| With the increasing demand for higher quality of agricultural products,including eggs and fruits,the automatic quality inspection and grading machine have been widely used in the production and processing industry of eggs and fruits.As an important grading factor,weight is used to measure the quality of eggs and fruits.As the demand increases for higher throughput and weighing accuracy,the commonly used low-pass filtering and a subsequent average-based mass estimation method did not obtain satisfactory weighing accuracy,therefore,it is of great significance to study the fast dynamic weighing data processing algorithm to improve the performance of the weight sorter.On account of the egg commercial processing equipment and fruit internal and external quality inspection equipment developed by our research group,an arc-shaped rail based dynamic egg weighing apparatus and a free pallet belt conveyor based dynamic spherical fruit weighing apparatus were developed.A CompactDAQ data acquisition card was applied to construct a data acquisition platform.A MATLAB GUI software was developed for data acquisition and processing.The dynamic behavior of egg and fruit dynamic weighing mechanism were deeply studied.The asymmetrically trimmed mean based egg dynamic weighing algorithm and multilayer perceptron(MLP)based egg and fruit dynamic weighing algorithm were studied using eggs and apples.The main research contents and results are as follows:1.Studying the dynamic behavior of egg and fruit dynamic weighing machine and the properties of vibration disturbance,including measuring and studying the natural frequency,studying the properties of vibration disturbance which came from no-load operation and the dymamic weighing process.The results are as follows:(1)Measuring and studying the natural frequency of dynamic weighing machine.The impulse excitation response method and Fourier transform were applied to measuring and studying its natural frequency.The results are as follows:1)The egg dynamic weighing mechanism had two natural frequency of 72 Hz and 118 Hz.2)The fruit dynamic weighing mechanism had one obvious natural frequency of 38 Hz and an unconspicuous natural frequency of 34 Hz.These results indicated that the fruit dynamic weighing mechanism was not an ideal second-order mass spring damping system.(2)Studying the properties of vibration disturbance that came from no-load operation of dynamic weighing machine.The results are as follows:1)The separating mounting platform of load cell module in egg dynamic weighing apparatus can effectively reduce the vibration noise.2)Although a higher processing speed clearly had more disturbances,the dominant frequencies showed little dependence on the processing speed or egg mass.It only affected the strength of vibrational frequency about 90 Hz at three processing speed.3)The processing speed direct impacted the intensity and position of vibrational frequency in fruit dynamic weighing machine.There were several obvious formants in vibration.4)There were some low frequency vibrating disturbance in the frequency range from 0 to 10 Hz that would be effected by processing speed.(3)Studying the properties of vibration disturbance during dymamic weighing process.The correlations of signal in the time and frequency domain of load cell and accelerometer were studied,then the smoothed-pseudo-Wigner-Ville distribution(SPWVD)was applied to perfom the time-frequency analysis of load cell and accelerometer signal.The results are as follows:1)There were strong correlation of time and frequency domain signal of load cell and accelerometer.2)The time domain and frequency domain signal of the no-load operation of fruit dynamic weighing machine also had strong correlation at 0.5 m/s and 0.9 m/s.3)The time domain and frequency domain signal of the no-load operation fruit dynamic weighing machine had low correlation at 0.7 m/s because of the resonance between the load cell mounting platform and the vibrating disturbance form the ground.4)The vibration noise of the no-load operation egg and fruit dynamic weighing machine was non-stationary.5)The vibration noise during the weighing process of egg and fruit dynamic weighing machine was also non-stationary and it was effected by the processing speed.2.This paper proposed a sorting-based weight estimation method(SME)that composed of digital filters and asymmetric trimmed mean,and the parameters were optimized using grid search.SME reduced the influence of strong vibration interference from the egg unconstrained rolling in the dynamic weighing rail that degraded the weighing accuracy.The results are as follows:(1)SME could significantly improve the dynamic egg weighing accuracy.Among the four digital filters with SME,the SME with FIR had the highest weighing accuracy.The average of the mean error was less than 0.02 g,the maximum mean error was 0.2 g,the maximum mean standard deviation of the error was 0.29 g,the maximum standard deviation of the error was 0.47 g,and the maximum measurement error was-0.92 g.(2)Compared with the average-based mass estimation methods(AME),the improvement ratio of mean error were 92.6%,86.6%,93%and 85.7%.And the improvement ratio of standard deviation of the error are 42.6%,49.7%,40.8%and 45.1%,respectively.(3)In terms of computational burden,the SME can be readily implemented.The execution time of the program of SME with FIR digital filer was 0.1313 ms.The execution time of the program of SME with IIR digital filer was 0.0572 ms.The execution time of the program of SME with ALMS adaptive digital filter was 2 ms.The execution time of the program of SME with TVLPF filter was 0.849 ms.Compared with the weighing time of 200 ms at 5 eggs/s,the proposed SME fully satisfy the real-time requirements of on-line weighing.Therefore,the SME is sufficiently capable of being an effective mass estimation method for dynamic weighing.3.Develping the method that saw the mass estimation prccess as the nonlinear fitting using the statistical distribution properties of dynamc weighing signal,and an two-layer feedforward artificial neural network,was employed to improve the egg dynamic weighing accuracy.The results are as follows:(1)MLP0 could significant improve the egg and fruit dynamic weighing accuracy.1)In egg dynamic weighing,the maximum error was-1.133 g,the mean of weighing errors was 0.025 g,the standard deviations of the error was 0.276 g.Compared with the average-based mass estimation methods,the improvement ratio of mean error were 90.5%,81.8%,and 85.5%.And the improvement ratio of standard deviation of the error were 55.4%,29.2%,and 39.3%,respectively.2)In fruit dynamic weighing,the mean of weighing errors of the test set was-0.126g,the standard deviations of the error was 1.374 g.Compared with the average-based mass estimation method(AME),the improvement ratio of mean error were 86.1%,89.5%,and 88.8%.And the improvement ratio of standard deviation of the error are 14.3%,36.5%,and 29.0%,respectively.(2)The neural network model MLP1 MLP2 and MLP3 could also significantly improve the egg and fruit dynamic weighing accuracy.The MLP0 and MLP1,MLP2,MLP3 had similar weighing accuracy,so the results could be concluded that the two-layer feedforward artificial neural network(MLP)can effectively improve the weighing accuracy of egg and fruit dynamic weighing at different processing speed. |
PDF Full Text Request