Fusion Connect IoT Platform to Embed Artificial Intelligence Modeling Engine

A.I. software CTO discusses engineering benefit of Eureqa analytics tool in Autodesk Fusion Connect.

Eureqa extrapolating data based on a predictive model. (Image courtesy of Autodesk and Nutonian.)

Eureqa extrapolating data based on a predictive model. (Image courtesy of Autodesk and Nutonian.)

Engineers connecting their devices and industrial equipment through Autodesk’s Fusion Connect Internet of Things (IoT) platform will soon be able to utilize an artificial intelligence (A.I.) modeling engine to perform predictive analytics on their big data.

The Eureqa software, from Nutonian, sifts through the sea of big data to find predictive models which engineers can then utilize in their efforts to optimize products, workflows and industrial processes.

“Engineers have two main challenges when leveraging their industrial IoT (IIoT) data. First is collecting, formatting and dashboarding that data into a usable and analyzable format. This is where Fusion Connect comes in,” said Michael Schmidt, CTO at Nutonian. “The second challenge is using those numbers and sensor values for alerts or analytics. Combining Eureqa and Fusion Connect makes it easy for engineers to perform this task. You don’t have to be a data scientist.”

So, what makes Schmidt’s software key to an engineer’s workflow? And how does it differ from all the other IoT analytical tools flooding the market? According to Schmidt, his analytical tool targets the IIoT market, and goes beyond the ability to automate tasks.

“A lot of people are using analytics to automate tasks, but we are using them to interpret things, to find causes and effects and give transparency to see why things will happen,” Schmidt added. “Engineers can use it to perform computational analysis, but the added benefit is higher accuracy of incoming data. Additionally, engineers can extrapolate data in a new regime.”

Another engineering challenge that Eureqa aims to solve is the difficulty of assessing the time series data that engineers face when working with production lines and equipment with multiple sensors. For instance, say that an operator changes the position on a valve. The effects of this valve aren’t instantaneous, like flipping a light switch; sometimes, the overarching effects can take hours to make themselves known.

However, Eureqa attempts to find precursors to events, like that of the valve example. For a human to find these causes and effects on the grand scale of a production facility would be exceptionally difficult.

How Eureqa Works with Fusion Connect

Eureqa is designed to automatically find connections in data that a human wouldn’t be able to see. The software creates multiple permutations of predictive models, then tests to see how well these models describe a data set based on the accuracy and simplicity of the model. “Even if you don’t know the question your data might answer, Eureqa will give you a good starting point,” Schmidt said.

With the creation of Eureqa, Schmidt has made a name for himself as one of the “World’s Top 7 Data Scientists” according to Forbes, and as one of the “35 Innovators Under 35” according to MIT. As a result, it’s clear why Autodesk chose his software to integrate into their IoT platform.

“I wanted to use A.I. to help scientists discover how things work, to see the impact of variables on a system and how to generate a hypothesis and test it,” said Schmidt.

Engineers can task Eureqa to find the core relationship between the data collected by their products over the IIoT. The first order models created by the software can then be transferred, via application program interfaces (APIs), to other tools such as MATLAB or system simulation software for further analysis and system optimizations. As the software is embedded into Fusion Connect, the model can immediately be used to set up dashboards, alerts and alarms based on the IoT data collected by the system.

The general workflow of the Fusion Connect Eureqa integration will include:

  1. Collect data from your IoT system. This will center around Fusion Connect collecting data from the IoT devices and sensors. The IoT platform will then mould the information into an actionable data set. Alternatively, the data set can come from MATLAB, Excel or other Statistical Analysis Systems (SAS).
  2. Engineers access Eureqa directly within Fusion Connect. Eureqa will automatically assess the data to see what predictive models can be created for the given set of variables.
  3. Users choose which predictive models they want the software to create based on the list of variables.
  4. Eureqa iterates through multiple models to determine which model will accurately and simply predict how the system works.
  5. Engineers use tools to visualize and interpret the model to assess its appropriateness.
  6. Once the model is verified, it can be shared directly with Fusion Connect. To share with other programs, engineers can use APIs.
  7. Engineers can then integrate the verified predictive model into the alert and dashboarding tools within Fusion Connect.

Since both Fusion Connect and Eureqa operate on the cloud, engineers will not need to invest in expensive high performance computing (HPC) hardware. This will help to keep the use of the technology affordable.

Autodesk estimates engineering firms can save a significant amount of money choosing predictive analytics software such as Nutonian over hiring a data scientist. (Image courtesy of Autodesk.)

Autodesk estimates engineering firms can save a significant amount of money choosing predictive analytics software such as Nutonian over hiring a data scientist. (Image courtesy of Autodesk.)

But don’t you still need a data scientist for all of this number crunching? Schmidt suggests that hiring data scientists will not be necessary, as the Eureqa software includes failsafe technology that helps users avoid garbage-in, garbage-out results. He explains that the software can detect a loss of data signal and the edge cases that a data scientist would traditionally handle.

“One test we have performed is giving Eureqa random data,” said Schmidt. “The answer that comes back is the software saying the best you can do is predict the mean; it tests whether the data you have is predictive or not. It can determine what is noise, and what is regular ordinary behavior. Eureqa also partitions data sets to see if the model has been exposed to over-fitting problems. It validates the results for you.”

Why Engineers Need Artificial Intelligence to Crunch IoT’s Big Data

“Engineers are using [Eureqa] to predict production line failure or how a design’s dimensions can influence the stresses and strains of a product,” said Schmidt. “The software allows engineers to characterize their manufacturing processes and product designs, which leads to stronger materials in those environments.”

In essence, optimization is the biggest application for engineers using this software. For example, take a manufacturing process. It will involve thousands of knobs, valves, switches and various other settings which can affect both yield and the rate of throughput.

By bringing this production facility onto the IIoT through Autodesk Fusion Connect, engineers can collect data on these various input settings and output. This data can then be sent to Eureqa and used to determine predictive models that will link the input and output.

What about those engineers that already know their operation like the back of their hand? How could Scotty possibly learn anything more about the Enterprise when he can feel the warp speed in the rattle of a floor panel? The advantage for these engineers is that they can input their own prior knowledge into Eureqa before the program creates its statistical model. The software can then optimize the model and help these engineers discover even more hidden quirks in their operations.

“In the IoT world, we are helping people collect new information from sensors,” said Schmidt. “This creates high dimensional data sets. It could include data from hundreds of thousands of things at a time. So how do you know one sensor is more important to another? What sensor predicts a certain outcome? We can help you scale to that level with the data based on our results.”

Unfortunately, Eureqa isn’t able to perform the next step, optimization. Schmidt explains that this functionality is in the works, but for now he suggests using tools such as MATLAB to discover the maxima and minima. However, without the model Eureqa creates, this optimization study would be difficult to perform by hand.

Eureqa can also be useful to IoT product designers and engineers. For instance, it can help them find alternative use cases for their products. The software performs this task using anomaly detection algorithms.

Since the software maps out a predictive model for the use of the product, the software can then determine when there is usage data that deviates from said model. Once the engineers and product designers investigate the anomalous use case, they can optimize the design of their product to this niche market.

To learn more from Schmidt’s data analytics genius, watch the webinar: How Engineers Can Leverage A.I. and IIoT Data to Automatically Optimize Design.

Written by

Shawn Wasserman

For over 10 years, Shawn Wasserman has informed, inspired and engaged the engineering community through online content. As a senior writer at WTWH media, he produces branded content to help engineers streamline their operations via new tools, technologies and software. While a senior editor at Engineering.com, Shawn wrote stories about CAE, simulation, PLM, CAD, IoT, AI and more. During his time as the blog manager at Ansys, Shawn produced content featuring stories, tips, tricks and interesting use cases for CAE technologies. Shawn holds a master’s degree in Bioengineering from the University of Guelph and an undergraduate degree in Chemical Engineering from the University of Waterloo.