AI in manufacturing – Transforming aerospace and semiconductors

Manufacturing is on the verge of a radical revolution and technological disruption in its end-to-end value chain, with the use of AI technologies and applications, from virtual assistants to advanced robotics.

Intelligent manufacturing is a new emerging trend. It is a smarter production approach, where machines connected with the Internet assemble parts and adapt with minimal guidance from human operators to new processes.

Intelligent manufacturing enables the industry to leverage the large volume of data to optimize processes in real-time. It can shorten development cycles, improve engineering efficiency, prevent faults, reduce inventory costs with better supply and demand planning, enhance safety by automating risky activities, and increase sales with better lead identification and price optimization.

The scope of AI compels many manufacturers to adopt new plant designs, reshape their manufacturing footprints, and devise new supply chain models. In this post, we will discuss six ways how AI transforms manufacturing in two industries: aerospace and semiconductors.

1. Quicker turnaround and fewer iterations

Engineers and researchers face difficult challenges in manufacturing, from the rapid growth in demand to market fragmentation driven by consumers’ taste and preference for new customization. At the same time, they need to work within the budget constraints and improve their productivity and efficiency in addition to ensuring quicker turnaround with fewer iterations without compromising the predictability of product performance. Thanks to the AI-powered technologies, they can help deliver more efficient designs than previously achievable. Innovations can hit the market much faster as AI facilitates lower process cycle times and real-time negotiations.

Notably, after Intel deployed a team of data scientists to speed up data integration in its R&D department, the company achieved a 10 percent higher yield for integrated-circuit products, compared to other players who were pursuing similar designs at the same pre-production development stage.

Predictive analytics using machine learning is a powerful tool to reduce the time required to solve design problems. An artificial intelligence startup named Motivo managed to compress its semiconductor’s design process from years to a few weeks, saving a lot of money in iterations and testing. Manufacturers can apply machine learning to consolidated comprehensive data, enhancing the design process from beginning to end. AI allows manufacturers to integrate production reviews and client feedback in real-time to refine the product design.

2. Untangling the procurement process

Manufacturing inventory and stock maintenance is a complex challenge. Manufacturers must source thousands of different parts from tens of thousands of suppliers all over the world. When they are digitally linked with the suppliers, they can balance the supply chain and optimize inventories in real-time, bringing better transparency and visibility into the stock management.

One aerospace manufacturer solved the problem by applying advanced algorithms on their spending data to search for discrepancies in product prices and understand differences across suppliers. It helped them develop effective procurement levers and reduce administrative costs. Similar supplier research and analysis can be automated with machine learning algorithms, and full automation could reduce IT staff numbers by 39 percent.

3. Effectiveness of review processes

It is needless to say that program reviews are highly critical in manufacturing. Any failure in detecting emerging problems or prioritizing critical decisions can result in delays that can cost millions of dollars. The communication gap between manufacturing functions and marketing can also hinder the process.

Notably, a leading aerospace manufacturer sought to improve the effectiveness of program reviews by using advanced analytics that can predict key performance indicators and identify “traffic lights,” such as heavy email traffic, which could be a bellwether of later-appearing problems. In doing so, the manufacturer generated an impact of around €40 million on pre-tax run-rate cash flow.

Manufacturers can use deep learning technology to optimize the key performance indicators of program reviews in real-time. It can predict, identify, and prevent material and staffing bottlenecks. Virtual agents can also alert program leadership and team members when problems arise and recommend solutions.

4. Redesigning assembly lines

Assembly and manufacturing inefficiencies or inaccuracies in fault detection and classification can cost billions of dollars every year for manufacturers. They can cause expensive and unnecessary interruptions on the assembly line.

Factory managers can apply deep learning to information flows in real-time to update and enhance the accuracy of standard operating-procedure predictions, with visibility on component availability and risk management, particularly during ramp-ups. The reliability of assets can also be enhanced with AI tools, thanks to machine learning for improving the predictive accuracy of defaults, and production interruptions.

In order to reduce assembly errors and flatten the learning curve for new operators, virtual agents will provide instructions and information on tablets or other interactive personal communication devices. In real-time, assembly lines will be significantly automated and optimized, with control and scheduling linked directly to real-time dispatching systems.

To solve these problems, an aerospace manufacturer applied AI technologies and saved €350 million. Nearly 60 percent came from using machine learning algorithms and advanced analytics to review data from every step of the assembly process, and then rewrite standard results-based operating procedures. By using machine learning to propose the best time to leave the office or warehouse, a semiconductor maker reduced its material delivery time by 30 percent. It also increased its output by 3 to 5 percent.

5. Delivering after-sales service

Aircraft maintenance and service is a significant part of the aerospace value chain. Maintenance, repair, and overhaul sales leads are cumbersome, resource-heavy, and always optimized manually. It results in unnecessary grounding of aircraft, non-revenue generation, and inefficient allocation of expensive engineering labor.

GE broke new ground in this space in the late 2000s with its service concept of “power by the hour,” which calls on operators to pay for aero engines only when planes are flying. This business model is becoming ever more dominant in aerospace manufacturing, thanks in particular to the Internet of things. The sensor and actuator network allows companies to monitor the actual use of their products, provide customized pay-as-you-go services, and better prevent disruption of service and downtime.

Ideally suited for leveraging the vast amount of data collected from operating engines, machine learning algorithms can make judgment calls on when to deploy drones and smart micro robots to conduct aircraft inspections and quality controls on intermediate and final products with real-time feedback between airplane and ground support facilities. Manufacturers with the best algorithms and data will be able to offer service contracts at a lower cost, which promise better performance.

Improving the accuracy of MRO work forecasting and focusing sales efforts on the most promising leads can have a significant impact on the EBIT of manufacturers. Advanced analytics and AI tools are effective in optimizing processes around unplanned maintenance events, allowing the manufacturer to respond to disruptions more effectively, and increasing uptime.

For example, GE turned to Kaggle, a platform for competitions in predictive modeling and analytics, and invited data scientists to design new flight planning routing and machine learning algorithms. This optimized fuel consumption by looking at variables like weather patterns, wind, and restrictions on airspace. The winning routing algorithm showed an increase in efficiency of 12 percent over actual flight data.

6. Collaboration and communication

The key feature of the future manufacturing is collaborative agility – an ability to adapt to the changes in demand, regulations, input prices, technologies almost instantaneously. From the perspective of human capital, manufacturing will become more collaborative, based on ever more complex activities and interdependent workers. With a global digital backbone, manufacturing plants around the globe, supply chains, and value chains will become more interconnected and collaborative from a technological standpoint.

It will combine highly automated plants that use smart robotics extensively for mass production, a customer-centered plant network close to higher-end market segments, and mobile plants that can produce a limited product range at competitive cost. Machine learning applications will help all players to collaborate and leverage the opportunities in real-time.