Predikto Streamlines Operations with Predictive Analytics
Friday, October 9th, 2015
When a large European railroad network needed assistance in predicting the probability of train delays caused by mechanical malfunctions, they contacted Atlanta-based Predikto, Inc., a predictive analytics software company.
Predikto deployed MAX, an intuitive self-learning artificial intelligence solution that retrieved large amounts of data from the sensors of the high-speed trains. MAX created thousands of algorithms from the raw data that were then used to identify and predict patterns of delays and mechanical performance issues.
“The high speed train operator in Europe wanted to know when failure was going to happen,” said Mario Montag, CEO of Predikto, “and they needed five to seven days warning of when the equipment would break. The five to seven days of warning is enough time to create a process to send the technicians when the train was scheduled to be in the depot.”
Montag explained that the delays and decreased performance caused from equipment failure would occur without warning. These unplanned delays are costly and can be related to safety, customer experience, reactive maintenance, and much more. Bottom line, if a commuter train is inoperable or operating at a reduced capacity, then these delays and the cascading effects often result in loss of revenue, reduced customer satisfaction, and possibly safety issues.
Montag said within the transportation industry there are expensive pieces of equipment that have sensors that measure hundreds of variables like the temperature or vibration of the motor.
“Train sensors come installed from OEM manufacturers, but the data seldom leaves the trains,” said Montag. “Some warnings were there for the operator to read. Now when the train reaches the maintenance depot, they can connect these sensors to the computer and data flows out.”
Designed by Predikto, the MAX system creates algorithms based on real-time sensor data, past maintenance records, and other historical data.
“Now people have access to what is happening live, and this is called the Internet of Things (IoT), and it allows the physical world and digital world to connect,” said Montag. “I now have data that provides insight into what is happening with a physical piece of equipment. We also have access to this rich sensor data and what will we do with it?”
Montag said this is where predictive analytics comes into play and where the use of artificial intelligence (MAX) is able analyze the data and provide insights regarding the components causing the delays.
Highly accurate predictions allow the railroad company to triage maintenance and prevent unplanned downtime and the negative side effect of delays.
Predikto has use cases where they are able to predict more than 98 percent of monitored component failures up to two weeks in advance and with a high degree of precision.”
Montag explains that with rapid advances in technology and faster dissemination of information from devices and sensors, companies now have access to real-time information regarding equipment and other conditions.
According to a report by CSC, an IT services and solutions company, only 80 percent of data captured is stored, and only 3 percent of the stored data is prepared for analysis. Even more astounding, only 0.5 percent of the data is being analyzed and then operationalized.
“We now have faster and easier ways to push the data out and companies are sitting on so much data they are frozen, and that is where Predikto comes in to help,” said Montag. “What does the company want to know? If it is hydraulics breaking in a plane, we can configure the software to process that data from the plane and give predictions three days in advance.”
Predikto Collaborates with New York Air Brake
New York Air Brake, an innovation leader and supplier in the rail industry since 1980, is working with Predikto to integrate a new predictive analytics component to its Advanced Train Control Technology LEADER (Locomotive Engineer Assist/Display & Event Recorder).
Predikto’s MAX system will pull LEADER train data, in addition to capturing external data from the train including weather and line of road conditions. The ability of MAX’s to adapt to rapid changes in near real-time will allow for the most accurate forecasts possible across a variety of use-cases.
“On the trains there are LEADER boxes, similar to a black box on a plane, which captures a lot of data from the trains,” Montag said. “It was designed as an autopilot and the LEADER box can tell the train operator how to basically drive the train, at what speed, when to brake, etc., in a way to maximize full efficiency. There is so much data that is gathered from the LEADER boxes and they [New York Air Brake] were only using a small percentage, so they have partnered with Predikto to build applications on top of their LEADER box to use the data in different ways.”
A Leader in Predictive Analytics
A relatively new company, Predikto was founded in 2012 and was recently named as one of the five Atlanta Startups to Watch in 2015 by Venture Atlanta.
Montag said he started Predikto with a “simple website and few slides” and his first client was a large steel manufacturer.
“This company had some hydraulics system failures and wanted me to determine when they would fail next,” he said. “I went searching for a partner and hosted a competition and received more than 50 applications from global data scientist.”
One contestant, Robert Morris, Ph.D., an associate professor of criminology at the University of Texas at Dallas, won the competition and became a co-founder. A few months later, David Bettinger, an application designer with Gulfstream Aerospace, joined Predikto as a co-founder.
“We are growing—starting with three and now more than 20 employees,” said Montag. “What we are doing is incredibly cutting edge.”
Montag cautions that predictive analytics is not the complete answer but another tool for companies to use to improve operation and efficiency.
“We do not replace maintenance people or operations staff,” said Montag. “We are just empowering them with additional information to be more proactive. I think predictive analytics will be deployed slowly, and those companies that start early and work with a company like Predikto will see success and have an impact on their business.”