With the rise of digital, manufacturers are finding themselves rich in data. Meanwhile, computing has emerged as the cheapest, most abundant resource that we can deploy against any problem.
The problem in manufacturing is not the lack of new ideas and products, but the ability to design and build new products efficiently. A 2015 IDC Big Data user study found that operations related processes were the top priority for analytics investments.
This next wave of IT innovation, the rise of digital, is providing manufacturing with the engine to improve efficiency. IT has become an integral part of a product. This is because of cheap sensors and processors, cheap storage, purpose-built software, purpose-built clouds enabling data storage and ubiquitous connectivity.
Simulating new innovations is the idea behind the digital twin in manufacturing. We can use stochastic simulation to generate future “what if” scenarios and use those scenarios to avoid costly product quality issues, speed time to market, and increase throughput.
This may sound exotic, but it is really just a modern twist on a very old idea — the scientific method. Build stochastic simulations, generate experiments, and use those experiments to minimize risk and innovate in the process.
Tesla is an excellent example of this concept. Tesla has a digital twin of every VIN they manufacture. Data is constantly being transmitted back and forth from the car to the factory. If a driver has a rattle in a door it can be fixed by downloading software to adjust the hydraulics of that particular door. Tesla regularly downloads software updates to their customer’s cars based on the data they are constantly receiving from each VIN.
Figure 1: Using the digital twin as a source of manufacturing insight
Prescriptive Data, and Pipelines
Creating a digital twin starts with establishing new pipelines of manufacturing data. We can automate the collection, for example of materials and design data. When integrated with historical operations performance data, we have the raw data required to support the creation of a digital twin.
The next step is to take the manufacturing process and model it using rules. But instead of using the more common retrospective models, digital twin simulation uses prescriptive models. Retrospective models, like those commonly used in predictive modeling, try to calculate the future based on past trends. Models like that have been successful in some areas of manufacturing prediction, but they take us away from breakthrough innovation and keep us stuck in optimizing. Instead, we need to build stochastic simulations, or prescriptive models. We create rules for mapping from design to performance and add randomness to simulate risk.
The prescriptive data from the simulations tells us how new products will work. We can detect design flaws early. We can predict and minimize cost. Because randomness is inherent in our models, we can simulate the kinds of uncertainty we encounter in the real world. Computer power is cheap and we can afford to run millions of scenarios. We can anticipate an entire spectrum of possible outcomes rather than just a single expected result.
Continuous Insights through IoT and Industrial Machine Learning
We can learn as much from the digital twin as we can from the real-world original. Internet of Things (IoT) technology allow us to augment the manufacturing process with sensors and automatically generate data about operations, performance, and maintenance. If we use Industrial Machine Learning to build and deploy, we can turn the streaming variant of the digital twin in to a continuous source manufacturing insight.
Figure 2: When deployed according to CSC’s Industrial Machine Learning, the digital twin becomes a continuous source of manufacturing insight
Digital twin really sits in the continuum of the IoT. If we agree that the foundation of IoT is connectivity, sensors and analytics, predictive maintenance is an established IoT application. Predictive maintenance is case-based reasoning enabled by data for mitigation and repair. Digital Twin incorporates product data from design to operation and beyond, including maintenance history. Harnessing all the data to enable a complete digital twin isn’t there yet. But there are examples and pilots showing the steps along the way are certainly relevant.
The Digital Twin Comes to Life
This idea is beginning to take hold in several major manufacturers. GE is piloting a “digital wind farm” concept, which it uses to inform the configuration of each wind turbine prior to procurement and construction. Once the farm is built, each virtual turbine is fed data from its physical equivalent, and software enables optimization of power production at the plant level by adjusting turbine-specific parameters, such as torque of the generator or speed of the blades. The hope is to generate 20% gains in efficiency.
PTC has developed a “Smart Connected PLM” software product called “Windchill”. The Swiss solar panel manufacturing company Oerlikon uses Windchill to automatically track system metrics and keep account managers apprised of the condition of their customers’ systems. PTC describes it as a FRACAS process or a failure reporting, analysis and corrective action system.
Dessault Systemes has built an aerospace and defense specific manufacturing operations management product called “Build to Operate”. It provides the ability to monitor, control and validate all aspects of manufacturing operations—ranging from replicable processes and production sequences, to the flow of deliverables throughout their supply chain—each on a global scale. Airbus Helicopter has deployed this system for current and future helicopter manufacturing.
In every industry, we see the increase of data with both higher velocity and volume. This is leading to the creation of more machine learning algorithms designed to learn incrementally over new data. In manufacturing, these new algorithms will take the form of digital twins capable of helping manufacturers design and build new products more efficiently.
What do you think? Are there any compelling manufacturing products or processes that would benefit by creating a digital twin?
JC Brigham — CSC ResearchNetwork Analyst
Joan-Carol (JC) Brigham has been an analyst within CSC’s ResearchNetwork for 8 years. She has lead strategy work and managed much of the start up of industry research within the ResearchNetwork. Right now she is a Principal and Business Manager analyzing the Manufacturing industry. Prior to CSC she worked in the Services area at Sun Microsystems, and before that at a small digital marketing company. She stumbled into the high tech market analysis profession during her 15 years as an IDC analyst. JC lives in the mountains of Colorado, loves the outdoors and travels as much as she can.
Jerry Overton — Distinguished Engineer
Jerry Overton is head of Advanced Analytics Research in CSC’s ResearchNetwork and the founder of CSC’s FutureTense initiative, which includes the Predictive Modeling Research Group, the Advanced Analytics Lab and the Predictive Modeling School.
See Jerry’s bio.