Artificial Intelligence is Becoming a Game-changer for Potato Sorting Companies

Artificial Intelligence is here. It is not just a buzzword. Companies are working with it every day to improve their business. If we look at what has changed in the past 12-24 months, we see two major changes accelerating the adoption of AI in potato processing – the increase in data storage capacity and the access to tools.

Data is the cornerstone of AI innovation – more data, meticulously structured, means more powerful AI models. As the demand for data surges globally, we’re witnessing an unprecedented expansion in data centers, a trend accelerated by the COVID-19 pandemic. This booming market is projected to skyrocket from USD200bn in 2021 to an astounding USD343bn by 2030.

Access to a diverse array of tools and products from leading service providers has improved dramatically. Microsoft Azure, a popular choice for building, deploying, and managing AI applications, is just one of many options in today’s market. Google Cloud and AWS Bedrock are also making significant strides, offering powerful solutions that are reshaping industries.

These advancements have driven food industry process engineers to enhance their skills and adapt to the evolving AI landscape. While potato specialists may now collaborate closely with data scientists, they don’t need to become data scientists themselves. The user-friendly AI tools at their disposal allow them to effectively harness this technology, providing valuable insights and guiding the work of data scientists. Together, they create a synergy where the specialist’s expertise informs and interprets AI-driven predictions, ensuring precision and innovation in their field.

Potato Business: Please discuss the Artificial Intelligence factor in an optical sorter and how equipment integration (software and hardware) is performed in the case of optical sorters.

Eamonn Cullen: Artificial Intelligence is revolutionizing optical sorting, becoming a game-changer for potato sorting companies. By integrating AI into research and development, companies are enhancing camera image analysis and spectral image interpretation, leading to unprecedented precision in detection. During sorter operation, AI processes and classifies every pixel in just 20 milliseconds, treating clusters of pixels with precision thresholds.

With the innovative CIT technology in Insort sorters, processors can now differentiate between wet wood and potatoes with remarkable accuracy, without needing to train the sorter to detect foreign materials like glass, stones, wood, and plastic. Instead, AI recognizes and captures these foreign objects, ensuring they are swiftly and effectively removed from the product stream. This rapid and crucial process showcases the immense power of AI in sorting, giving customers confidence that their products are thoroughly inspected and safe.

P.B.: Big tech companies help introduce potato processors to AI foundation models. Please tell us more about Insort’s experience in managing its huge amount of data generated daily.

E.C.: The vast amount of data generated at potato processing facilities is a goldmine waiting to be tapped. To unlock its full potential, potato processors need to establish a comprehensive data lake or data store, where all valuable information can be seamlessly collected. This data repository would integrate insights from every critical point in the supply chain, including farmers, logistics, quality inspection, utility supply, and customer feedback, driving smarter decisions and optimized operations.

Insort technology is used at different stages of the process and collects as much as 1GB per second of data. Potato processors are taking this data and merging it with all their data to develop process line AI models to determine the process line settings for each of the process steps.

Companies leverage AI foundation models – large-scale machine learning models trained on vast datasets – for a variety of transformative purposes across industries. They integrate data management, advanced AI technologies, and human expertise. By following these steps, companies can extract valuable insights from their data and drive informed decision-making. Companies such as AWS offer a wide range of products and foundation models. The great value of this is that their teams of experts work with potato process experts to choose the correct solution for their business.

P.B.: What are the main benefits associated with an optical sorter in a potato processing line?

E.C.: Optical sorting is a critical component of any potato processing line, ensuring that foreign materials like stones, wood, metal, and glass are effectively removed. Every customer demands the assurance that their brand is safeguarded by an advanced automated inspection process. Thanks to cutting-edge technology, an impressive 99.5% of foreign materials can now be detected and eliminated. Beyond just safety, optical sorting also provides unparalleled quality analysis throughout every stage of potato processing – from unpeeled whole potatoes to premium French fries, chips, or other final products.

The Insort sorter excels in detecting not only visible defects but also those invisible to the human eye, using AI to accurately identify and classify every imperfection. Additionally, automated inline process monitoring offers real-time insights into every potato and every piece of potato in the production line. This enables processors to implement SMART programs that optimize the entire operation, from raw material selection to peeling and frying processes.

Insort sorters inspect across a range of 400nm to 1700nm, capturing extensive data that drives each of these three key benefits to peak performance, delivering unparalleled results in potato processing.

P.B.: What is the per-hour capacity range for an optical sorter?

E.C.: The capacity of a sorter is influenced by several factors, with potato size being a crucial element, as capacity is measured in tons per hour. Insort’s sorters, with a maximum width of 2.4 meters (7.87 feet), deliver the highest capacity available in the potato processing industry.

Today’s potato processors are laser-focused on maximizing output while minimizing inputs like raw material, water, and energy. A key question they face is: why operate two parallel processing lines when a single line can meet capacity requirements? Running two lines not only takes up valuable factory space but also increases maintenance demands and the overall Cost of Ownership (CoO).
 

As a result, many processors are turning to larger sorters to boost throughput on a single line, aiming to maximize potato utilization and increase yield. This approach ensures that they use the least amount of floor space, the least amount of maintenance, while also minimizing potatoes, dry matter, and water to achieve their desired output.

Artificial Intelligence plays a pivotal role in the optimization of a single process line with a high capacity sorter. With AI, processors gain a comprehensive map of the entire production journey, from raw material input to the final product delivered thousands of kilometers away. They can accurately predict the amount of input required to fulfill customer orders, anticipate raw material size and quality, and use AI models to forecast yield, storage duration, and transport times. This level of precision empowers processors to meet their goals with unparalleled efficiency.

P.B.: Which are the hardest to detect defects in potatoes for an optical sorter, during processing and why is that?

E.C.: Defect is a relative term – what one customer considers ideal might not meet another’s standards. The real challenge lies in addressing skin, scab, and rot defects, which vary by potato variety and size. These can all be effectively removed, but when it comes to detecting internal defects and specific potato properties, CIT imaging and other advanced inspection technologies become essential.

For processors, the primary objective is to maximize the utilization of every potato entering the production line. This means not just discarding defective potatoes, but intelligently aligning them with the right finished product specifications. By identifying and re-purposing potatoes based on their unique attributes, processors can optimize yield and reduce waste.

In today’s modern processing facilities, producing a diverse range of potato products with distinct specifications is common practice. This approach not only ensures the optimal use of every potato but also boosts the overall efficiency and versatility of the production process.

Even when a potato doesn’t meet the required specifications, it still contributes to sustainability. Non-conforming potatoes can be transformed into biogas through anaerobic digestion, providing a renewable energy source. This process not only minimizes waste but also aligns with processors’ sustainability goals by generating clean energy, reinforcing the commitment to a greener future

P.B.: How is a perfect 360-degree scan of each potato achieved? How many cameras are needed and what technology do your optical sorters use?

E.C.: While we haven’t yet achieved a perfect 360-degree scan of each potato, INSORT sorters come remarkably close, inspecting 90% to 95% of the surface area (depending on the size and shape of the potato) as the potato is presented to four high-resolution cameras while in mid-air. For optimal inspection, viewing each potato off the belt offers the most comprehensive surface analysis available today.

Insort sorters are equipped with cutting-edge technology, featuring four high-resolution cameras that operate across a broad spectrum from 450nm to 1700nm – equivalent to 240 lasers per pixel, viewing every object. This advanced system integrates two RGB cameras with two CIT cameras, providing an in-depth view of the potato’s surface while uncovering hidden attributes.

By combining these two powerful camera types, Insort sorters not only capture an extensive surface area but also detect invisible elements such as high sugars and other chemical properties. This ensures a thorough and precise inspection process, elevating the standard of quality and consistency in potato processing.

P.B.: How many sorting algorithms can be programmed on your optical sorting equipment in the case of potatoes? (size, color, types of defects, bruising, foreign material, etc.).

E.C.: Insort has developed a specialized AI algorithms tailored to both the sorter and the customer’s specific product requirements. These algorithms are continuously improved by adding more images captured from the customer’s live production, enhancing its precision over time. The Insort sorter uses multiple different algorithms.

The Insort sorter also includes multiple programs that integrate potato characteristics—such as size and surface quality – to determine the best fit for each of the customer’s final product specifications. This allows the customer to customize and produce the optimal program for their desired product specification and end-user requirement.

By utilizing AI to map the entire production process, process engineers can accurately predict the yield of finished products from each potato. This enables them to leverage AI insights and use the insights to customize and fine-tune sorter program parameters while managing the production process with exceptional precision, ensuring that downstream processing steps are highly efficient and that waste is minimized.

P.B.: What are some of the most common demands you get from existing and potential customers referring to optical sorting machinery?

E.C.: Every customer is unique and each customer has specific potato sorting needs. Processing flexibility is most common request we receive from customers. Can you detect this piece of foreign material or can you detect this type of defect?

Today’s key challenge in potato processing is managing the wide range of potato varieties, size and quality and while also managing different potato diseases and defect types. Each potato is unique and each potato growing area produces different challenges. Customers sorting needs vary greatly – some are internal, some are external, and others only become visible after thermal processing.

The complexity of sorting has increased as the number of potato issues continues to rise due to various factors, such as pests, nutrient deficiencies and disease. However, industry leaders like Insort are at the forefront, continually advancing their inspection technologies to empower potato processors to deliver the highest quality products.

P.B.: How do you reconcile the required speed of operation with the continuous improvement of the ‘good-in-bad’ ratio in sorting systems?

E.C.: GiB (Good in Bad) is a critical concern for customers who are keen to minimize the waste of quality product. When a piece of good potato is removed along with every foreign material or defect, it is not acceptable. The ultimate goal is to ensure that no good product ends up in the foreign material stream.

At Insort, our sorters are fully customizable to meet specific business needs. For instance, if a piece of glass enters the sorter, our ejectors are designed to divert it with utmost precision by extending the ejector’s open time, to guarantee that the piece of glass is removed. While the removal of glass might occasionally result in the loss of a potato, our system puts the control in the customer’s hands.
Customers have the power to adjust the good-to-bad ratio according to their desired sorting performance. This ensures they achieve the highest quality results while minimizing the loss of good potatoes.

P.B.: Lastly, please discuss food waste reduction and sustainability overall. What role do optical sorters play during the processing of potatoes in this regard?

E.C.: Maximizing the use of every potato that enters the production process is the top priority for everyone in the industry. As consumption continues to rise, processing volumes are increasing year after year, driving up the demand for potatoes. However, this growth presents challenges, particularly with water utilization for crop irrigation. The cost and availability of water are likely to become the key factors limiting potato production.

To address this, it’s crucial to minimize potato waste and ensure that every potato is fully utilized. Efficient sorting is the cornerstone of achieving this, but it requires a comprehensive approach. Integrating AI into processing factories is essential for elevating yields to the next level.

At Insort, our customers are already leading the way by using advanced sorters and food analyzers throughout the production process to detect dry matter. This enables processors to ensure that every finished product meets precise dry matter content targets, guaranteeing consistent quality and optimizing resource use.

Increasing the dry matter content of a finished product by just 1% equates to using 2% more raw material (potatoes) in the production line. Traditionally, operators had to take samples and send them to a lab for dry matter analysis. Today, this process has been revolutionized. Operators now leverage AI-powered reports to access real-time process data, monitor yield values, and pinpoint opportunities for process optimization.

The role of process operators is rapidly evolving. With AI support, their ability to drive enhanced performance, sustainability, and profitability is significantly amplified, making their contribution to business success more impactful than ever.