The collection-disassembly problem plays an important role in a reverse supply chain. It coordinates the collection and disassembly activities for end-of-life (EOL) products. Most existing works consider the deterministic problems. However, in practice, demands of reusable components in EOL products may be uncertain. Besides, it is usually difficult to exactly obtain probability distributions of uncertain demands, due to inadequate historical data. This paper studies a collection-disassembly problem under partial known distributional information of component demands, in which equipments of the disassembly site, corresponding to different disassembly capacities, have to be selected. The objectives are to minimise the system cost and to maximise the customer service level. For the problem, a novel distributionally robust bi-objective formulation is proposed. Based on the Monte Carlo simulation and an ambiguity set, a sample average approximation (SAA) model and an approximated mixed integer programming (MIP) model are constructed, respectively. Then the two approximated formulations are solved, via the ε-constraint framework, and compared in numerical experiments.