Time series scientific simulation on supercomputers generates huge amounts of data at each time step. In the big data era, these data become impossible to be stored anymore, so simultaneous analysis of these data is strongly demanded. In order to realize such an on-the-fly intuitive analysis of simulation, this paper showcases a multimodal visualization and steering system with visual and haptic interfaces for a real-time GPU-based coupled fluid-structure simulation. Since the nature of touching sense of human beings requires extremely fast and continuous refreshing to form authentic feeling, parallel techniques were utilized to speed up haptic updating to around one millisecond. A middle-layer interface for a haptic device was developed to realize the ease of use of this device. Furthermore, a model of palpation was proposed to allow users to touch, push and sense the dynamic fluid motion inside a deformable tube.