Simultaneous reconstruction of the perfusion coefficient and initial temperature from time-average integral temperature measurements. Applied Mathematical Modelling, 68. pp. 523-539.
AbstractInverse coefficient identification formulations give rise to some of the most important mathematical problems because they tell us how to determine the unknown physical properties of a given medium under inspection from appropriate extra measurements. Such an example occurs in bioheat transfer where the knowledge of the blood perfusion is of critical importance for calculating the temperature of the blood flowing through the tissue. Furthermore, in many related applications the initial temperature of the diffusion process is also unknown. Therefore, in this framework the simultaneous reconstruction of the space-dependent perfusion coefficient and initial temperature from two linearly independent weighted time-integral observations of temperature is investigated. The quasi-solution of the inverse problem is obtained by minimizing the least-squares objective functional, and the Fréchet gradients with respect to both of the two unknown space-dependent quantities are derived. The stabilisation of the conjugate gradient method (CGM) is established by regularising the algorithm with the discrepancy principle. Three numerical tests for one-and two-dimensional examples are illustrated to reveal the accuracy and stability of the numerical results.
IntroductionThe inverse problem of identifying the space-dependent perfusion/radiative coefficient from integral observation was previously studied in [1,2,3]. This unknown coefficient was numerically determined in the one-dimensional bio-heat equation with heat flux or time-average temperature measurement by minimising the Tikhonov regularisation functional using the NAG routine E04FCF together with the finite-difference method (FDM), [4]. Recently, the space-dependent perfusion coefficient was recovered by the CGM from the final or time-average temperature measurement in [5]. Also, the inverse problem of determining the initial temperature from temperature measurements at a later time was extensively studied, e.g. [6,7]. Besides, there are many numerical techniques that had been developed to reconstruct the unknown initial temperature, including the iterative CGM [8,9], the boundary element method (BEM) with regularisation [10], the elliptic approximation together with the BEM [11], the Tikhonov regularisation approach [12], the Fourier regularisation method [13] and the self-adaptive Lie-group adaptive method [14].In [15], the space-dependent radiative coefficient and the initial temperature were simultaneously reconstructed from temperature measurements at a fixed time θ > 0 and in ω × (0, T ), where ω is a subregion of the space domain Ω; the stability of the inverse problem was established, the existence of the minimizer of Tikhonov's first-order regularisation functional was proved, and the numerical results were obtained by using a nonlinear gradient multigrid technique. Similarly, t...