重量選別機重量量測系統性能之改良

Abstract

蔬果重量分級機是蔬果集貨包裝場中最常見的機器。雖然應用荷重元等電子式重量秤重商品很多，但在台灣電子式的蔬果重量分級機仍很少見。因重量分級機機構原件的運動與機台之振動嚴重影響到重量動態量測精準度。本論文中，在裝置荷重元為重量感測器的蔬果重量選別機進行動態量測。量測所得訊號作快速傅立葉轉換，識別出實際重量訊號分佈於1Hz∼15Hz；而機台振動雜訊分佈於30Hz附近。研究中使用以下兩種方法以改善動態量測誤差。首先，使用電路模擬軟體Multisim模擬並設計雙T電橋帶拒濾波器，其濾波截止頻率為30Hz，利用電阻電容與運算放大器等原件實現雙T電橋帶拒濾波器再裝置於動態量測系統進行實驗。實驗結果發現，高品質因數（Ｑ＝0.65）帶拒濾波器有最佳的量測誤差改善結果，其動態量測誤差可改善至±5％以內。第二種方法是利用Simulink模式設計反摺機濾波器。在Simulink環境下模擬並實現動態量測輸出訊號經反摺積濾波器的濾波結果。利用數位低通濾波器將輸出訊號作平滑化處理以過濾輸出訊號中的高頻雜訊，再將平滑化後的訊號輸入至反摺積濾波器作濾波處理。濾波結果發現，反摺積濾波器可將動態量測誤差改善至±2％的動態量測誤差，此性能可以滿足商用蔬果重量分級機的需求。

Weighting sorter is one of the most popular machinery found in produce packers. Although load cell has been commercially applied in weight measurement for a long time, but in Taiwan it is still limited to be used in produce weighting sorter. Noise occurred by movement of mechanical elements and vibration of frame of the sorter greatly interference the weight measurement accuracy. In the thesis, by the use fast Fourier transform for the dynamic measurement signals of the load cell in a commercial sorter were identified as weighting signals distributed from 1 to 15Hz and the vibration noise mainly around 30Hz. Two methods were then implied to reduce the vibration noise in order to decrease the dynamic measurement error. First, a commercial software Multisim was used to simulate and design Twin-T band reject filters with cut-off frequency at 30Hz, filters were then realized through RC and Op-Amps circuits and were examined in dynamic measurements. The high quality Twin-T band reject filter with Q=0.65 gave the best results, the dynamic measurement error were lowered to ±5％. The second approach was to design a deconvolution filter using Simulink schema. Based on the dynamic response of the measurement system a deconvolution filter was simulated and materialize in the Simulink environment. The dynamic measurement signals were first digitally smoothed to eliminate high frequency noise and then processed with the deconvolution filter. With these process the dynamic measurement error were lowered to ±2 ％, that is fairly acceptable in the commercial measurement for a produce weighting sorter.