Online tool wear monitoring is an important component of intelligent milling. Integral end mill is one of the typical high-value cutting tools which has been widely used in aerospace, automobile, mold and other industries. Its cutting edge may produce inhomogeneous wear after suffering variable cutting depth experience. The existing methods are mainly focused on monitoring the maximum value of the tool wear, which cannot identify the inhomogeneous wear state and results in insufficient accuracy and practicality. Therefore, a physics-informed method is proposed to online identify inhomogeneous tool wear state. Firstly, a milling force mechanism model considering tool wear is established. The force model are expressed with matrix formulation so that the time-domain signals of the forces considering inhomogeneous wear can be easily simulated. Then, a total of 11 groups of single-factor simulation experiments are carried out to provide data support. Accordingly, 48 features for each group are extracted, including time-domain and frequency-domain features. By analyzing the Mean Absolute Percentage Error (MAPE) of the extracted features, it is found that the inhomogeneous wear has significant effect on the feature of skewness. Finally, the conclusion is verified by practical experiments through comparing the extracted features in homogeneous and inhomogeneous wear state. The study will provide theoretical and experimental supplement to the engineering application and improve the online wear monitoring accuracy of end mill.