AVEVA’s new Vision AI Assistant uses video and image inputs to provide computer vision solutions for industrial applications.
Within manufacturing industries, the adage “time is money” rings true. When companies can increase their efficiency, it often leads to direct savings in capital and increased revenue. But identifying bottleneck issues within operations and finding meaningful solutions can often be expensive and time-consuming. The promise of Industry 4.0 is to integrate artificial intelligence (AI), machine learning (ML) and advanced data analytics to help companies gain meaningful insights from their data.
Many software companies are working on AI tools to support Industry 4.0 within manufacturing industries. The goal is to deliver real-time data analytics to support industrial decision-making, improve product quality, enhance employee safety and reduce unexpected downtimes. However, realizing this goal remains a challenge for companies without access to a dedicated data science team.
AVEVA, an industrial software company, recently introduced its Vision AI Assistant. Using existing general-purpose cameras, the AI-driven analytics tool can use image classification algorithms to identify anomalies before they can cause expensive and time-consuming damage.
“The Vision AI Assistant enables enhanced process optimization, thereby reducing time, wastage and costs,” said Jim Chappell, AVEVA global head of AI and Advanced Analytics. “Cameras can also be used in areas that are not suitable for humans for security or safety concerns, freeing up workers to focus on high-value tasks instead of continuously monitoring live camera feeds.”
What Is an Image Classifier and Computer Vision?
Image classifiers have a vast number of applications within the world of AI tools. A massive amount of visual data is readily available within industrial settings, and AI offers a powerful tool for taking advantage of this information. Image classifiers are a core component of vision AI, also known as computer vision (CV).
In a general sense, an image classifier relies on a set of training images that help an AI algorithm learn to identify specific objects or categorize different inputs. For example, an image classifier can learn the difference between an apple and an orange and then sort a series of images within those categories. This is incredibly useful for industrial applications, helping companies scale and automate quality control and identify bottleneck areas to improve efficiency.
When it comes to vision AI, the image classifier algorithms can handle increasingly complex visual data to assist companies with answering more complex questions. For example, video footage can be monitored in real-time for anomalies or other inconsistencies that may indicate a safety concern or possible disruption to production. By incorporating CV into industrial applications, companies can benefit from real-time data analytics without relying on manual surveillance of live footage. Resources can then be better allocated to focus on innovation and higher-level operations within a company, leaving manual tasks to be automated.
AVEVA is not the only company working on CV applications for industry. Google, IBM and other major software companies are also developing applications to help industries better utilize visual data.
Repurposing Existing Cameras to Optimize Operations
AVEVA’s Vision AI Assistant can use standard digital camera video and photo inputs to generate user-friendly insights and real-time warnings for industrial operations. The AI software leverages deep learning to gain insights about visual information already collected by companies. The software was designed for low latency industrial applications to help companies monitor core equipment and manufacturing processes to enhance efficiency, safety and sustainability. The AI assistant employs two main detection methods with broad applications in industry: anomaly detection and discrete state detection.
To detect anomalies, an unsupervised machine learning algorithm is employed to identify baseline trends in the visual data. Then, statistical analysis determines if a given image or video represents a significant change. The tool is trained using a collection of images curated by a company to feature expected or acceptable outcomes. Therefore, anomalies are anything the AI tool decides is distinct from the expected training set. A company can then fine-tune the learning process to ensure the acceptable outcomes fit within their operational vision.
For discrete state detection, the AI assistant uses a supervised deep-learning algorithm to distinguish an object between two defined states. To construct a model for each category, companies provide training images that are already annotated and include diverse examples of both states. The algorithm can then classify input images within one of the defined categories. The final product is a classification system that can sort images within discrete categories to assist with operational efficiency.
The Vision AI Assistant is integrated into the AVEVA System Platform and Operations Interface and AVEVA Insight to provide AI outputs without requiring a data science team. The software can be purchased within the company’s Flex subscription program to simplify licensing.
Applications of the software are diverse and include everything from increasing product quality to improving industrial efficiency. For example, companies can use the image classifier to identify defective products on a manufacturing line by automating the detection of items that pass or fail quality control.
Beyond standard visual inputs, the AI tool can assess infrared or thermal data to identify anomalies that may indicate a gas or fluid leak, reducing failure incidents and supporting employee health and safety. The software can even identify hot spots in data centers to prevent overheating of servers, processing units or other core equipment before they break. This can save time and money that would normally be required to fix equipment when broken.
“AVEVA’s Vision AI is monitoring and analyzing the 1.5 km main drive chain for our Lexington, Ky., smart factory, which manufactures electric breaker panels,” said Kenneth Labhart, Schneider Electric innovation leader, Global Supply Chain North America. “If that chain goes down, our entire factory comes to a halt, which typically happens a few times per year. Vision AI has already provided early detection of issues with parts of this chain on two occasions, giving us the ability to take proactive action. By doing this, we avoid revenue-impacting, unplanned downtime. We’re looking forward to expanding our usage of this AI system.”
AVEVA Is Working on Solutions to Deliver a More Sustainable Future
An overarching goal of AVEVA is to provide software solutions that support the sustainable operations of companies. With its software suite, including the vision AI assistant, the company is working on helping industries achieve UN Sustainable Development Goals. By improving operational efficiency, advanced data analytics can help customers reduce emissions and maximize energy efficiency.
One exciting example is Network Rail, which operates the United Kingdom’s rail system and supports more than 1.7 billion passengers. With AVEVA’s Unified Operations Center, Network Rail built 90,000 critical assets into one operational system to reduce failures that affect train service by 6.6 percent. The solution helps improve energy efficiency and resource management throughout the rail network.
In partnership with Henkel, AVEVA also created a proprietary cloud system to support the company’s global operations and reduce energy usage by 16 percent over the past six years.
Vision AI is just one of many ways companies are looking to achieve sustainability goals in the coming years. By increasing energy efficiency and reducing operational losses, companies can increase both sustainability metrics and revenue. With investments in advanced data analytics, companies have the unique opportunity to exploit their own data to glean meaningful insights to improve operations. A recent study found widespread interest in CV adoption across industries, and it will be exciting to see how this technology transforms industrial operations in the future.