Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: schemat Voronoi

Sortuj według:

Ogranicz wyniki do:

Quantification of Individual Tree Competition Index Taking Chinese-Fir Plantations in Subtropical Low Hilly Area as an Example

Zhang Yifei, Deng Xiangwen, Huang Yaqi, Li Yanfei, Xiang Wenhua, Yan Wende

Polish Journal of Ecology

2019

Vol. 67, nr 1

1--16

Competition among trees is a fundamental interaction process within plant community, which is the theoretical basis of thinning. Plant competitive intensity is generally measured using a competition index (CI) that can be classified into two major categories: distance-independent and distance-dependent. The current study used Cunninghamia lanceolata (Lamb.) Hook as the test subject and used Hegyi's CI (distance-dependent), to quantify individual CI and their relationship with tree diameter at breast height (DBH). Five different criteria were used to select potential competitors for the calculation of CI. Seven basic linear and nonlinear mathematical functions were used to test and quantify the relationships between DBH of the target tree and the individual CI. Results showed that individual CI was negatively correlated with target tree DBH: as DBH increased, competition intensity weakened. The adjusted R2 with five different criteria of selection competitors simulated by seven functions ranged from 0.30 to 0.82. Considering the root mean square error (RMSE), P-value, and adjusted-R2, our results suggested that the best model to simulate the relationship between individual CI and focal tree DBH was power function (CI = 43.98 × DBH-1.08, adjusted R2 = 0.81) and with the Voronoi diagram method as the criteria for selecting competitors. These results can demonstrate a clearer understanding of the spatial structure of forests, and can be used to guide the selection of thinning trees in the process of thinning practice.

Analiza dwuczynnikowego planu doświadczenia nad siatką nieregularną generowaną schematem Voronoi przy aproksymatorze nauronowym

Pietraszek J.

Górnictwo Odkrywkowe

2003

R. 45, nr 2-3

21-24

Stosowana zazwyczaj w planowaniu eksperymentu metodologia powierzchni odpowiedzi (ang. response surface methodology) zakłada w ogromnej większości przypadków użycie aproksymatorów wielomianowych oraz planów eksperymentu utworzonych na podstawie złożonych reguł statystycznych. W sytuacji, gdy nie jest możliwe modelowanie powierzchni odpowiedzi przy pomocy pojedynczego płata wielomianowego należy jako aproksymatora użyć sieci neuronowej. Aproksymatory neuronowe wymagają jednak odmiennych procedur identyfikacji planu eksperymentu. W2 niniejszym artykule autor proponuje zastosowanie do tego celu powszechnie znanego schematu generowania wieloboków Voronoi, dzięki czemu możliwe jest iteracyjne utworzenie żądanego planu eksperymentu. Rozważany jest przypadek identyfikacji planu doświadczenia dla dwuwyjściowego obiektu badań.

Traditional response surface methodology is based on polynomial approximator and designs prepared according to statistical requirements. The fundamental of this methodology were founded by Robbins and Manro, Kiefer and Wolfowitz. Academic methodology initially developed for industrial utilization was formalized in the last years and coded as ISO 3534. In those cases when response surface cannot be modeled with single polynomial plane, a neural network approximator should be used. It is very useful in industrial applications where theoretical model is unavailable and the errors of empirical polynomial model are too large. If data obtained from model do not satisfy the assumed error level in comparison to measures from verification experiments, it is necessary to enhance the experiment design. These neural approximators require a different iterative procedure of design enhancing. The autor proposed to use an area splitting based on well-know scheme of generating the Voronoi's polygons first described by Voronoi in year 1908.