Jason Rios, Senior Vice President, and General Manager
Coupling cutting-edge simulation technologies with AI is a game-changer in analyzing the performance of mechanical systems and prototype designs. It offers companies a golden opportunity to reduce the substantial cash and time consuming efforts associated with building physical prototypes for every mechanical system design. When it comes to system maintenance, simulation can demonstrate the functional how’s and why’s but AI can enable predictive maintenance.
When combined, simulation and AI reveal possibilities for improvement while helping prevent the undesirable.
For industry verticals that rely heavily on mechanical systems, such as aerospace and engineering, this combination would be the Holy Grail and Sentient Science delivers just that.
AI-powered Physics-based Modeling & Simulation
Sentient Science’s flagship physics-based modeling and simulation technology, DigitalClone, enables users to predict the life of mechanical systems. Sentient Science has partnered with Amazon Web Services (AWS) to make its complete cloud solution accessible through a web browser. Whether it is an OEM trying to evaluate the life impact of various design options or an operator forecasting the 18-24 month demand for critical components, the team at Sentient Science believes that the physics that powers the DigitalClone software will fulfill the long-anticipated promise of the ‘digital twin.’
Digital twins are essentially virtual surrogates for physical systems, and organizations like the US Department of Defense and other advanced users are racing to incorporate them into the sustainment enterprise. Many industry solutions have focused on machine learning and AI models that consume vast amounts of data to create these surrogates. Sentient Science approaches it differently. They offer a physics-based approach that doesn’t just anticipate on ‘when’ systems will start to fail but also anticipates ‘why’ and ‘how’ those failures will occur.
To Sentient Science, aerospace digitalization entails more than just computerizing existing processes and technical data. The company adopts a different way of thinking, one that truly embraces innovative technologies to find more powerful and efficient ways of operating. While data is imperative to drive AI application effectiveness, there still are instances where the data is insufficient or not yet available. Take, for instance, next-generation technology programs that may have not yet generated ample data for analysis or even typical performance updates. Such instances are not uncommon for mechanical systems infrastructures.
“Those scenarios require a first principles approach that considers the physics of the design, the characteristics of the materials, and the operational loads that the system is expected to withstand.
The ability to effectively support not only mature, data rich environments but also next-generation technology exploration is what sets DigitalClone apart from other solutions,” mentions Jason Rios, Senior Vice President and General Manager, Sentient Science.
Brilliant Nuances of a Winning Technology
Rios explains that mechanical systems break when materials fail—a behavior that can be predicted by using advanced physics-based modeling and simulation to analyze the microstructure of the underlying materials and its ability to withstand operational stresses. “We start by analyzing the system to determine the critical loading areas, and then we do a deep dive into those parts of the design to predict the probability distributions associated with the initiation of damage – the first microcracks that signal the beginning of the failure,” explains Rios.
Sentient Science’s computational testing is based on numerous proprietary technologies developed over the course of several years, funded in part by US government grant programs with a strong emphasis on the aerospace drive train industry. These technologies have been codified into advanced computational and analytical software tools in material microstructure quantification, traction models, and damage initiation predictions—all leading to predictions of certain failure mechanisms. The tribological and damage mechanics techniques are particularly applicable to gears and bearings, demonstrating a combination of rolling contact fatigue and wear due to the inherent contact forces and sliding behaviors, respectively.
For these scenarios, Sentient Science’s Component Life Prediction (CLP) entails six key processes: determining component hotspot, building material microstructure models, creating surface traction models, computing material microstructure response, calculating the time for mechanical failure, and predicting fatigue life distribution.
Sophistication Encapsulated and Simplified
DigitalClone offers an intuitive user interface for assembly and specification of a system for analysis. In addition, insightful, interactive three-dimensional visualizations of assembled systems assist in validating the model as well as understanding its behavior. Moreover, system-level analysis is complemented by detailed contact analysis for gears and bearings. Contact stresses, sliding velocities, and, in the case of gears, tooth bending stresses are computed and provided to the user in interactive visualizations. Finally, traction and fatigue life analyses are conducted while sampling the full random space of uncertainty to capture the spread in expected failure times and failure modes of the component under the prescribed loading conditions. “Although we aren’t suggesting that using DigitalClone eliminates the need for physical testing during qualification, it has clearly shown the ability to accelerate rapid prototyping and enable more efficient testing. We’ve seen cost savings in the 15-35 percent range, and schedule acceleration of up to 65 percent in the pre-qualification phases,” says Rios. The company enables faster time to market for new designs, inventory reduction, increased asset reliability/availability, and brings other outcomes that align with their customers’ objectives.
In this regard, Sentient Science has participated in a number of advanced development programs for the US DoD, working with OEM experts to evaluate new materials for gears/bearings and quantify the life benefits of proposed design changes (such as superfinished gears). The value that DigitalClone brought to each of those programs really comes down to two things: cost and schedule.
DigitalClone brings in the best of both worlds. Most of the market is heavily focused on Big Data solutions, which take the vast streams of data generated by fleet operations and use advanced algorithms to predict the trends. While Sentient Science has that capability, it is not their primary means of predicting the health risk of systems like a helicopter drive system.” The problem with Big Data solutions is that they rely on the output from sensors and accelerometers to drive predictions, and when those outputs reflect a ‘normal’ condition there is not much insight to be gained about the future failure risk – you’re not able to look ahead and forecast the point at which your usage is likely to cause damage to initiate,” asserts Rios. When there is an abnormal vibration or a temperature increase, something is already failing, and the operator has limited options at that point to manage the life of that system. DigitalClone looks ahead at a much further horizon, taking into account the usage profile of a system and forecasting the long-term changes to the health risk that will eventually result in gears cracking or bearings spalling.
"We want to enable our customers to see the failure coming, long before it actually presents itself and is detectable by sensors"
“We want to enable our customers to see the failure coming, long before it actually presents itself and is detectable by sensors. That is the kind of insight that a fleet manager needs. That is not to say that there isn’t value in the Big Data approach. We do this ourselves in industrial applications such as wind energy, where we have thousands of assets integrated into our DigitalClone framework – pulling in terabytes of data that we analyze using advanced AI and machine learning tools to improve our understanding of current state and refine our long-term forecasts,” adds Rios. “The physics-based and AI-enabled approaches are complementary, and both provide essential insight to system health – they just focus on different time horizons.”
The Innovation Continuum
Having carved their niche in the space, Sentient Science is actively exploring funding opportunities that will trigger the next phase of their growth and redefine the approach to mechanical systems maintenance.
Sentient Science has recently accelerated a SBIR (Small Business Innovation Research) development program that will enable them to commercialize their world-class capabilities to predict the material microstructure, residual stress, porosity, and life expectancy of components built using additive manufacturing (AM). They anticipate it will accelerate design optimization by 75 percent and reduce the cost of qualification by up to 50 percent, compared to existing processes. “With the emergence of AM as a key enabler in the aerospace community, we have very high hopes for the market opportunities that will follow,” concludes Rios.