics can be organized by what questions they address. Analytics can be descriptive, diagnostic, predictive and prescriptive.
The first type of analytics is descriptive. Descriptive analytics uses historical data. It takes that data and reconfigures it into easy-to-read formats so you’re able to describe the state of business operations and learn from the past to see what has happened. It answers the question of what: What happened, what occurred?
The next type of data analytics is diagnostic. This uses historical data and identifies anomalies that are happening inside the data, highlights the different trends that it sees and allows an investigation of those underlying issues. Oftentimes, data analytics answers the question, “Why did that occur?”
The next type of data analytics is predictive. Predictive uses historical data, but it fills the gaps in the available information. It creates data models based on what it’s seen happen in the past, and it forecasts potential future outcomes based on that past performance. It answers the question of what may happen in a process when certain things occur.
Lastly is prescriptive analytics. Just like the others, this uses historical data, estimates outcomes based on variables. It also tries to optimize and set the system to be able to provide suggestions about how to improve on those outcomes. It uses a lot of algorithms, artificial intelligence (AI) and machine learning structures.
Using different industrial analyses together
Each component of analytics is designed to answer a question of what, why, what might happen, followed by what to do. How does that fit though in the factory? When applying analytics to a facility, what questions, what, why and what might happen? What are the answers? Answers fit into and apply to the manufacturing triangle of constraints. Everybody’s familiar with time, quality and cost as the main things that drive a manufacturing facility.
In theory, only two of them are attainable, but analytics provides a path a little closer to getting all three at the same time. With quality, analytics can answer questions around, how good is it, how is the scrap being produced, what rework do I have and how close am I to the specification and targets?
It can dig into how that happens, why that happens and what happens to that. From a time standpoint, it digs into how long things take. It talks about to analyze and dig into cycle times. Time spent is not redeemable, but future time can be used more effectively. Analytics can help with digging into the cycle time, changeover times, how long it takes to clean, and how long things take to schedule. Analytics can help answer the question, how much does it really cost? What is the yield? What is the magnitude of waste? How much downtime exists, why is it occurring and what are the overall operational costs are of making a product?
Facilities contain opportunities to answer the really hard questions that operational and financial personnel may want to answer around the performance of operations. Examples of the different types of analytics can help.
Analytics: 4 types with examples
The first one is descriptive. Descriptive is a snapshot in time. It’s common to do descriptive reporting as the first path to analytics. It leverages existing data collection. It ties into the historian and production info and figures out how to report either compliance or results. It’s designed to provide information and insights on meeting the compliance requirements or the overall results of production.
The next type of analytics is diagnostic. It’s a real time visualization and digs into the troubleshooting aspect of your process and your problems. It fits in the common dashboarding that most use in facilities to measure productivity. It’s an emphasis on the key performance indicators (KPIs) in the moment as things happen. It’s organized based on processes or lines or various components of systems. It provides a path to drill into the data and do a deep diagnostic dive into why things are occurring.
The third type of analytics is predictive. Predictive analytics are looking to optimize based on real-time feedback. Most everybody’s familiar with the time-based maintenance approach that happens in most facilities and factories today. The idea and the goal is to move scheduled maintenance more towards a predictive structure, which is the desired state. It requires a bunch of physical information on the assets to manage and improve the maintenance of. It comes from a lot of different locations. It’s mostly a lot of sensor data, things like temperature and vibration. A lot of that can be collected via Internet of Things (IoT) sensors that can grab the information from wireless networks and various different protocols on communication.
Videos provide a way to see what has happened in the past and what’s happening in the present and compare the two. People often miss traditional I/O when looking at predictive analytics. Multiple sensors in a provide a wealth of information to leverage without actually leveraging all the IoT. There’s a path to do more. After gathering all that information, it’s possible to do math and model it based on the feedback of failures and when failures occur. From there, it’s possible to alert and modify in a predictive model and then look at integrating maintenance systems. That provides ability to schedule preventive maintenance or predictive maintenance based on those alerts, rather than a predetermined timetable.
The last example of analytics is prescriptive. The idea of prescriptive is to optimize production based on the past. In one case, a customer had a process that required a changeover from product to product. In the best possible world, it wouldn’t require a wash or a tooling change. The next best situation would be if it only needed to be washed, followed by a lot of different data that exists in the facilities already. They had existing manufacturing execution system (MES) software that was providing stock-keeping unit (SKU) and other information.
In the case of this prescriptive model, information from the MES database could be run through some algorithms and some analytics applications and come up with a recommendation report based on those SKUs and then hit optimize in the scheduling department.
Industrial data analytics were able to do that. Analytics provided the insight to be able to set up schedule for the day, week, month and year.
– Edited by Tyler Wall, assistant editor, CFE Media and Technology, firstname.lastname@example.org, from a Control Engineering RCEP/PDH development hour webcast.
Industrial data analytics types
Have you used industrial data analysis to advance how information is used in processes?
Do you have experience and expertise with the topics mentioned in this content? You should consider contributing to our CFE Media editorial team and getting the recognition you and your company deserve. Click here to start this process.