Imagine if a manager arrived at their desk on Monday morning and the computer dashboard flashed a message — a key component in the company’s production system has an 85 percent chance of breaking within the next three months.
 
Now, let’s say the manager clicked on the part onscreen and was able to view pictures of the component from multiple angles, data about throughput slowing, and information that indicates a problem. They would have everything needed to discuss ordering a replacement part and deciding the best time to schedule a repair. In the meantime, they could check to see if other maintenance tasks or repairs could be scheduled simultaneously to reduce downtime. 
 
This scenario is no longer science fiction; necessity has made this a requirement for many organizations. According to a Forbes Insights research report, 51 percent of executives surveyed have significant internet of things (IOT) programs in operation; 49 percent have programs in the early stages, including pilots.
 
“The IOT has come from nowhere to being a part of our industry in just two years,” Satyam Priyadarshy, technology fellow and chief data scientist at energy company Halliburton, says in the Forbes report. “This trajectory cannot be ignored — the digital architecture will become the norm in our industry and many others in a short amount of time.”
 
Companies can’t afford to have plants closing for inspections, maintenance, and repairs for weeks or months, but the alternative — surprise broken parts and sudden breakdowns causing zero output — isn’t affordable either.
 
Automating inspections and creating “digital twins” to create predictive models may be the best alternative. Avitas Systems, a GE Venture company, found that automated inspections can drop inspection costs by up to 25 percent.
 
Companies also experience additional benefits, such as increased personal safety (including reduced health risks) from toxic chemicals and hazardous situations; improved efficiency (up to a 15 percent reduction in downtime and 25 percent increase in inspection turnaround time), and improved product quality. 
 
Getting started
 
How does a company get started automating inspections? Depending on the organization, this can be a complicated venture that requires a lot of resources — money, time, staff, knowledge, and equipment.
 
Avitas Systems, launched last year by GE Ventures (the massive company’s startup and incubator arm) is one company working with companies in the oil and gas, aviation, electric power and rail industries to automate inspections, using the internet of things, artificial intelligence, machine learning and robotics.
 
The inspection services industry requires cutting-edge technologies to help avoid unplanned asset downtime and to deliver new, valuable insights,” Alex Tepper, founder and head of corporate and business development at Avitas Systems, said at the announcement of the company. “We will use state-of-the-art robotics, automated defect recognition and cloud-based technology to give customers the customized service and data they need to advance from reactive to predictive repair.”
 
To start, they meet a client to figure out what is needed on the technical front. Avitas Systems partners with vendors, inside and outside GE, who have the right hardware to achieve a clients’ goals and help hurdle one barrier to entry, which is identifying the right technology and knowing how to use it.
 
“The Avitas Systems Platform deploys next-generation, automated processing of time-sensitive industrial IOT data using advanced analytics, which are based on our complex machine learning and artificial intelligence capabilities,” says Alok Gupta, vice president of engineering at Avitas Systems.
 
Data access is the key to any work in IOT, AI and machine learning, as without data, companies can’t create models. Avitas Systems works to solve this by identifying data to use, including existing and third-party data; gathering new data with robots, drones and sensors; creating 3-D models and “digital twins” of important systems; and using the models for predictive maintenance.
 
For example, Avitas works with snake robots from the company Sarcos, which can fit through small pipes where humans can’t and carry a variety of sensors that feed back into the system. It partnered with Kraken Robotics to tie that company’s autonomous underwater vehicles and sensors into its Avitas Systems platform, creating new inspection options tailored to the oil and gas industries.
 
It also partnered with computer-maker Nvidia to use artificial intelligence to bolster inspections by enabling advances such as automated defect recognition. For industries such as the rail industry, that could be critical, as it allows them to quickly find problems in track alignment, broken rails, washouts, broken or missing crossties, or more.
 
It’s the data
 
Determining the key data elements that will provide actionable information is probably the most difficult aspect of measurement and predictive systems. Getting new data isn’t the only answer, as data collected from previous inspections, ongoing operations and maintenance schedules can also be valuable. 
 
Companies tend to reference this historical data over and over again because they know it, understand it and see value in it, according to Avitas. However, there may be some “dark data,” existing in virtual data warehouse silos, with no current relationship to other data points, which could be useful.
 
For instance, data about weather, ocean currents and wind currents may not be useful by themselves, but when correlated with other sources using AI or machine learning can provide insights such as the impact of weather or water damage on equipment. They can also help identify the causes of additional wear. 
 
“To date, companies have explored only a tiny fraction of the digital universe for analytic value,” Deloitte Technology Consulting officials wrote in an article last year. “IDC [International Data Corp.] estimates that by 2020, as much as 37 percent of the digital universe will contain information that might be valuable if analyzed. But exactly how valuable? IDC also projects that organizations that analyze all relevant data and deliver actionable information will achieve an extra $430 billion in productivity gains over their less analytically oriented peers by 2020.”
 
Data, modeling, predicting
 
New data is also valuable, and there are more ways to get it than ever before. If a company can put a sensor on it, they can measure it, even if a human can’t access it. In the past, most companies focused on gathering data that could be accessible through measuring equipment. With the accessibility of IOT through sensors and robots and the ability to leverage unstructured data, companies have expanded their perception of what data is usable. Now, they can use devices including drones, robots and sensors to measure sound, define visible wear through photos, or determine surface changes through measurement.
 
These devices don’t need to be built with a single purpose in mind; they can have alternative tasks, such as helping with routine maintenance. 
 
Once a company has gathered data and collected insights, they can create a model. Some will go so far as to create a digital twin, basically a 3-D model that helps them understand what’s happening with their systems. There are a number of cloud-based platforms that can help create these models. The Avitas Systems cloud-based platform is hosted on GE’s Predix (pronounced like “predicts”), which accommodates edge computing and can process the tremendous amount of data received during inspections.
 
Once a model is developed, the system can start doing predictive analysis, such as described at the beginning of this article. Previously, a company might infer this type of information about a pending component failure based on intuition, past experience, and accessible facts. Now, a system can accurately measure if there is a status change due to weather conditions, a change in air or water quality, or visible corrosion, and if something bad is about to happen.
 
According to GE, the use of sensors and AI helps industry move away from rigid inspection schedules based on regulations and focus on the actual health of a given asset.
 
However, the benefits of automated inspections go deeper with the digital twin concept and its ability to model system possibilities. 
 
For example, GE launched the Digital Wind Farm initiative to help improve productivity in that industry. GE studies the wind patterns at a given location and then calculates the best turbine configuration and site layout for it. The turbines are then linked to the virtual model, so operators can monitor performance, and potential problems, on a “unit by unit” basis.
 
GE says one U.S. customer for the system was able to increase its annual power production by 16 percent.