Stäubli Robots and AI Inspect 100,000 Peaches an Hour at Greek Processing Plant

At the Agrophoenix fruit processing plant in Greece, eight Stäubli SCARA robots perform fruit grading at remarkable speed, processing a staggering 15 tons of peaches per hour during the season. That equates to approximately 100,000 peaches, each cut in half, inspected by an AI- and vision-based system.

The robots are part of the STAQ inspection system developed by Dutch startup QING. Eight parallel STAQ cells perform fruit grading with the primary goal of removing peaches containing pit fragments.

Agrophoenix was founded in 2018 in Greece's Imathia region as "not just another fruit processing company" with the aim of "creating something new and fresh," according to Lazaros Ioannidis, partner and project manager. The company established a new production line for fruit balls in pouches and packs aseptic diced fruit in 200-liter bins for year-round use.

After the fruit is washed, steam-peeled, cut into halves and depitted, it is pre-inspected by a vision system, mainly for color. The peach halves then arrive at one of eight conveyors, each leading to a compact cell containing an AI-assisted vision system and a food-grade Stäubli HE robot. In the cell, the camera captures every peach half and inspects it in milliseconds. Immediately after screening, the robot picks the rejected halves and places them on belts running to the right and left of the main conveyor, while the majority exit the cell for further processing.

QING uses AI to train the vision system, meaning the user can easily determine and teach it the difference between acceptable and defective products, and adjust the parameters as needed. "The robots can pick one peach half per second," says Will Uijting, Commercial Director of QING. Since the rejection rate is low because the fruits have already been graded, 15 tons of peaches can pass through the eight STAQ cells each hour.

Lazaros Ioannidis explained:

"We calculate in weight, not pieces. A peach weighs approximately 150 g, so 15 tons means 100,000 peaches per hour and 200,000 halves, more or less. And 3,600 of them can be rejected in that time."

Uijting added:

"At the moment, we only look for pits and pick them out in two channels: those with small pits and those with bigger ones. Those with bigger pits are directed to automated re-pitting, and the smaller ones go to manual inspection. Then both are fed back into the STAQ stations."

In the STAQ cells, a Stäubli TS2-80 SCARA robot in the HE version selects the rejected peaches. The HE version offers washdown capabilities, with protected joints, corrosion-resistant materials, and inner pressure that prevents vapor, steam or water from entering the housing during cleaning. QING opted for H1 oil filling because the robots are top-mounted over the conveyors.

The cells need frequent cleaning. During operation, every four hours the lines stop for 20 minutes and an automatic CIP-like cleaning starts. Each Sunday, a more intense cleaning with a food-grade detergent takes place. The Stäubli robots in the EHEDG-compliant HE version are designed for frequent cleaning without compromising longevity.

As for why it took two seasons to get the eight QING cells production-ready, the answer lies in the gripper and the material that comes into contact with the peaches. "The peeled peaches are slippery, so we had to find a conveyor material on which the peaches hold their position. Even more difficult was the effective picking, grabbing, and releasing of the peaches with their cups upside down," Uijting said. The engineers had to find the right vacuum gripper with the right contact material.

In the end, they found the right combination, and the AI-based peach grading, with Stäubli HE robots picking the rejected peaches, successfully completed its first harvest season at the Agrophoenix plant. The collaboration with QING will continue, as Ioannidis explains: "The next step is a difficult one, but we will try. Together with QING, we want to develop an automated system that is able to remove pit fragments from the peach halves – before or after the sorting. That's our next big thing."

The complete solution is an automation framework called STAQ, developed specifically for the food industry. The acronym stands for See, Think, Act. "We look at the product with sensors or cameras, we determine what to do with it, and we act upon it. This works for different and difficult products, not only fruit like peaches, but also meat and biscuits," Uijting said. "The software and the functions are always the same, while the grippers and the algorithms are adapted to the products and sorting parameters."