Top 5 Uses of Artificial Intelligence Impacting Manufacturing Today

Companies across all sectors leverage artificial intelligence (AI) to gain valuable insights that improve processes, products, and optimize performance. This recent boom in AI applications is largely due to two factors: the ubiquity of powerful and inexpensive computing capabilities, thanks to cloud and edge technology, and the ever-increasing sophistication of AI algorithms and data science. This article examines five uses of AI that are transforming manufacturing processes and business strategy: 

1) Analytics for Intelligent Asset/Process Monitoring

The value of remote monitoring is not in the capture of data but the insights that can be gained from its analysis. Manufacturers who have implemented smart factory and Industrial Internet of Things (IIoT) initiatives understand this and are addressing their increased volume and complexity of data with increasingly powerful and sophisticated analysis tools to maximize value.

At a high level, modern AI has enabled manufactures to shift from descriptive and diagnostic analytics to predictive and prescriptive analytics. The former are employed to answer the questions, “What happened?” and “Why did it happen?”, while the later answer questions, “What will happen?” and “What should I do?” With the ability to predict behavior and additionally to prescribe actions based on those predictions, manufacturers can improve KPIs and limit the frequency and impact of negative incidents; for instance, by identifying an optimal combination of process variables to improve production yield or addressing early signs of machine failure to reduce unplanned downtime. The value gained here can be significant. For example, Deloitte estimates that by using predictive analytics manufacturers can reduce maintenance planning time by 20-50%, and overall maintenance cost by 5-10%.

Unfortunately, complex analytics often requires the expertise of experienced data scientists. Although this was a common hurdle for many early adopters of IIoT, some modern solutions, such as the ThingWorx IIoT Platform, have been purpose-built to simplify a robust analytics strategy. This has enabled organizations with deep expertise in their own operations but cursory understanding of sophisticated data analytics to benefit from artificial intelligence.

2) Generative Design for Product Development

Generative design is a growing AI-driven application that autonomously creates optimal designs from a set of system design requirements. Within a computer-aided design (CAD) environment engineers can specify design conditions such as loads, constraints, and materials as well as performance goals and the AI will produce a selection of geometries which meet that criteria.

With recent advances in generative design technology, the types of conditions and performance goals that can be incorporated and solved for have broaden significantly. For example, PTC’s generative design software, Creo Generative Topology Optimization, allows engineers to indicate the intended manufacturing process as a design constraint. The additional functionality expands the scope of generative design beyond design for performance to include design for manufacturability.

This powerful AI is changing how companies approach new product development by enabling engineers to rapidly explore a design space and quickly evaluate dozens of candidates that meet performance and manufacturing requirements. Entry-level mechanical engineers who lack years of practice implementing standard methods and rules of thumb can now, with generative design, create viable designs as quickly as their senior counterparts. Meanwhile, experienced engineers can use generative design to explore novel geometries that they might never discover using traditional approaches to design. In either case, generative design accelerates the early stages of the product development process, reduces the incidence of late stage re-design, and facilitates the creation of differentiated and optimized products.

3) Computer Vision for Quality Control & Augmented Reality

With advances in AI, specifically machine learning algorithms, the variety, complexity, and value of computer vision applications have grown significantly in areas of quality inspection and employee training and productivity.

Traditional computer vision applications for quality inspection are largely based on feature detection; identifying edges, corners, and colors and comparing them to pre-defined thresholds.  In AI-driven applications, the pass/fail criteria is no longer hard coded into quality inspection systems, but instead is discovered by the system based on reinforcement learning using known-good and known-bad samples. As a result, modern inspection systems can respond with high levels of accuracy to many different and far more subtle quality characteristics.

This complex object and feature recognition ability is also a key component of cutting-edge augmented reality (AR) applications. By incorporating AI-powered computer vision, AR applications like PTC’s Vuforia Expert Capture can provide digital tools and information within the context of the surrounding environment. For example, AR with object recognition can guide an operator through a series of complex assembly steps or reveal to a technician the location of a failing component in a downed asset. Even simpler applications, like displaying IIoT data at relevant locations of a manufacturing line, make it easier for factory workers to quickly identify and react to the state of operations. In a recent white paper, Forrester identified significant benefits attributed to the adoption of Vuforia solutions, including reduced training time by as much as 50%, and reduced overtime spend by 10 to 12%.

4) Autonomous Mobile Robots for Material Conveyance

The movement of material through factories and warehouses is a fundamental component to process efficiency and a great opportunity to apply artificial intelligence. The most prominent example of this is the adoption of autonomous mobile robots (AMRs). In fact, deployment of AMRs has doubled year over year and there is no sign that this growth will slow down soon. 

Unlike automated guided vehicles (AGVs), AMRs do not require a guidance system to be built into the environment where they operate. Instead these modern material conveyance solutions rely on spatial computing technology to determine their location in a factory or warehouse and advanced AI to navigate their environment, including the traffic created by people and other robots. With the freedom to safely move without restriction and the power of artificial intelligence to determine optimal routing, entire swarms of AMRs can be dropped into new or shifting environments and operate with incredible efficiency. This is the same technology that has allowed Amazon to boost efficiency in its fulfillment centers and sorting facilities.

5) Service Parts Optimization for Maximizing Asset Availability

For many manufacturers, service excellence has become a key competitive differentiator. As predictive and prescriptive analytics become the norm across industries, the organizations that rely on these new technologies to address issues quickly and proactively will need their suppliers to be just as agile in supplying service parts. This demand for more responsive service is being met by incorporating increasingly powerful artificial intelligence into material planning systems to ensure the right parts can be delivered to the right place at the right time.

It’s easy to make sure all service locations are sufficiently stocked if you don’t care about the carrying cost of inventory, or the opportunity cost of overstocking. So, for suppliers of service parts, the goal of material planning is to optimize the balance between cost and availability and when and where to stock. Achieving this optimal balance becomes exponentially harder as supply chains grow more complex, and customers more demanding. sophisticated new analytics tools, such as multi-echelon optimization found in PTC’s Servigistics software, address these complexities by considering all parts and service locations across multiple tiers of a supply chain to optimize them simultaneously. These decision support systems use AI algorithms in all aspects of supply chain optimization, from forecasting future demand, to determining optimal placement of parts in a complex interconnected supply chain with thousands of locations, and finally matching supply and demand on any given day.

Final Thoughts

It’s evident that artificial intelligence plays an increasingly important role in manufacturing, especially as organizations mature in their digital transformation journeys. But AI must be recognized as a means to an end. Many who adopt nascent AI technology find themselves grappling with the complexities of the data science in an almost academic manner. By partnering with solution providers, like PTC, that have deep domain expertise in both AI and manufacturing, organizations can instead focus on leveraging artificial intelligence solutions to address the challenges that truly matter to them: making better products and providing better service more efficiently.