Due to the great challenges of measuring forces directly, identifying dynamic loads based on accessible responses is a crucial problem in engineering, helping ensure integrity and reliability of mechanical structures. Dynamic load identification is a difficult inverse problem due to matrix ill-posedness, noise-sensitivity and computational scale, especially in uncertain structures. Unexpected inaccurate or non-unique solutions may be found if these problems are not well addressed. During the past decades, many methods have been proposed to deal with these problems. This paper tries to provide a comprehensive review of techniques for dynamic load identification, including under ill-posedness and uncertain parameter processing; with an emphasis on the statistical, data science, machine learning, and artificial intelligence aspects. Classical physics-based dynamic load identification theories in frequency and time domain are also introduced. Research challenges and prospects of dynamic load identification in mechanical systems are discussed finally. This review may offer guidelines for dynamic load identification in practical complex structures, as well as possibilities for further researches. Some methods could have broader applicability to other inverse problems.

由于直接测量力的巨大挑战，基于可访问的响应识别动态载荷是工程中的关键问题，有助于确保机械结构的完整性和可靠性。由于矩阵不适定性，噪声敏感性和计算规模，动态载荷识别是一个困难的逆问题，尤其是在不确定的结构中。如果不能很好地解决这些问题，则可能会发现意外的不准确或不唯一的解决方案。在过去的几十年中，已经提出了许多方法来解决这些问题。本文试图对动载荷识别技术进行全面综述，包括不适定和不确定参数处理。重点放在统计，数据科学，机器学习和人工智能方面。还介绍了基于古典物理学的频域和时域动态载荷识别理论。最后讨论了机械系统中动态载荷识别的研究挑战和前景。这篇综述可以为实际复杂结构中的动态载荷识别提供指导，并为进一步的研究提供可能。某些方法可能对其他反问题具有更广泛的适用性。