Monitoring real-time concurrent systems is a challenging task. In this paper we formulate (model-based) supervision by means of hidden state history reconstruction, from event (e.g. alarm) observations. We follow a so-called true concurrency approach using time Petri nets: the model defines explicitly the causality and concurrency relations between the observable events, produced by the system under supervision on different points of observation, and constrained by time aspects. The problem is to compute on-the-fly the different partial order histories, which are the possible explanations of the observable events. We do not impose that time is observable: the aim of supervision is to infer the partial ordering of the events and their possible firing dates. This is achieved by considering a model of the system under supervision, given as a time Petri net, and the on-the-fly construction of an unfolding, guided by the observations. Using a symbolic representation, this paper presents a new definition of the unfolding of time Petri nets with dense time.