By analyzing the current development process of human-computer collaboration (HRC) and combining the realistic development needs of complex collaborative environments, an experimental paradigm for detecting the abnormalities of the HRC system is constructed by taking collaborative assembly of mobile phone panels as a case study. Brain-Computer Interaction (BCI) technique is employed to assess the similarities and differences between normal and abnormal signals, thereby addressing the issue of the inability to predict and model abnormalities in the work of human-computer collaborative assembly (HRCA). The results demonstrate that smaller perceptual differences result in significant similarities and differences between major electroencephalogram (EEG) components, such as N1, N2 and P3. This suggests that the design of the paradigm and experiment is reasonable and feasible. This paper also discusses the next part of the work, which is to generate a mapping model of thinking and intention by extracting multi-dimensional data features, transforming the process of human behaviour - neural response - data features into abnormalities - intention model - system task, which will provide an important idea for the systems and robots to better understand and perceive human beings.