October 26, 2017
By: Lyat Avidor Peleg
The full potential of combining sensor data and automated expert knowledge is yet unknown
In a unique interview, Dr. Hila Nachlieli, Head of Research at Precognize, reveals what motivates her, how Precognize is making the world a better place, and why women should practice sports
Q. What is your role at Precognize?
A. Precognize has a unique combination of both sensor data and automated expert knowledge. We have the unique ability to transfer experts’ knowledge into an automated model within a period of two weeks, which makes it possible for us to obtain this new type of information. The full potential of this combination is yet unknown. As the chief researcher, my role is to identify the benefits that can be harvested from this unique combination.
One such benefit is overcoming the “no free lunch” theoretical limit, which enables us to harvest the data beyond traditional data mining, cracking yet unsolved challenges. True predictive maintenance monitors the system as a whole, since one-sensor detections are bound to miss the non-trivial problems. In systems with thousands of sensors, the number of possible states is bigger than the estimated number of atoms in the universe. In such a big state-space it is unlikely that a specific state occurred in the past, and differentiating a normal state from an anomaly is prone to error, which is a challenge we can overcome. Another important benefit is our ability to accumulate many anomalies into a few informative issues.
Q. What is your expertise?
A. My expertise is machine learning and data mining. My PhD is in physics, in Non-Equilibrium Statistical Mechanics. My advisor was Prof. Dov Levine, who provided the mathematical explanation for the development of quasicrystals. I then worked at HP Labs Israel, focusing on image processing and machine learning.
Q. How do you see Precognize’s technology changing the process industry?
A. The aspect which is most important to me is preventing accidents: zero casualties in system malfunctions or explosions, a pollution-free environment, and the reduction of related health problems. With continuous predictive maintenance, equipment will last longer, which is the ultimate recycling, as there is no need for replacements. Moreover, the more you are able to trust your equipment, the less backup equipment you need and the leaner your factory is, and regions that were once occupied for redundant facilities can host trees and foliage.
Precognize provides the quality alerts required for predictive maintenance in complex factories, but full integration of its capabilities by industry goes far beyond predictive maintenance: imagine a compressor that gradually stops to avoid pressure building up to a possible explosion, as the following valve in the pipeline behaves suspiciously; or a tap that closes automatically because of a detected danger of oil leakage in the pipe; or a forge that cools itself because something in the input flow is just not right.
Precognize’s real strength is beyond focused alerts – it is in its ability to automate expert knowledge and build a computer model of the facilities. Advanced use of these models borders on science fiction, such as the automatic design of machines, a factory that changes its manufacturing lines to address changes in product specification, or to adapt the production mix to changing product prices.
Q. Do you see Precognize’s technology being applied to other industries? If so, which?
A. Sure. Precognize technology is general, and can be applied to any system. Precognize’s focused alerts and short integration time are crucial in complex systems, but they are beneficial in simpler machines as well, as it replaces the long customized development of a specific alert system or rule design. Instead of building an alert system from scratch, all the customer needs to do is to insert the expert knowledge into the Precognize system, which will take far less than two weeks for a simple system. On the other side of the scale, it would be highly beneficial, though far from the company’s focus, to apply the technology to the ultimate machine: the human body.
Q. Where do you see the predictive maintenance industry ten years from now?
A. In ten years, maintenance will include small adjustments and tunings, yielding zero accidents, zero casualties, and a pollution-free environment. Precognize will be embedded in the factory work routine, structures, and machinery, resulting in increased security. Machines with potentially life-threatening problems will not start, and planes with a significant alert will “refuse” to take off, until the problem is fixed, which will be fast because the problem has already been identified.
Q. And on a more personal note, how has your experience been as a woman in traditionally male-dominated STEM?
A. There are semi-cultural differences between a man and a woman, where women are much more likely to think and talk in “we” terms than men. This makes men’s contributions much more visible, recognized and compensated, which serves as feedback for women. As women adapt to the male-oriented culture, we learn to suppress our collaborative tendencies, which is a big miss. One thing I like about Precognize is that we are a one big team, where each person has his or her unique role and expertise, and it is the combination of all these qualities that enables us to act on important goals, such as helping to reduce the pollution in our city.
Q. What would your advice be to women who are interested in getting into STEM?
- Learn mathematics. Besides being the core of the STEM subjects, it will teach you to exercise perseverance and dedication to the completion of challenging missions.
- Practice sport and take part in competitions to experience a competitive environment.
- Your obligations are just as important as those of your spouse, and you are both equally obligated to your kids. Make sure you fulfill all of your obligations, and do not hesitate to hire professional help.