The rise of Artificial Intelligence of Things (AIoT) makes everything connected smartly, which can enrich industrial productivity. With the help of learning-based wireless communications, terminals in AIoT can communicate with each other and collaborate intelligently. However, one of the critical issues faced in AIoT deployment is the lack of available large datasets for the training of artificial intelligence algorithms in the industrial environment. In this paper, we propose a channel data augmentation algorithm for the dataset limitation cases in intelligent industrial wireless communication systems using generative adversarial networks (GAN). Specifically, we first show the performance of deep learning-based channel state information (CSI) feedback algorithm trained with different size of datasets. Then we develop a GAN for the channel data augmentation to enhance the performance of CSI feedback algorithm. Finally, we apply the enhanced approach to an insufficient dataset for the performance evaluation. Experimental results show that our method can achieve at most 3dB performance improvement than other traditional data augmentation approaches in increasing the accuracy of CSI feedback algorithm when the size of dataset is limited to 10000.