This paper reviews and maps the development of intelligent fault diagnosis (IFD) using machine learning theories, focusing on traditional machine learning, deep learning, and transfer learning. IFD aims to automatically recognize the health states of machines by applying machine learning theories, reducing reliance on human labor. Traditional machine learning methods, such as ANN, SVM, and kNN, were used in the past for IFD, involving data collection, feature extraction, and health state recognition. These methods required manual feature extraction and relied on expert knowledge, which limited their effectiveness with large datasets. In recent years, deep learning has transformed IFD by enabling end-to-end diagnosis models that automatically learn features from raw data, improving diagnosis accuracy. Deep learning techniques, such as stacked autoencoders (SAE), deep belief networks (DBN), convolutional neural networks (CNN), and ResNet, have been applied to fault diagnosis, allowing models to directly connect raw data to health states. These methods have significantly improved the performance of IFD in handling large and complex datasets. Looking ahead, transfer learning is expected to play a crucial role in IFD by leveraging knowledge from one or multiple tasks to improve diagnosis in new scenarios. Transfer learning approaches, such as feature-based, GAN-based, instance-based, and parameter-based methods, aim to overcome the challenges of limited labeled data in engineering applications. These methods can enhance the generalization ability of IFD models and improve their adaptability to different machine conditions. The paper systematically reviews the development of IFD from the past to the present, highlighting the evolution of IFD methods and their applications. It provides a roadmap for future research, emphasizing the potential of transfer learning in expanding the applications of IFD in engineering scenarios. The review also discusses the challenges and opportunities in IFD, offering insights into the future directions of research in this field.This paper reviews and maps the development of intelligent fault diagnosis (IFD) using machine learning theories, focusing on traditional machine learning, deep learning, and transfer learning. IFD aims to automatically recognize the health states of machines by applying machine learning theories, reducing reliance on human labor. Traditional machine learning methods, such as ANN, SVM, and kNN, were used in the past for IFD, involving data collection, feature extraction, and health state recognition. These methods required manual feature extraction and relied on expert knowledge, which limited their effectiveness with large datasets. In recent years, deep learning has transformed IFD by enabling end-to-end diagnosis models that automatically learn features from raw data, improving diagnosis accuracy. Deep learning techniques, such as stacked autoencoders (SAE), deep belief networks (DBN), convolutional neural networks (CNN), and ResNet, have been applied to fault diagnosis, allowing models to directly connect raw data to health states. These methods have significantly improved the performance of IFD in handling large and complex datasets. Looking ahead, transfer learning is expected to play a crucial role in IFD by leveraging knowledge from one or multiple tasks to improve diagnosis in new scenarios. Transfer learning approaches, such as feature-based, GAN-based, instance-based, and parameter-based methods, aim to overcome the challenges of limited labeled data in engineering applications. These methods can enhance the generalization ability of IFD models and improve their adaptability to different machine conditions. The paper systematically reviews the development of IFD from the past to the present, highlighting the evolution of IFD methods and their applications. It provides a roadmap for future research, emphasizing the potential of transfer learning in expanding the applications of IFD in engineering scenarios. The review also discusses the challenges and opportunities in IFD, offering insights into the future directions of research in this field.