Simulation has long been used to improve the design of nearly every physical product or process by providing the opportunity to evaluate a wide range of alternative designs prior to building physical prototypes. Simulation has also long been used to model different operating scenarios to develop control strategies that are incorporated into control algorithms to improve operations. The emergence of the Internet of Things (IoT) has created the potential for a transformational journey in which a simulation model of the product or process is tied through the Internet to sensors capturing data and to actuators controlling its operation. The result is a so-called digital twin of the physical product or process that can be used to analyze and diagnose its operation and optimize its performance and maintenance in real time. By using simulation in conjunction with the IoT, companies can analyze the performance of products in real-world operating conditions and make confident predictions about future performance to improve product operation and productivity, and to reduce the cost and risk of unplanned downtime.
Emergence of the Internet of Things
The IoT is changing the way companies approach the entire product lifecycle from development, testing and manufacturing to operations and maintenance. Smart connected products leverage connectivity with the cloud and other devices to deliver unprecedented functionality. The proliferation of smart connected products offers exciting new capabilities for their users and enormous opportunities to companies developing them. There are already more smart devices connected to the internet today than there are human beings in the world. IoT devices are creating massive opportunities for existing businesses and giving rise to brand new markets and companies. The potential economic impact by 2025 has been estimated to be up to $11 trillion per year.
Early IoT applications focused on relatively simple applications, such as determining the state of assets and issuing simple commands, such as turning an asset on or off. But manufacturers and users of smart connected products are demonstrating that the power of the IoT can be increased through integration with simulation technology. Simulation can perform diagnostics and troubleshooting in real time, anticipate breakdowns and determine the optimal point to perform maintenance, tune the product to optimize its performance, and capture information that can be used to improve the next-generation design.
Role of simulation in the IoT
Industry leaders use simulation to create complete virtual prototypes of complex products and systems comprising mechanical, electronics and embedded software components, incorporating all the physical phenomena that exist in real-world environments. For example, computational fluid dynamics (CFD) software is used to model and predict fluid flow, which is critical to optimizing the efficiency of so many products and processes, ranging from the combustion of gases in an automobile engine to the movement of a chemical solution through pores in a shale gas formation. Structural analysis software is used to predict how a product will react to forces, heat, electromagnetic fields, abrasion and other physical effects to determine whether a proposed design will meet design requirements. Electromagnetics simulation predicts the signal integrity, power integrity and thermal integrity of products such as computer chips, circuit boards, cell phones, automobile electronic components and entire communications systems, making it possible to quickly optimize the design without wasting time building and testing costly prototypes. Furthermore, software engineers leverage software development tools and certified code generators to ensure the high level of quality needed to prevent product failures in the embedded software that is increasingly being used to oversee and control the operation of many products.
Simulation is also used to design simplified reduced order models that are embedded in control algorithms to manage the operation of automobiles, power plants, machine tools, printing presses, chemical reactors and others. As an example, let’s look at the difficult challenge of operating electric batteries in electric vehicles (EVs) and hybrid electric vehicles (HEVs). Batteries provide the primary drive power for the vehicle as well as power for numerous electrical auxiliary systems. The operation of the battery must be carefully managed to avoid overheating, which reduces power-generating efficiency and shortens battery life. Engineers typically employ an air cooling strategy for cylindrical cells that use housings shaped for optimal cooling and a blower and guiding vanes to provide adequate airflow. For rectangular cells, cooling generally is done with liquid circulating through heat exchanger elements in contact with the cells. A control algorithm varies the loads on different cells based on temperature readings and the charger status.
Simulation is typically used to evaluate and optimize the thermal performance of the cells under a wide range of conditions. Fluid dynamics solvers are used to analyze the complex 3-D cooling flows and conjugate (two way solidfluid) heat transfer. Engineers use electronic circuit simulation technology to evaluate control algorithms that manage the thermal performance of the battery while also preventing overcharging, high-current charging/ discharging, external shorts and other electrical problems that could reduce battery life and risk battery explosion. Engineering simulation is ideal for designing such algorithms due to its ability to tightly integrate 3-D physical models (fluid dynamics and mechanical) into the control circuit simulation.
Methods such as parameterization and design of experiments are used to identify the best control system parameters at each set of operating conditions. The simulation results are then used to generate reduced order models that are incorporated into the control logic used in a battery’s electronic control units (ECUs).
Advent of the digital twin
The IoT connects simulation to the product or process in near real-time, justin- time or in replay mode to aid in operating and maintaining the product or process. The simulation-based digital twin concept incorporates the physical product or process, the simulation models and the connections that facilitate communications between the two. The digital twin may consist of a simulation model that has been developed to duplicate the current condition of the product or process, such as by incorporating wear or degraded performance into the simulation model. The data from sensors connected to the product or process may be used to provide real-time
boundary conditions for the digital twin. The digital twin results can be calibrated based on the operation of the actual product or process. These enhancements to the digital twin can improve its predictive capabilities far beyond what can be obtained in the product design process. The predictions made by the digital twin can be used to determine the root cause of performance problems, evaluate results of different control strategies, determine optimized maintenance schedules, etc. The digital twin can also provide information about the product or process that cannot be measured with sensors, such as flow velocities through internal passages. The result is that digital twins can be used to substantially increase the performance and reliability of the product or process while reducing its operating cost.
Pump application demonstrates digital twin concept
ANSYS recently demonstrated how a simulation model can serve as a digital twin to process sensor data generated by an instrumented product or process to predict failures and diagnose problems so that action can be immediately taken over the IoT to fix problems and optimize performance. The demonstration showed a motorized pump operating in a hydraulic system with valves on the suction and discharge sides. The motor and pump were instrumented with sensors to measure key operating parameters such as mass flow rate pressure, vibration and current draw. Actuators on the valves were used to control their operation based on instructions generated by evaluation of the simulation model.
In the recent demonstration, a machine learning algorithm predicted the number of operating days left until failure. Then an operator introduced an anomaly by closing the suction valve to 50 percent. The sensor readings on the physical product immediately indicated that something was amiss. For example, inlet pressure, outlet pressure and flowrate through the pump decreased drastically, while pump noise rose to higher than normal values. But the sensor readings provided minimal diagnostic information, and it was not possible to look inside the pump and see why it was vibrating. Furthermore, the sensor readings provided little or no help in determining what would have happened if various actions were taken to solve the problem.
So the digital twin was used to address these challenges. The sensor readings from the demonstration unit were transmitted over the internet and used as boundary conditions for the system and component level simulation models. The simulation model immediately began exhibiting the same symptoms as the physical model, such as reduced pressure and flowrate. While engineers could only view the physical product from the outside, the digital twin enabled them to look inside the virtual product and understand what was going on. The digital twin showed that the fluid in the interior of the pump was cavitating. The drop in pressure inside the pump was forming vapor cavities – essentially bubbles – where the pressure was low. In locations where the liquid was subjected to a higher pressure, the voids imploded and generated noise. Next, engineers used the digital twin to evaluate the impact of changing the operating conditions. After evaluating the effect of different valve settings, they determined that opening the suction valve to 100 percent would restore pressure and flowrate to normal levels.
The digital twin concept is being extended far beyond this simple example to encompass large and complicated assets such as petroleum refineries, automobile assembly plants, distribution centers, wind farms, large scale construction projects, etc. In each case, sensors and other devices capture data that is fed into the simulation model to provide a detailed understanding of the current state of the asset. Machine learning algorithms running on the edge or in the cloud access information from the physical asset and the simulation model to optimize the performance of the asset by scheduling maintenance, setting control points, sending alerts to operators, providing reports to management, etc. The information generated by the simulation model can also potentially be communicated by overlaying the flow contours of the fluid flow inside the pump onto an image of the pump so operating personnel can quickly understand and diagnose problems.
The simulation-based digital twin will help companies analyze smart machines in real-world operating conditions and make informed decisions that will improve their performance far above what is possible today. Physics based and system simulations with big data analytics and industrial devices augmented with embedded intelligence can reduce risk, avoid unplanned downtime and speed up new product development. The resulting efficiency and productivity gains will have a dramatic effect on an organization’s bottom line, as well as on the global economy. The combination of machine connectivity with a data lifecycle management platform powered by engineering simulation will enable organizations to perform diagnostics and troubleshooting, determine the ideal maintenance program based on the characteristics of the individual asset, optimize the performance of their assets, and generate insightful data that can be used to improve the next generation of the product.