| Tea is an economic crop,and China is the largest tea producer in the world.The moisture content of tea is an important index to evaluate the quality of tea in the manufacturing.Tea was distributed in stack shape,and its moisture content changed in real-time and matched process parameters in the continuous production process.Therefore,realizing real-time detection of tea moisture content is very necessary in the manufacturing.Due to time-consuming,laborious and destructiveness,the traditional detection method of tea moisture content is not suitable for online detection.In view of this situation,this paper developed an on-line detection instrument and established a rapid non-destructive detection method for tea moisture content based on spectral technology.The main results of this research are as follows:(1)Based on the visible-near infrared(Vis/NIR)spectroscopy,a real-time online detection equipment for tea moisture content was developed.The equipment was composed of a frame,a light source,a spectrometer,a Y type optical fiber,a quartz plate,a tea container and an electric push rod.At the same time,the frame is provided with a shell composed of an upper,lower,left,right,and back side plate.The equipment was applied to collect the spectral of Anji white tea and Zisun tea in each manufacturing stage(pluching fresh tea leaves→spreading→green removing→rolling→drying 1→drying 2).Among them,Anji white tea was collected in two batches and Zisun tea in one batch.The results show that the equipment can collect the spectral of tea and provide hardware support for achieving the real-time online detection of the moisture content of tea.(2)Partial least squares(PLS)was used to establish model for tea spectral and the moisture content of tea.In order to obtain the best modeling results,multiple scatter correction(MSC),moving average smoothing(MAS),standard normal variate transformation(SNV),baseline offset correction(BOC)and normalization were used to preprocess the original spectrum for eliminating the background and noise signals of the interference sources in the original spectral,such as the equipment,the experimental environment,etc.The processing results of the five preprocessing methods were evaluated by coefficient of determination(R~2)and root mean square error(RMSE).The results show that the modeling result after preprocessing by the normalization function was the best,which R~2 is greater than 0.97,RMSE is less than 0.05.The accuracy and stability of the model were high,and the equipment can achieve rapid non-destructive detection of tea moisture content.(3)In order to reduce the workload of establishing models in moisture content for different tea and the impact of testing environment,the direct standardization(DS)algorithm was used to correct the differences between Anji white tea and Zisun tea,and different batches of Anji white tea.So,the goal of using a model to detect the moisture content for all tea samples was achieved.After the correction of DS algorithm,the R~2 of models by mutual prediction in moisture content of Anji white tea and Zisun tea,and Zisun tea of different batches of was significantly increased(From 0.8962,0.7856,0.4118,0.7678,0.5553,0.9427 to 0.9457,0.9161,0.8505,0.9711,0.8857,0.9625,separately),while RMSE was reduced(From 0.0931,0.1253,0.2065,0.1304,0.1795,0.0692 to 0.0673,0.0783,0.1041,0.0491,0.0910,0.0524,respectively).DS algorithm can eliminate the differences between different varieties and batches of tea,the robustness and universality of models for tea moisture content were improved.The adaptability of the equipment to the detection of moisture content of tea in the actual production environment was obviously improved. |
PDF Full Text Request