by Steve Griffith, Industry Director, NEMA

Multinational automaker Nissan manufactures vehicles in 20 countries worldwide, with production volume exceeding 5.6 million vehicles. And while its production assets were generating an abundance of operational and production data, the company lacked sufficient skilled resources to perform analysis on all of that data adequately.

However, through an engagement with its vendor, Senseye, and using artificial intelligence (AI), Nissan was able to analyze that data and generate predictions on when its production machines would need maintenance. Before, the automaker used static metrics (such as the number of process cycles or the number of hours in service) to determine when a machine needs to be taken out of service for maintenance.

Nissan used machine learning (ML) algorithms to monitor and spot patterns in the operational data of more than 9,000 connected assets and more than 30 different machine types, including robots, conveyors, drop lifters, pumps, motors, and press/stamping machines. Maintenance users can track and predict machine failures in advance. As a result, the company can optimize maintenance activities and make repairs months before a machine fails and slows or halts production. According to the company, Nissan reduced unplanned downtime by 50 percent and generated a payback period for the project of fewer than three months.

Nissan is hardly alone in its use of AI. Manufacturers within the electroindustry and medical imaging sectors of all types and sizes are increasingly using technology that enables components of AI, such as ML, deep learning (DL), computer vision (CV), and natural language processing (NLP) to enable a wide range of cases. AI can provide companies with enhanced operational visibility, allow enhanced pattern analysis that can drive predictions, support more autonomy, and tie disparate machine and information systems together.

Trends Driving the Use of AI

Several trends are driving manufacturers to incorporate these new technologies. There has been a shift toward cloud-based analytics and ML (both in the public and private cloud) and the use of hybrid environments, mainly when data security, privacy, and confidentiality are an issue. Further, manufacturers of all types are investigating or actively using manufacturing execution system (MES) solutions that are infused with ML algorithms. To support these efforts, vendors offer solutions that capture, clean, and prepare data for analysis, enabling data science professionals to work on more interesting data analysis problems. The solutions also allow operations professionals to interact with algorithms, even if they are not experts in data science or programming.

Potential Use Cases in Manufacturing

While there are dozens of potential AI use cases, a few are commonly deployed by manufacturers today. While many manufactured goods are already constructed using precision techniques and materials, ML and DL allow systems to review vast amounts of process data collected by sensors on the machine and in the production environment to identify variances, patterns, and other anomalies that may affect the efficiency of a process or the end quality of a product. This means that manufacturers can more easily modify functions at a granular level, resulting in efficiency gains, improvements in quality, and reduced waste materials.

AI also is useful for managing and combining smart technologies and services and can be used as a catalyst for integrating disparate networks and cutting costs. By incorporating ML and DL, systems can identify patterns in the data generated by machine logs, system logs, and data from the larger production ecosystem, identify inefficiencies or problems, and then apply an optimal solution based on historical data patterns or simulations.

Manufacturers are also utilizing AI to improve workplace safety. Using a combination of machine vision and ML, images from video surveillance cameras can be used to recognize humans, vehicles, or objects. Machine vision incorporates a series of algorithms that compare an object seen with hundreds of thousands of stored reference images of objects in different postures, angles, positions, and movements. Once trained, the algorithm can determine whether the observed object moves like the reference images and whether it appears like the reference images (as well as other characteristics, including speed and gait). It combines all the values to ascertain what the object is and can apply various preprogrammed rules to safety alerts when the object or its attributes fall outside the accepted norms. 

At their heart, manufacturers are companies that must market and sell their products. AI is increasingly being used to drive marketing and sales processes, allowing manufacturers to understand better, engage with, and satisfy existing and new customers. Utilizing a virtual assistant (which can be set up to handle inquiries on a 24/7 basis) to handle a first-call inquiry reduces the possibility of a lost sale or inquiry due to a prospect or customer dropping off and seeking out another manufacturer.

Another way to utilize AI is to capture data about customer behavior at each touchpoint of the customer journey. A more personalized and customized experience can be created by analyzing this data to find specific patterns, allowing the relevant delivery of particular content, product recommendations, consumable reminders, or other marketing messages.

Any AI-driven marketing and sales strategy aims to improve the accuracy and responsiveness of routine tasks, thereby freeing up humans to handle higher-value interactions. By intelligently automating initial or regular customer touchpoints, operational efficiency should improve, which would enhance customer experience and retention.

Keys to Success with AI

Manufacturers that have successfully incorporated AI technology generally have a firm grounding and knowledge in how analytics and AI can complement each other and which tasks or use cases are best suited to being handled separately. In addition, developing AI projects requires setting realistic goals and benchmarks for success for each use case and establishing processes to confirm or override outputs that do not deliver expected or valuable results. To develop valuable goals and metrics, AI development teams must identify and understand the problem to be solved, how an AI algorithm can deliver results and methods for verifying or modifying the algorithm quickly.

Many manufacturers also leverage private clouds to gain access to APIs and services without creating custom ML models. And many are focused on the issue of “explainable” AI, which uses ML techniques that produce transparent models with audit trails while maintaining prediction accuracy.

Benefits of AI to Electroindustry and Medical Imaging Makers

If deployed correctly, AI can provide benefits to manufacturers and industrial companies, both for their internal operational practices (and the products they make) and the customers they serve. AI can improve quality assurance by deploying ML and DL to allow systems to review vast amounts of process data and identify variances, patterns, and other anomalies that may affect the efficiency of a process or the end quality of a product.

Other algorithms can be used to identify patterns in the data generated by machine logs, system logs, and data from the larger production ecosystem, identify inefficiencies or problems and then apply an optimal solution based on historical data patterns or simulations.

AI also is increasingly being used to drive marketing and sales processes by using ML algorithms and enhanced data analysis to conduct more detailed and granular research of engagements, activities, and purchasing processes. Leveraging the pattern-recognition power provided by AI, manufacturers can develop a more one-to-one relationship with each customer and enhanced automation of interactions and responses.

All told, AI is opening up a wide range of benefits for manufacturers of all types. However, while AI-knowledgeable enterprises may be able to ensure a solution that meets their exact needs, most manufacturers should honestly assess their level of skill and knowledge with AI and consider using pre-built AI solutions and deploy them on the cloud. Pre-built and embedded AI solutions can be rolled out quickly and updated and refined on a rolling basis.

“AI is an emerging technology that is becoming increasingly prevalent across the NEMA Sectors,” said Steve Griffith, NEMA Industry Director and staff lead for the NEMA AI Strategic Initiative. “Additional insights into the drivers, technologies, use cases, regulatory issues, and deployment strategies used by manufacturers and is available to Members in the NEMA white paper, AI in the Electroindustry and Medical Imaging Sectors.  The concepts described in this whitepaper will serve as a basis for our AI work in 2021.”  For more information on how to participate, please contact Griffith directly at Steve.Griffith@nema.org.