Optimisation of an In-Line Chemcial Lean Instrument for the Australian Meat Industry

Summary

Background
The chemical lean of cartoned manufacturing meat is currently determined in most small and medium sized plants by drilling 10 cores from a random selection of cartons produced, pooling cores from groups of 5 to 10 cartons, mincing them and drying samples in a microwave oven.
Coring and analysis is both labour intensive and time consuming, leading plants to restrict sampling to the minimum required. There is a mandatory requirement in export plants to sample at least 5% of cartons of each type or grade of manufacturing meat produced. However adherence to this minimum requirement does not provide a statistically acceptable level of confidence that the product will meet the chemical lean specifications. This leads to the "giving away" of lean meat and to claims from customers for incorrect labelling, costing the industry millions of dollars.
Early Research (pre 2000)
The 1994 project RPDA.409a evaluated the use of time domain reflectometry equipment for the measurement of lean meat content.
Recent Research (after 2000)
In these projects, configuration of a CSIRO laboratory prototype sensor potentially suited to the measurement of chemical lean (CL) in meat trimmings on conveyors was completed. The sensor size was compatible with conveyors used in meat processing applications. The sensing volume also contained a non-conducting sheet used to simulate flat plastic conveyors used in the meat processing industry.
Initial tests on simple meat simulant samples confirmed signal detection in the sensor. While the CSIRO sensor had a very homogeneous response across a large volume, and was sensitive to relatively small trimming size, the variations in trimming shape that are likely to occur in a real application caused problems with obtaining a linear response from the sensor.
Attempts were made to linearise the response by changing the dimensions of the sensor, the configuration of active elements in the sensor, as well as excitation frequency. Unfortunately, in all cases studied, the variations due to shape differences ruined the correlation to CL that needed to be established for use in online control.
The attached reports describe the measurement problems encountered for variable sample shape, the reasons for failure of the CSIRO sensor configured for this application and a description of the generic problems likely to affect a range of different electromagnetic methods for measuring CL in trimmings.
In project A.TEC.0064, an online lower cost measurement solution for chemical lean, QVision, using Near Infra Red (NIR) technology, developed in Norway, was the subject of evaluation in this MLA study with Nortura. The NIR scanner was found to give reliable and accurate fat estimates of batches of beef trimmings, including when the batches are quite small. For the concept to work properly it was a requirement that the trimming or layers of trimmings on the belt are not too thick.