Machine Learning (ML) has proved to be successful at identifying and representing underlying relationships in large data sets which would be difficult to process manually. However, the large amounts of data required for unsupervised learning mean that these traditional approaches encounter problems where data is sparse. In addition, these models are often used with insufficient regard for the details of the underlying optimization process. This poses a problem in engineering where the ability to explain model predictions (explainability) is often a prerequisite. There is a particular issue where ML methods may reach a conclusion which does not agree with existing physical understanding. Further, for problems where some of the underlying physics is already known, the traditional ML approach is effectively using large data sets to “re-learn” existing physical understanding. A potential solution to these issues is the incorporation of physical domain knowledge into the model or its training process to produce Informed Machine Learning. This paper provides an overview of the current state of informed machine learning for application in engineering. Firstly, the definition of explainable machine learning is explored. A selection of methods that incorporate physical priories into the machine learning pipeline is then described, leading to a review of current applications of informed machine learning in engineering. As a result of this analysis, a taxonomy is developed which provides a potential path for method development.