That’s right, machines don’t just analyse data but can talk to humans like other humans would. Except, as L&T Technology Services says, they can move from talking like a layperson to spitting the kind of intelligence Albert Einstein and Stephen Hawking are known for.
All because artificial intelligence (AI) has matured. From $272.5 million in 2016, revenues from AI in manufacturing are all set to shoot up to $4,882.9 million by 2023 at a CAGR (compound annual growth rate) of no less than 52.42 percent. Artificial intelligence (AI), machine learning, deep learning, and neural networks are driving technology innovations, the kind we’ve only imagined before. Deep learning techniques for instance are helping companies with signal recognition, data mining, voice and image recognition, while machine learning is enabling them to make sense of technological resources such as data from sensors and the Internet of Things (IoT).
On the industrial front, AI-powered machines are using structured and unstructured data to overhaul manufacturing business models and strategies. With data volumes only burgeoning, enterprises are even identifying performance improvement areas to make smart factories even smarter.
AI-driven performance improvement strategies
Real-time monitoring and machine learning are together optimising factory operations by providing actionable insights into workloads at the machine level and production schedule performance. Obtaining this knowledge on a real-time basis has only helped engineers take better decisions for managing machines and overall operations. If we go by predictions, manufacturers will adopt machine learning and analytics to improve predictive maintenance by 38 percent over the next five years.
A German industrial manufacturing major has already started using neural networks to monitor, record, and analyse its steel plants’ operations. Sensors embedded in its machines consistently measure different variables and enable data-driven decision making. The AI system has managed to improve the performance of gas turbines and reduce emissions by 10 to 15 percent beyond what experts could achieve.
The oil & gas (O&G) sector is another prime example. In a prominent study, a global management consultancy collected hundreds of gigabytes of data over three years from a mature production platform equipped with 5,000 sensors. The data scientists working on the project used advanced analytics to enhance the offshore plant’s predictive maintenance practices. They were able to predict the occurrence of oil-in-water incidents, gas compressor train failures with over 70 percent accuracy, and the probability of pressure build-up in the well.
AI as a facilitator in smart plants
For some time now, smart manufacturing plants have been leveraging industrial robotics and automation to enhance operational efficiencies. In 2017, these technologies witnessed remarkable growth compared to 2016. Come 2018, the situation has only brightened as AI has made robots and automated machines more intelligent, perceptive, adaptive, and reactive.
Take the case of Amelia, an intelligent virtual engineer, created by an American technology start-up working on cognitive technologies and enterprise automation. This virtual agent utilises advanced machine learning models to advise clients without any human assistance. To input data into her system, Amelia has been empowered to read documents, learn from observations, and follow processes based on business analytics.
Machine learning technologies can assist manufacturing plant operators to do even more. These solutions not only help in distilling data-driven insights and running predictive maintenance and machinery inspection, but also moving materials and implementing production planning, field services, reclamation, and quality control.
The automotive industry was among the first to harness AI in manufacturing operations. Carmakers have deployed cobots with computer vision technology that enable human-machine collaboration on the same factory floor, eliminating the need to alter factory design. For quality control, companies have been quick to use AI-enabled visual quality checkers, which have boosted defect detection by about 90 percent. Besides this, AI has helped increase R&D productivity by 10 to 15 percent and save inventory costs by reducing forecasting errors by 30 to 50 percent.
The future is competitive …
While manufacturing organisations globally are still in the process of learning what AI can do for their business, China and the US are racing ahead to gain competitive advantage through AI. The US has about 850,000 employees working on AI with more than half of them having over 10 years of experience. China, on the other hand, has 50,000 employees with 40 percent of them having less than five years of experience. The tables may turn soon, considering China’s dogged efforts to mass-manufacture neural network processors and use the chips to enhance manufacturing operations. By 2025, we might just see China dominate the AI market.
The phenomenon has already found its footing. A leading Chinese smartphone manufacturer has employed an AI consultancy to help improve their factory efficiency and deployed more than 40,000 industrial robots to work with humans to produce smartphones.
… And about learning opportunities
As AI becomes commercialised, data scientists are finding more opportunities to closely study the technology’s potential and applications. The next logical step is toward developing energy-efficient deep neural networks and building an AI-powered automated factory where only robots work with humans at a safe distance. To realise the latter, an American automotive major recently bought a German engineering firm that specialises in fully automating factory floors.
Research is also underway to make AI more human-like. A non-governmental organisation that develops friendly AI applications is working with reinforcement learning algorithms to train AI agents to learn from their mistakes and act accordingly.
Despite the spike in the implementation of AI-enabled machines, there remains no regulatory authority to manage machine intelligence at the government level. While AI has still not achieved human intelligence capabilities, it is a good time to begin exploring this road. At the institutional level, investments have tilted toward researching influential algorithms and exploring other AI opportunities.
It’s hard envisioning the future with AI now at our behest. We’ll leave it to your imagination to extrapolate today’s possibilities to those of tomorrow.
This blog is by L&T Technology Services. Don’t miss their webinar on Tuesday, April 17th, 2018 – Perceptive, Adaptive & Reactive: AI in manufacturing
Comment on this article below or via Twitter: @IoTNow_OR @jcIoTnow
