On-line sorting of meat trimmings into targeted fat categories

A system for on-line sorting of meat trimmings into categories with different fat levels was developed and tested by simulations and pilot-plant trials. The system consists of a conveyor belt, a NIR imaging scanner (QV500, Tomra Sorting Solutions, Asker, Norway), a flow weigher and grader (both Marel hf, Iceland) and a host computer containing synchronising software and a sorting algorithm. The sorting algorithm is based on desirability functions, which makes it flexible when it comes to selecting number of categories, target values, limits for deviations and other restrictions. The results showed that the sorting algorithm works when the fat measurements are accurate, giving deviations from target lower than the selected ±1 percentage point limits. In reality there are some inaccuracies in the on-line fat measurements due to inhomogeneous meat trimmings. This leads to a systematic under-estimation of the fat percentage in low-fat categories and over-estimation in the high-fat categories. These biases can be reduced by e.g. improving the on-line fat measurement technology. However, simulations showed that the bias for either category was generally low (below 2 percentage points) and the current system therefore has potential for on-line implementation.

► An on-line system for fat-dependent sorting of meat trimmings was developed.
► The sorting algorithm was based on desirability functions.
► The system was tested by both simulations and pilot-plant trials.
► The sorting algorithm worked perfectly when fat measurements were accurate.
► Inaccuracies in fat measurements led to a small, systematic bias.