33 Bull. Min. Res. Exp. (2020) 161: 33-47 2D inverse modeling of the gravity feld due to a chromite deposit using the Marquardt’s algorithm and forced neural network Ata ESHAGHZADEH *a , Sanaz SEYEDI SAHEBARI b and Alireza DEHGHANPOUR c a Danesh Tadbir Zima Institute, Chaloos, Iran b Roshdiyeh Higher Education Institute, Tabriz, Iran. c Islamic Azad University, Science and Research Branch, Tehran, Iran. Research Article Keywords: Chromite deposit, Finite vertical cylinder, Forced Neural Networks, Gravity, Marquardt’s algorithm. Received Date: 20.09.2018 Accepted Date: 20.01.2019 ABSTRACT In this paper, two modeling method are employed. First, a method based on the Marquardt’s algorithm is presented to invert the gravity anomaly due to a fnite vertical cylinder source. The inversion outputs are the depth to top and bottom, and radius parameters. Second, Forced Neural Networks (FNN) for interpreting the gravity feld as try to ft the computed gravity in accordance with the estimated subsurface density distribution to the observed gravity. To evaluate the ability of the methods, those are employed for analyzing the gravity anomalies from assumed models with different initial parameters as the satisfactory results were achieved. We have also applied these approaches for inverse modeling the gravity anomaly due to a Chromite deposit mass, situated east of Sabzevar, Iran. The interpretation of the real gravity data using both methods yielded almost the same results. * Corresponding author: Ata ESHAGHZADEH, eshagh@alumni.ut.ac.ir Citation Info: Eshaghzadeh, A., Sahebari, A.S., Dehghanpour, A. 2020. 2D inverse modeling of the gravity feld due to a chromite deposit using the Marquardt’s algorithm and forced neural network. Bulletin of the Mineral Research and Exploration, 161, 33-47. https://doi. org/10.19111/bulletinofmre.589224 Bulletin of the Mineral Research and Exploration http://bulletin.mta.gov.tr BULLETIN OF THE MINERAL RESEARCH AND EXPLORATION CONTENTS Foreign Edition 2020 161 ISSN : 0026-4563 E-ISSN : 2651-3048 1. Introduction Non-uniqueness is a common problem in the inverse modeling of the residual gravity anomaly. IT can assign a set of the measured gravity feld data on the ground to the geometrical distributions of the subsurface mass with various shapes or physical parameters such as density and depth. One way to eliminate this ambiguity is to put a suitable geometry to the anomalous body with a known density followed by inversion of gravity anomalies (Chakravarthi and Sundararajan, 2004). Although simple models may not be geologically realistic, they are usually are suffcient to analyze sources of many isolated anomalies (Abdelrahman and El-Araby, 1993a.b). The interpretation of such an anomaly aims essentially to estimate the parameters such as shape, depth, and radius of the gravity anomaly causative body such as geological structures, mineral mass and artifcial subsurface structures. Several graphical and numerical methods have been developed for analyzing residual gravity anomalies caused by simple bodies, such as Saxov and Nygaard (1953) and Bowin et al. (1986). The methods include, for example, Fourier transform (Odegard and Berg, 1965; Sharma and Geldart,1968); Mellin transform (Mohan et al., 1986); Walsh transforms techniques (Shaw and Agarwal, 1990); ratio techniques (Hammer, 1974; Abdelrahman et al., 1989); least- squares minimization approaches (Gupta, 1983; Lines and Treitel, 1984; Abdelrahman, 1990; Abdelrahman