Separating in shell pistachio nuts from kernels using impact vibration analysis

Authors: Haff, Ronald - Ron; Pearson, Thomas

Submitted to: Sensing and Instrumentation for Food Quality and Safety
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/8/2007
Publication Date: 8/23/2007
Citation: Haff, R.P., Pearson, T.C. 2007. Separating in shell pistachio nuts from kernels using impact vibration analysis. Sensing and Instrumentation for Food Quality and Safety. 1(4):188-192

Interpretive Summary: A prototype machine has been built that sorts pistachio nuts with shells and those without shells. The machine recognizes the type of nut by analyzing vibrations that occur when the sample falls onto a steel plate. Both the amplitude and frequency of the resulting signal are used in the analysis. If an in-shell nut is identified, an air nozzle is activated that diverts the sample. Testing was done at different speeds of 0.33, 10, 20, and 40 nuts per second using a mix of 10% in-shell and 90% kernels. The slower the nuts were fed through the machine the better the sorting accuracy, which ranged from 84% to 99% depending on the speed.

Technical Abstract: A sorting system has been developed for the separation of small in-shell pistachio nuts from kernels without shells on the basis of vibrations generated when moving samples strike a steel plate. Impacts between the steel plate and the hard shells, as measured using an accelerometer attached to the bottom of the plate, produce higher frequency signals than impacts between the plate and the kernels. Signal amplitudes, on the other hand, were highly variable and by themselves were not useful for the separation of samples. An algorithm was developed using both amplitude and frequency information to classify the signals. The algorithm activated an air nozzle to divert in-shell nuts away from the kernel stream. A prototype sorter was tested at throughput rates of 0.33, 10, 20, and 40 nuts per second using a mix of 10% in-shell and 90% kernels. At the lowest throughput rate, classification accuracies were 96% for in-shell nuts and 99% for kernels. For throughput rates between 10 nuts/s and 40 nuts/s, correct classification ranged from 84% to 90% for in-shell nuts. For kernels, accuracy was 95% at 10 and 20 nuts/s and 89% at 40 nuts/s.