Abstract: Mechanical equipment is widely used in fields such as aerospace, rail transportation, energy and chemical engineering, and advanced manufacturing. Its condition monitoring technologies can help ensure the safe and reliable operation of various equipment. In recent years, research on intelligent machinery condition monitoring methods has received much attention. Especially, research on machinery condition monitoring based on deep learning has emerged and gradually become a hot topic. However, deep learning often requires massive amounts of high-quality data, and the abstract fault features it extracts lack physical interpretability, making it difficult to effectively solve the challenges faced in practical engineering condition monitoring scenarios, such as strong background noise, a large amount of health data with few fault data, a small overall data volume, and the urgent need for physically interpretable fault features to support monitoring diagnosis results and maintenance decisions. This has limited the widespread application of deep learning theory and methods in machinery condition monitoring. Therefore, it is urgent to carry out research on interpretable and trustworthy intelligent machinery condition monitoring. In recent years, newly proposed data-driven interpretable optimized weights have physical interpretability and can indicate the key information of fault frequencies, and a series of methods developed based on the optimized weights show great potential application value for practical intelligent machinery condition monitoring. Therefore, to further promote related research, this article aims to review and summarize the research of interpretable optimized weights theory and related methods, and discuss the future research directions of interpretable intelligent machinery condition monitoring.