Modelling and simulation of human motions are of great interest for a number of industrial applications such as ergonomics and production planning. Over time, efforts have been made to provide computerized human models with biomechanically accurate underlying skeleton and realistically rendered volumetric representation of muscles, skin, clothes, etc. known as computer manikins. Despite considerable progress in computational methods for human motion generation in these models, there are still major challenges to generate natural looking motions for daily routine tasks for example in manual assembly lines. This research work proposes a software platform, related methods, and data structures to support use of real motion data for simulating and analysing routine work tasks. One contribution of this thesis is to find ways to deploy motion data in a uniform and efficient manner. The outcomes are a data schema, standards for data conversions, and procedures to aggregate motion data in a unified database. Another part of this thesis is dedicated to 'generating new motions by re-using stored data'. This is done by a synthesizer platform consisting of modules which are able to decompose tasks into primary motions, to search and retrieve motion pieces from the motion database, and to compose a new motion based on the required specifications. In addition, this thesis, presents a method to analyse the generated motions using time-varied motion data. Results of this work contribute in extending current tool functionalities by simulating complicate routine motions which, if not impossible, are very hard to simulate using today's computational algorithms. This work also improves the use of time-varied direct measurement analysis tools against traditional static observational methods.