<sec>Optical phased array (OPA) technology, as a pivotal component of laser detection and ranging (LiDAR) systems, plays a crucial role in augmenting the application efficiency in fields such as autonomous driving, precision measurement, and remote sensing detection. With the escalating demands for high-resolution imaging, the array size of OPAs is continuously expanding, imposing higher requirements on the calibration precision and efficiency of the output beam. Existing calibration algorithms, such as the simultaneous perturbation stochastic gradient descent (SPGD) and the Gerchberg-Saxton (GS) algorithm, often face challenges of prolonging calibration times and insufficient precision when dealing with large-scale OPA systems.</sec><sec>In order to address this problem, our study introduces the Adam optimization algorithm, renowned for its adaptive learning rate feature, into the calibration process of OPA output beams. Through simulation modeling and experimental validation, this work comprehensively examines the differences in performance between the Adam algorithm and conventional SPGD and GS algorithms in beam calibration, especially under various OPA array configurations. For a 16×16 OPA array, the application of the Adam algorithm significantly enhances the peak side lobe ratio (PSLR) to over 15.98 dB, while notably reducing the number of iterations to less than 600, thereby shortening the calibration cycle and improving calibration precision effectively.</sec><sec>Furthermore, this work provides an in-depth analysis of parameter selection, convergence speed, and stability of the Adam algorithm in OPA calibration, offering detailed guidance for achieving more efficient and high-quality beam calibration. Through comparative analysis, this work not only demonstrates the substantial advantages of the Adam algorithm in enhancing OPA calibration efficiency, reducing calibration duration, and optimizing output beam quality but also emphasizes its critical role in advancing OPA technology.</sec><sec>The main contribution of this work lies in providing an innovative algorithmic approach for achieving efficient calibration of OPA output beams, which has important theoretical and practical significance for advancing the LiDAR technology, particularly in the field of high-precision beam control. Moreover, by applying optimized algorithms, this study not only improves the performance of OPA technology within existing domains but also paves new ways for its application in emerging fields such as optical communication, optical networking, and high-resolution imaging.</sec>