| Changes in the water quality of aquaculture water bodies have a greater impact on the growth of aquatic products.According to"Environmental Quality Standards for Surface Water",water quality can be divided into I to V according to the content of various elements in the water body,of which,the water body suitable for aquaculture is III to IV,whose main factors are ammonia nitrogen,total nitrogen,total phosphorus,COD,etc.In the Aquaculture process,timely detection of water quality factors and timely adjustment according to the changes in aquaculture is very significant.However,the present water quality testing methods commonly used such as the volumetric analysis method,spectrophotometry,etc.need to have testing techniques for testing personnel in the laboratory,and a method can only test a parameter,time-consuming and laborious.Electronic tongue is a new electrochemical detection means,through the combination of multi-sensor array and pattern recognition method,can realize the qualitative detection of liquid and multi-parameter quantitative detection.Voltammetric electronic tongue in recent years in the field of water quality monitoring has a number of applications.However,the existing electronic tongue system is generally determined at the beginning of the design of the sensor array,the detection process can not replace the sensor,the subsequent removal of the sensor is not convenient;detection process requires frequent sample replacement according to the detection needs,time-consuming and labor-intensive;secondly,the existing electronic tongue system sensor cleaning method is mostly submerged cleaning,time-consuming and requires multiple cleaning to ensure the cleaning effect.All these factors restrict the automation and detection efficiency of electronic tongue.To address these problems,this paper designs and develops an automatic electronic tongue system with automatic sensor replacement,automatic sample cleaning,and automatic qualitative and quantitative testing of farmed water based on farmed water testing.The main study includes.(1)In response to the problems of inconvenient sensor replacement and low automation such as inability to automatically feed and clean the electronic tongue system,an automatic electronic tongue system for aquaculture water testing is designed,including automatic detection module,automatic sample feeding and cleaning module,automatic sensor replacement module,data collection module and the upper and the lower computer control module as the control system.Among them,the upper computer is built based on Lab VIEW technology;the lower computer communicates directly with the hardware devices of each module to control the work of each module device;the automatic detection module controls the movement of the array platform by controlling the ball screw movement to realize automatic detection and sample tank cleaning;the sample feeding and cleaning module realizes the sample feeding in the detection tank by controlling the solenoid valve opening and closing in the pipeline system and the movement of the ball screw,as well as realizing the sensor feeding in the cleaning tank.The automatic sensor replacement module uses a five-axis robotic arm to realize the clamping and installation of the sensor,and the automatic sensor wire collection and discharge device combined with a relay to realize the automatic collection and discharge of the sensor wire;the data acquisition module mainly uses a data acquisition module to transfer the acquired data to the host computer through the USB port,so as to facilitate the next step of data processing.The data acquisition module mainly uses the data acquisition card to transfer the collected data to the host computer through the USB interface for the next step of data processing.(2)The automatic replacement sensor module of the automatic electronic tongue system was tested for the performance of the classification of the electronic tongue system by testing the collected culture water samples and collecting data using the electronic tongue system.The kinematic analysis of the robotic arm in the automatic sensor replacement module and the workspace animation using Monte Carlo method were used to locate the location of the robotic arm in the automatic sensor replacement module.Cartesian spatial trajectory planning of the robotic arm was performed,and the smooth operation process of the robotic arm was determined by analyzing the operation results of joint angle,angular velocity,and angular acceleration.Further,the clamping and installation accuracy test was conducted,and it was determined that the clamping and installation accuracy of the robotic arm was within 3mm,which met the expected target,and the overall clamping and installation time of the sensor was about 20S,which met the design requirements.The collected farming water samples were classified into four categories according to the Surface Water Environmental Quality Standard,and after data collection using the designed automatic electronic tongue for 25 sets of collected button crab farming water samples,PCA analysis was further carried out,and the results showed that the proposed automatic electronic tongue could successfully discriminate between different types of button crab farming water.(3)Based on the button crab farming water data collected by automatic electronic tongue and the physicochemical indexes of button crab farming water,the qualitative and quantitative analysis of button crab farming water was realized by combining the convolutional neural network(CNN)framework.Based on the Surface Water Environmental Quality Standard,the collected 25 sets of button crab farming water samples were classified into four water quality categories of II,III,IV and V.The qualitative detection model was built by using Alex Net-CNN structure and combined with different optimization means to qualitatively analyze the button crab farming water,and the results were compared with GA-BP qualitative analysis,and the results showed that the combination of Re Lu6 and BN normalization layer was used to The Alex Net-CNN qualitative detection model built by the optimization method achieved 95.5%recognition rate for four types of water quality samples,which is better than the qualification analysis results of GA-BP.The single-output quantitative analysis model and the multi-output quantitative analysis model were built separately using Inception V2-CNN structure,and the four main water quality indicators of button crab culture water,such as total nitrogen,ammonia nitrogen,total phosphorus and CODMn,were forecasted and the quantitative prediction results were compared with PLS,and the results showed that the validation set with single-output Inception V2-CNN quantitative analysis model RMSE ranged from 0.111 to 0.139,and R~2 ranged from 0.974 to 0.991;the validation set with multi-output Inception V2-CNN quantitative analysis model RMSE ranged from0.134 to 0.185,and R~2 ranged from 0.977 to 0.989,all of which were better than the prediction results of PLS quantitative analysis model.The analysis results demonstrate the feasibility of automatic electronic tongue combined with convolutional neural network for farmed water detection. |
PDF Full Text Request