مدل‌سازی پارامترهای موثر بر دقت سامانه‌های اندازه‌گیری هدایت الکتریکی خاک به روش شبکه عصبی rbf در شرایط آزمایشگاهی

از جمله سامانه‌هایی که در تهیه نقشه هدایت الکتریکی خاک مزارع به‌کار می‌روند، سامانه‌های مبتنی بر روش تماس مستقیم الکترود با خاک می‌باشند. در این تحقیق با علم به اینکه علاوه بر شوری پارامترهای فیزیکی و شیمیایی خاک نیز در هدایت‌پذیری الکتریکی خاک تاثیرگذارند، به کمک روش شبکه عصبی rbf در طرح آماری باکس- بنکن به بررسی تاثیر پارامترهای اثرگذار بر نتایج روش تماس مستقیم در اندازه­گیری هدایت الکتریکی ظاهری خاک پرداخته و مدلی جهت تخمین هدایت الکتریکی واقعی خاک با داشتن هدایت الکتریکی ظاهری، دما، درصد رطوبت و چگالی توده تعیین شد. اندازه‌گیری همزمان پارامترهای موثر می‌تواند مرحله کالیبراسیون را حذف کند. مدل شبکه عصبی به‌دست آمده توانست به‌خوبی با ضریب تبیین 0.99، ece را تخمین بزند. ضمن بررسی الگوریتم‌های مختلف آموزش شبکه عصبی عملکرد الگوریتم آموزشی بیزین بهتر از سایر الگوریتم‌ها تشخیص داده شد. نتایج تحلیل حساسیت شبکه نشان داد به‌ترتیب متغیرهای eca، رطوبت، دما و چگالی توده بیشترین تاثیر را در تخمین مقدار ece خاک دارند، به‌طوری‌که با حذف آنها از مدل ضریب تبیین از 0.99 به‌ترتیب به 0.30، 0.35، 0.56 و 0.63 کاهش می‌یابد. پس از مرحله مدل‌سازی، مدل شبکه عصبی به‌دست آمده با یک گروه داده مزرعه‌ای مورد اعتبارسنجی قرار گرفت. نتایج اعتبارسنجی مدل ضریب تبیین 0.986 بین خروجی مدل و مقادیر ece اندازه‌گیری شده در آزمایشگاه را نشان داد. بدین ترتیب با استفاده از این مدل ضمن اندازه‌گیری هم‌زمان پارامترهای ذکر شده همراه با هدایت الکتریکی می‌توان دقت سامانه‌های اندازه‌گیری هدایت الکتریکی ظاهری خاک در تخمین و تهیه نقشه‌های شوری خاک افزایش داد. همچنین با توجه به عدم نیاز به داده‌برداری مجدد جهت کالیبراسیون سامانه‌ها، استفاده از این مدل زمان تحلیل داده‌ها و هزینه تهیه نقشه هدایت الکتریکی خاک را کاهش می‌دهد.

Modeling the Effective Parameters on Accuracy of Soil Electrical Conductivity Measurement Systems Using RBF Neural Network

<p > <strong >Introduction </strong > </p > <p >Presently, the loss of ground water levels and the increase in dissolved salts have given importance to the determination of salinity and the management of their variations in irrigated farms. Soil electrical conductivity is an indirect method to measure soil salts. The direct electrode contact method (Wenner method) is one of the widely used methods to rapidly measure soil EC <sub >a </sub > in farms. However, soil scientists prefer soil actual electrical conductivity (saturated extract electrical conductivity) (EC <sub >e </sub >) as an indicator of soil salinity, though its measurement is only possible in the laboratory. The aim of this study was to find a relationship between the prediction of soil actual electrical conductivity (EC <sub >e </sub >) in terms of temperature, moisture, bulk density and apparent electrical conductivity of soil (EC <sub >a </sub >). Thereby, the estimation of EC <sub >e </sub > would allow the partial calculation of EC <sub >a </sub > that is dependent upon soil salinity and dissolved salts. </p > <p > <strong >Materials and Methods </strong > </p > <p >This study used RBF neural network in BoxBehnken statistical design to explore the impacts of effective parameters on direct contact method in the measurement of soil EC <sub >a </sub > and provided a model to estimate EC <sub >e </sub > from EC <sub >a </sub >, temperature, moisture content and bulk density. In this study soil apparent electrical conductivity (EC <sub >a </sub >) was measured by direct contact (Wenner) method. The present study considered four most effective factors: EC <sub >a </sub > (saturated paste extract EC), moisture, bulk density, and temperature (Baradaran Motie <em >et al </em >., 2010). Given the characteristics of farming soils in Khorasan Razavi Province (Iran), the maximum and minimum of each independent variable were assumed as 0.56 mS.cm <sup >1 </sup > for EC <sub >e </sub >, 525% for moisture content, 11.8 g.cm <sup >3 </sup > for bulk density, and 237°C for soil temperature. Considering the experimental design, three moisture levels (5, 15 and 25%), three salinity levels (0.5, 3.25 and 6 mS.cm <sup >1 </sup >), three temperature levels (2, 19 and 37°C) and three compaction levels with bulk densities of 1, 1.4 and 1.8 g.cm <sup >3 </sup > were assumed in 27 trials with predetermined arrangement on the basis of BoxBehnken technique. 13 common algorithms were explored in MATLAB software package for the training of the artificial neural network in order to find the optimum algorithm (Table 4). The input layer of the network designed by integrating a Randomized Complete Block Design (RCBD) with kfold crossvalidation. Using kfold crossvalidation, 20 different datasets were generated for training and validation of RBF neural network. </p > <p > <strong >Results and Discussion </strong > </p > <p >A combination of an RCBD and kfold crossvalidation was used. The results of both training and validation phases should be considered in the selection of training algorithm. In addition, R <sup >2 </sup > of T1 training algorithm had a much lower standard deviation than other training algorithms. The lower standard deviation is, the more capable the algorithm would be in learning from different datasets. Considering all aspects, trainbr (T2) training algorithm was found to have the best performance among all 13 training algorithms of the neural network. Table 7 tabulates the results of means comparison for R <sup >2 </sup > of RBF model for both training and validation phases resulted from the application of some combinations of S and L2 factors as interaction. As can be observed, R <sup >2 </sup > = 0.99 for all of them with no significant difference. However, the magnitude of order differed between training and validation phases. Given the importance of the training phase, L2=9 and S=0.1 were regarded as the optimum values. </p > <p >The sensitivity analysis of the network revealed that soil ECa, moisture, bulk density, and temperature had the highest to lowest impact on the estimation of soil EC <sub >e </sub >, respectively. This model can improve the precision of soil EC <sub >a </sub > measurement systems in the estimation and preparation of soil salinity maps. Furthermore, this model can save in time of data analysing and soil EC mapping because it does not need data recollection for the calibration of systems. A validation prose was done with a 60 field collected data set. The results of validation show R <sup >2 </sup >=0.986 between predicted and measured ECa. </p > <p > <strong >Conclusions </strong > </p > <p >The present research focused on improving the precision of soil EC <sub >e </sub > measurement on the basis of easily accessible parameters (EC <sub >a </sub >, temperature, moisture, and bulk density). In conventional methods of soil EC mapping, the systems only measure soil EC <sub >a </sub >and then calibrate it to EC <sub >e </sub > by collecting some samples and using statistical methods. In this study, Soil EC <sub >e </sub > was estimated with R <sup >2 </sup > = 0.99 by a multivariate artificial neural network model with the inputs, including EC <sub >a </sub >, temperature, moisture, and bulk density of soil without any need to collect further soil samples and calibration process. The Bayesian training algorithm was introduced as the best training algorithm for this neural network. Thereby, soil EC variation maps can be prepared with higher precision to estimate the spatial spread of salinity in farms. Also, the results imply that soil EC <sub >a </sub >, moisture, bulk density and temperature have the highest to lowest effectiveness on the estimation of soil EC <sub >e </sub >, respectively. </p > <p >   </p >