基于机器视觉的生化池监测技术

张亮; 杨博文; 刘宇衡; 张雨; 高瑜; 李佳诚; 李俊辰; 林斯杰

doi:10.13205/j.hjgc.202503001

基于机器视觉的生化池监测技术

doi: 10.13205/j.hjgc.202503001

张亮^1, ,,
杨博文¹,
刘宇衡¹,
张雨¹,
高瑜¹,
李佳诚¹,
李俊辰²,
林斯杰^2,3

1. 北京工业大学环境科学与工程学院城镇污水深度处理与资源化利用技术国家工程实验室, 北京 100124;
2. 南方科技大学环境科学与工程学院, 广东深圳 518055;
3. 南方科技大学工程技术创新中心(北京), 北京 100083

基金项目:

国家重点研发计划“基于智能感知物联网的生态环境高密度监测关键技术研发”（2021YFC1809001）

详细信息

作者简介:
张亮（1986—），男，教授，主要研究方向为污水生物处理新技术和新工艺开发优化及智能控制。zliang@bjut.edu.cn

通讯作者:
张亮(1986—),男,教授,主要研究方向为污水生物处理新技术和新工艺开发优化及智能控制。zliang@bjut.edu.cn

计量
- 文章访问数: 159
- HTML全文浏览量: 39
- PDF下载量: 9
- 被引次数: 0
出版历程
- 收稿日期: 2024-11-29
- 录用日期: 2025-01-03
- 修回日期: 2024-12-10
- 网络出版日期: 2025-06-07
- 刊出日期: 2025-03-01

A monitoring technique for bioreactors based on machine vision

1. National Engineering Laboratory for Advanced Municipal Wastewater Treatment and Reuse Technology, College of Environmental Science and Engineering, Beijing University of Technology, Beijing 100124, China;
2. School of Environment Science and Engineering, Southern University of Science and Technology, Shenzhen 518055, China;
3. Engineering Innovation Center of SUSTech (Beijing), Southern University of Science and Technology, Beijing 100083, China

摘要

摘要: 生化池的在线监测对污水处理厂的稳定运行具有重要意义。现有生化池监测依赖流量计、传感器以及人工巡查和综合判断。为降低生化池人工巡检的工作量，提升监测的连续性和准确性，提出基于机器视觉的监测方案，并以曝气量预测为目标，考察了该技术方案的可行性。以曝气量1，2，3，4，5 L/min的小试规模生化池为预测对象，通过采集生化池的曝气图像、构建数据库、使用卷积神经网络提取特征、建立模型的监测框架，实现了污水处理过程中曝气量变化的自动感知。模型分析表明，测试集的预测精度达到99%，且模型预测精度稳定性较高，满足自动监测的需求。进一步考察机器视觉技术从小试装置迁移到中试规模生化池的可行性，证明了该方法在不同生化池中均具有可行性，表现出良好的应用潜力。该研究利用硬件（摄像头）和软件（机器学习模型），实现生化池运行关键信息的在线监测和异常识别，可实现对人工巡视的部分替代，可为污水处理厂的智慧化运行提供可行思路。
- 生化池监测 /
- 曝气量 /
- 图像数据 /
- 机器视觉 /
- 可视化
Abstract: In the operation and management of wastewater treatment plants， the online monitoring of bioreactors plays a critical role in the stable operation of the entire process. Currently， the monitoring of bioreactors mainly relies on flow meters and sensors to obtain relevant data， supplemented by manual inspection and comprehensive judgment. However， this traditional monitoring mode has certain limitations. Manual inspection consumes a lot of manpower， materials and time， and due to the intermittent nature of manual operation， it is difficult to ensure the continuity of monitoring. Furthermore， manual judgment may be influenced by subjective factors， thus affecting the accuracy of monitoring to a certain extent. Therefore， this study proposed a novel monitoring program based on machine vision， and set the aeration volume prediction as the core objective to examine the feasibility of this technical program in practical application. In the specific experimental process， a lab-scale bioreactor biochemical tank was selected as the prediction object， with an aeration volume in the range of 1 L/min to 5 L/min. Firstly， the camera was used to collect the aeration images of the bioreactor， which covered various state information of the tank affected by different aeration volumes. Subsequently， a special database was constructed to systematically organize and store the collected image data for subsequent analysis and processing. Then， the advanced convolutional neural network technology was used to extract features from the image data， and the key feature information closely related to the aeration volume was mined out. Finally， a corresponding model was established based on these extracted features to form a complete monitoring framework， realizing the accurate perception of aeration volume in the wastewater treatment process. The analysis of the model revealed that the prediction accuracy of the test set was as high as 99%， which fully demonstrated the high stability of the model and fully met the actual needs of automatic monitoring of wastewater treatment plants. In order to further expand the application scope of this technical program， this study also focused on the feasibility of migrating the machine vision technology from the lab-scale bioreactor to the pilot-scale bioreactor， which strongly proved that the method had good feasibility in both， and showed the good application potential of the machine vision technology in the field of wastewater treatment. In the whole research process， the using of hardware （camera） and software （machine learning model）， not only efficiently completed the task of online monitoring of the aeration volume， but also sent out alarm signals in time when abnormalities were detected. The realization of such a program could partially replace the traditional manual inspection work， greatly reducing the manual workload， while improving the efficiency and accuracy of monitoring. Thus， some valuable and feasible ideas for the development of wastewater treatment plants in intelligent operation were put forward in the hope of promoting the technological upgrading and innovation of the entire wastewater treatment industry.
- bioreactor monitoring /
- aeration volume /
- image data /
- machine vision /
- visualization

HTML全文

参考文献(33)

[1]	GRESCH M，ARMBRUSTER M，BRAUN D，et al. Effects of aeration patterns on the flow field in wastewater aeration tanks［J］. Water Research，2011，45（2）：810-818.
[2]	PIOTROWSKI R，UJAZDOWSKI T. Designing control strategies of aeration system in biological WWTP［J］. Energies，2020，13（14）：3619.
[3]	LI J，LIU Y，JIANG H，et al. A multi-view image feature fusion network applied in analysis of aeration velocity for WWTP［J］. Water，2022，14（3）：345.
[4]	VILLEZ K，VANROLLEGHEM P A，COROMINAS L. Optimal flow sensor placement on wastewater treatment plants［J］. Water Research，2016，101：75-83.
[5]	LIU C M，CHAI L P，LI Y，et al. Aeration state detection and prediction analysis based on image recognition and regression calculation［C］// Procedings of the 2022 China Automation，Xiamen，2022. 刘成明，柴立平，李跃，等. 基于图像识别与回归计算的曝气状态检测与预测分析［C］// 2022中国自动化大会，厦门，2022.
[6]	LI Y H，WANG X，ZHAO Z X，et al. Lagoon water quality monitoring based on digital image analysis and machine learning estimators［J］. Water Research，2020，172：115471.
[7]	KANG D，LIN Q，XU D，et al. Color characterization of anammox granular sludge：Chromogenic substance，microbial succession and state indication［J］. Science of the Total Environment，2018，642：1320-1327.
[8]	LEE W H，PARK C Y，DIAZ D，et al. Predicting bilgewater emulsion stability by oil separation using image processing and machine learning［J］. Water Research，2022，223：118977.
[9]	RAWAT W，WANG Z H. Deep convolutional neural networks for image classification：A comprehensive review［J］. Neural Computation，2017，29（9）：2352-2449.
[10]	KRAUSE L M K，KOC J，ROSENHAHN B，et al. Fully convolutional neural network for detection and counting of diatoms on coatings after short-term field exposure［J］. Environmental Science& Technology，2020，54（16）：10022-10030.
[11]	MORENO-RODENAS A M，DUINMEIJER A，CLEMENS F. Deep-learning based monitoring of FOG layer dynamics in wastewater pumping stations［J］. Water Research，2021，202：117482.
[12]	WU Y，ZHANG X，XIAO Y，et al. Attention neural network for water image classification under IoT environment［J］. Applied Sciences，2020，10（3）：909.
[13]	HASSAN S I，DANG L M，MEHMOOD I，et al. Underground sewer pipe condition assessment based on convolutional neural networks［J］. Automation in Construction，2019，106：102849.
[14]	BAI R C，WANG K，YU F W. Intelligent sorting system for household waste based on ResNet_Vd model［J］. Environmental Engineering，2023，41（S2）：1030-1033. 柏润泚，王凯，俞范文. 基于ResNet＿Vd模型的生活垃圾智能分拣系统［J］. 环境工程，2023，41（S2）：1030-1033.
[15]	LIU Z，GAO D M. Application and comparison of different deep learning models in the identification of kitchen waste types［J］. Environmental Engineering，2024，42（03）：254-260. 刘志，高东明. 不同深度学习模型在餐厨垃圾类型识别中的应用与比较［J］. 环境工程，2024，42（3）：254-260.
[16]	JIN P W，YAO Y，LIANG X Y，et al. Research progress on garbage image recognition［J］. Environmental Engineering，2022，40（1）：196-206. 金佩薇，姚燕，梁晓瑜，等. 垃圾图像识别研究进展［J］. 环境工程，2022，40（1）：196-206.
[17]	PYO J，CHO K H，KIM K，et al. Cyanobacteria cell prediction using interpretable deep learning model with observed，numerical，and sensing data assemblage［J］. Water Research，2021，203：117483.
[18]	LI Y Y，LIU H L. Prediction of total phosphorus in rivers using time convolutional network based on attention mechanism［J］. Environmental Engineering，2023，41（5）：163-171. 黎园园，刘海隆. 基于注意力机制的时间卷积网络河流总磷预测［J］. 环境工程，2023，41（5）：163-171.
[19]	YU S T，LIU P. Prediction of PM_2.5 concentration in Beijing based on long short term memory network-convolutional neural network（LSTM-CNN）［J］. Environmental Engineering，2020，38（6）：176-180，166. 于伸庭，刘萍. 基于长短期记忆网络-卷积神经网络（LSTM-CNN）的北京市PM_2.5浓度预测［J］. 环境工程，2020，38（6）：176-180，166.
[20]	IOFFE S，SZEGEDY C. Batch normalization：Accelerating deep network training by reducing internal covariate shift［C］// proceedings of the 32nd International Conference on Machine Learning，Lille，2015.
[21]	HINTON G E，SRIVASTAVA N，KRIZHEVSKY A，et al. Improving neural networks by preventing co-adaptation of feature detectors［J］. Computer Science，2012，3（4）：212-223.
[22]	KRIZHEVSKY A，SUTSKEVER I，HINTON G. ImageNet classification with deep convolutional neural networks［J］. Advances in neural information processing systems，2012，25（2）：84-90.
[23]	YANG Y. An evaluation of statistical approaches to text categorization［J］. Procedings of the AMIA Annual Fall Symposium，1999，1（1/2）：358-362.
[24]	ZEILER M D，FERGUS R. Visualizing and understanding convolutional networks［J］. European Conference on Computer Vision，2014，8689：818-833.
[25]	ZHOU B，KHOSLA A，LAPEDRIZA A，et al. Learning deep features for discriminative localization［C］// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition（CVPR），Seattle，2016.
[26]	HUANG Y H. Research and Implementation of Sludge Settling Ratio Recognition based on Computer Vision［D］. Hangzhou：Zhejiang Gongshang University，2023. 黄勇昊. 基于计算机视觉的污泥沉降比识别研究与实现［D］. 杭州：浙江工商大学，2023.
[27]	GAO Y T. Research on Data-Driven Intelligent Recognition and Control Optimization of Sludge Expansion［D］. Beijing：North China University of Technology，2024. 郜煜涛. 基于数据驱动的污泥膨胀智能识别与控制优化研究［D］. 北京：北方工业大学，2024.
[28]	WANG J H，LIAO W S，LI H M，et al. Data augmentation method supporting intelligent management of sewage treatment plants under data defect conditions［J］. Environmental Engineering，2024，42（6）：153-159. 王建辉，廖万山，李慧敏，等. 数据缺陷条件下支持污水处理厂智能管理的数据增强方法［J］. 环境工程，2024，42（6）：153-159.
[29]	WANG Y，SU X L，LIU J，et al. AI based sewage treatment monitoring and automatic regulation technology［J］. New Technologies and Products in China，2024，（13）：115-117，127. 王勇，苏晓亮，刘佳，等. 基于AI的污水处理监控与自动调节技术［J］. 中国新技术新产品，2024，（13）：115-117，127.
[30]	QI M，CHEN Y B，ZHANG X P，et al. Construction and application of a smart sewage treatment plant control platform based on multi-source information fusion［J］. Water Supply and Drainage，2020，56（1）：120-124. 齐鸣，陈燕波，张辛平，等. 基于多源信息融合的智慧污水处理厂管控平台建设与应用［J］. 给水排水，2020，56（1）：120-124.
[31]	WANG X S，CHEN Q S，SHEN Y W，et al. Multi-data fusion online diagnostic system for wastewater plant fan failures［J］. Industrial Control Computers，2022，35（1）：31-34. 汪喜生，陈秋忠，沈怡雯，等. 多数据融合的污水厂风机故障在线诊断系统［J］. 工业控制计算机，2022，35（1）：31-34.
[32]	XIAO L Z，HU F. Research on spam detection based on LGD-YOLO high-precision lightweight target detection network［J］. Environmental Engineering，2024，42（6）：169-177. 肖立中，胡凡. 基于LGD-YOLO高精度轻量化目标检测网络的垃圾检测研究［J］. 环境工程，2024，42（6）：169-177.
[33]	WANG L，YU K，CHEN H，et al. Application of unmanned inspection of water supply pipe network based on the fusion of fiber optic sensing and video AI technology［J］. Environmental Engineering，2023，41（11）：59-63. 王蕾，俞坤，陈辉，等. 基于光纤传感与视频AI技术融合的供水管网无人化巡检应用［J］. 环境工程，2023，41（11）：59-63.