Abstract:

Traditional coagulation experiment methods have the weakness of slow data collection, which is difficult to support the demand of deep learning for massive training data. The phenomenon of low accuracy, over fitting and weak generalization ability of the coagulation effect prediction model based on deep learning algorithm greatly limits its performance in the field of water treatment. Based on the process elements of traditional coagulation experiment, this paper proposed an automatic cup coagulation and data acquisition system with continuous operation function. The system was controlled by PLC and upper computer software. It could perform 36 rounds of cup tank coagulation experiment within 24 hours, and complete the collection of floc image and water quality parameter data. For a fixed dosage range(3-1 000 mL) and dosing time(10-38 s), the system could control the dosing error of peristaltic pump below 1% by introducing linear programming algorithm. The least square method was used to achieve the accurate fitting between PLC control current parameters and peristaltic pump speed. At the same time, the U-shaped liquid level balance water injection stabilization technology was used to reduce the repeated fluctuation error of water injection to less than 3%. Through the anti-aircraft bag algorithm, the system realized the stability and high-speed transmission of high-definition images, ensuring the stability of the system operation and the repeatability of the experimental results.

Key words: coagulation, linear programming algorithm, least square method, air defense package algorithm

摘要：

传统混凝实验手段存在收集数据较慢的弱点，难以支撑深度学习对海量训练数据的需求，这导致基于深度学习算法的混凝效果预测模型出现低准确率、过拟合和弱泛化能力现象，极大地限制其在水处理领域中的性能表现。基于传统混凝实验的工艺要素，提出一种具有连续运行功能的自动杯罐混凝与数据采集系统。该系统由可编程逻辑控制器（PLC）及上位机软件控制，24 h内可执行36轮杯罐混凝实验，同时完成絮体图像和水质参数数据的采集。对于固定的加药量范围（3~1 000 mL）和加药时长（10~38 s），通过引入线性规划算法，系统将蠕动泵的药剂投加误差控制在1%以下；利用最小二乘法，实现PLC控制电流参数和蠕动泵转速之间的精准拟合，同时使用U型液位平衡注水稳定技术，注水重复波动误差被降至3%以下；通过防空包算法，系统实现高清图像的稳定和高速传输，确保系统运行的稳定性与实验结果的可重复性。

关键词: 混凝, 线性规划算法, 最小二乘法, 防空包算法

CLC Number: 

• X703