Open Journal Systems

OTIMIZAÇÃO DOS PESOS DAS OBSERVAÇÕES GEODÉSICAS PELO PROBLEMA DE VALOR PRÓPRIO INVERSO

Reginaldo de Oliveira

Abstract



Nesta dissertação apresenta-se o problema de valor próprio inverso,
aplicado à otimização dos pesos das observações geodésicas. Discorre-se sobre o
emprego de valores próprios da matriz de covariâncias que dentre outras aplicações,
existe aquela que indica a qualidade das redes geodésicas. O problema de valor
próprio inverso amplia estas aplicações possibilitando a determinação dos pesos a
fim de que se escolha os instrumentos adequados para efetuar as mensurações. O
propósito principal do trabalho é avaliar a aplicabilidade de tal problema quando se
pretende que a rede geodésica apresente precisão pré-estabelecida e ainda apresente
condições aos valores próprios da matriz de covariâncias quando estes são tomados
como função objetivo em uma otimização de pesos. Apresentam-se ainda os
conceitos teóricos que conduzem à formulação e solução do problema de valor
próprio inverso através do método de Newton.

Otimizing geodetic observation weight by the use of the inverse eigen value

Abstract


In this dissertation, the inverse eigenvalue problem is presented and
applied to the optimization of the geodesic observation weights. It is explained the
use of the covariance matrix eigenvalues. Among several applications, the one that
indicates the quality of the geodesic networks is presented. The inverse eigenvalue
problem enlarges these applications making easier the determination of the weights
so that it is chosen suitable instrumentation to do measurements. The main purpose
of the work is the evaluation of the usefulness of such problem when it is intended
that the geodesic network show the pre-established precision besides the conditions
to the covariance matrix eigenvalues when they are taken as objective function in an optimization of weights. The theoretical concepts that drive to the formulation and
solution of the inverse eigenvalue problem through Newton's method are also
shown.