Equipment health risk assessment without prior knowledge and complete observation information is still a hard problem in the prognostic and health management(PHM) field. To solve the problems caused by unlabeled, unbalanced data under uncertain initial condition upon the health states of engineering equipment, a multivariate deep forest based health assessment method is proposed. Firstly, a correlation and trend metric based feature selection strategy is proposed to remove redundant features. Then, the three-dimensional data is unfolded to be a two-dimensional data, and a quantum fuzzy clustering is utilized to define the health states, which can overcome the problem of initial condition uncertainty. Finally, a multivariate deep forest classifier is adopted for offline training and online health assessment. Case study results show that the proposed health assessment method is effective and feasible.