TOMRA launches high-capacity potato sorting system

TOMRA Food has launched the TOMRA 4A optical potato sorter, combining AI-assisted inspection with three-way material separation in a machine designed to handle up to 100 tonnes per hour during harvest and pre-storage operations.

Introduced on 5 November 2026, the equipment is intended for unwashed potatoes and uses the company’s LUCAi deep learning platform, dual-sided Pulsed LED+ optical inspection and mechanically controlled rejection fingers to separate acceptable potatoes from foreign material and defective tubers.

The sorter has been developed to accommodate different requirements before and after storage. During harvest, its principal function is to remove stones, soil clods, other foreign objects and unsuitable potatoes before the crop enters storage. When potatoes leave storage, the system can undertake more detailed quality sorting at a lower throughput.

TOMRA will demonstrate the machine publicly at Interpom in Kortrijk, Belgium, from 29 November to 1 December 2026. The company is positioning the 4A as a platform that can support several sorting applications, with onion sorting planned as a future addition.

Harvesting and receiving operations can involve large quantities of potatoes mixed with soil and stones. The proportion of unwanted material varies with field conditions, harvesting equipment and the way the crop is handled before it reaches the sorting line.

Removing that material before storage reduces the quantity of foreign matter passing through subsequent handling equipment. It also avoids occupying storage capacity with material that does not form part of the saleable crop.

Traditional manual picking systems require workers to identify and remove unwanted items as potatoes pass along a conveyor. The effectiveness of this approach depends on presentation, line speed, the number of workers and the ability to distinguish defects under the available inspection conditions.

Optical sorting replaces part of that inspection and removal process with cameras, classification software and automated rejection equipment. The machine must first obtain usable images of the incoming material, then decide which items should remain in the accepted stream and which should be diverted.

The TOMRA 4A uses Pulsed LED+ technology to inspect both sides of potatoes passing through the machine. Infrared inspection provides additional information beyond visible colour, helping the classification system distinguish between materials with different optical characteristics.

Dual-sided inspection is useful where an unwanted feature is visible on only one side of a potato. It does not eliminate the need for suitable product presentation, as overlapping material or unsuitable feed conditions can affect the information available to the sensors.

LUCAi processes the inspection information using deep learning models trained to distinguish acceptable potatoes from foreign objects and particular quality defects. The machine is supplied with pre-trained models, allowing operators to begin using its standard classifications without developing a new model from the outset.

TOMRA also provides the ability to train the system for site-specific conditions. This can be useful where a particular soil type, crop variety or recurring defect produces characteristics that differ from those encountered in the standard training data.

Additional model training must be evaluated against representative samples. A classification that performs satisfactorily on one batch cannot automatically be assumed to provide the same results with different crop conditions, defect concentrations or material presentation.

After classification, the sorter uses two sets of rejection fingers to direct material into separate discharge streams. One removes foreign objects including stones and clods, while the other separates potatoes that fail the selected quality specification.

The arrangement produces three outputs: accepted potatoes, foreign material and out-of-specification potatoes. Keeping rejected tubers separate from soil and stones means that they can be assessed for alternative uses rather than automatically becoming part of a mixed waste stream.

Depending on their condition, segregated potatoes may be suitable for a different processing specification, animal feed or biogas production. The available outlet will depend on quality requirements and the reason the potatoes were rejected.

Recovery is particularly relevant where the sorting criteria remove potatoes that remain commercially usable but do not meet the specification of the main product stream. Separating those tubers allows their condition and value to be assessed without mixing them with foreign objects.

The required sorting sensitivity creates a balance between product purity and recovery. Setting an inspection threshold too aggressively can increase the loss of acceptable potatoes, while insufficient sensitivity may leave unwanted material in the accepted stream.

TOMRA reports foreign-material sorting accuracy of 92% under internal testing that included incoming defect loads of 20%. The result is a supplier test figure rather than an independently established guarantee of performance across every potato variety, field condition or machine configuration.

The stated maximum throughput of 100 tonnes per hour applies to pre-storage sorting. More detailed quality inspection when potatoes are removed from storage requires a lower operating rate, reflecting the additional classification work involved.

That distinction is important when specifying equipment capacity. An installation intended primarily to remove soil and stones may have different throughput requirements from one expected to identify a wider range of defects before potatoes enter a processing factory.

Handling conditions can also affect the usable product recovered from a sorting operation. Potatoes may be damaged during transfers or rejection, particularly where impact occurs at unsuitable speeds or against hard receiving surfaces.

The 4A incorporates handling arrangements intended to limit damage to acceptable potatoes, with optional cushioned ejection fingers available for additional protection. The performance of those features will depend on the complete receiving and discharge arrangement rather than the sorter alone.

Operator control is provided through TOMRA’s XCD touchscreen interface, which allows sorting programmes to be selected and adjusted. The same interface is used on other TOMRA sorting equipment, providing consistency for operators familiar with the company’s existing machinery.

TOMRA says the 4A can operate with one person attending the sorting line. That claim concerns the machine’s sorting operation and should not be interpreted as eliminating staffing requirements elsewhere in a potato receiving, handling or storage facility.

The equipment forms part of TOMRA’s wider food sorting portfolio, extending established optical inspection and automated rejection technologies into a system developed specifically for high-volume potato handling.

The supplier is also developing the platform for additional crops. Onion sorting is intended to become available on TOMRA 4A machines in time for the next harvest season, although that application is not part of the confirmed November potato launch.

Its first public exhibition at Interpom will provide an opportunity to demonstrate the machine’s inspection and separation arrangement. Installation-specific results will be needed to establish throughput, rejection accuracy, product damage and recovery performance under commercial operating conditions.

The launch adds another option for automated sorting before storage and subsequent quality grading, with the machine’s value dependent on the quality of incoming material, the product specifications applied and the proportion of usable potatoes recovered from the rejected streams.