Abstract: Dredging project plays a critical role in the construction of port channels and marine exploitation, with cutter suction dredgers (CSDs) serving as the primary equipment for transporting dredged slurry to designated discharge sites through pipelines. Real-time monitoring of key parameters, such as flow velocity, is vital for optimizing the efficiency of sediment conveyance by CSDs. However, the challenging operational environment imposes significant constraints on traditional physical sensors, which are costly and require complex maintenance, thus limiting their widespread application. To address this issue, this paper introduces a soft sensing method ,ICBFormer, targeting flow velocity as measured variable. ICBFormer integrates Interactive Convolutional Block (ICB) and Transformer to replace traditional flowmeters. The ICBFormer utilizes the ICB block to capture complex inter-variable relationships and extracts multi-scale temporal features. Then, by leveraging the advantage of Transformer in long-sequence feature extraction, ICBFormer effectively handle the dynamic relationship within variable data sequences , achieving accurate pipeline flow velocity prediction. Validation experiments, conducted through a simulated dredge pump transport platform, demonstrate that the ICBFormer offers significant advantages in flow velocity prediction. This method provides a novel solution for reducing sensor costs and maintenance expenses associated with dredging projects.