Cavitation is a common fault phenomenon in hydraulic machinery, which can affect the performance of equipment, causing serious economic losses and safety accidents. Many methods for cavitation fault diagnosis have been proposed and validated to be effective, but due to the limited amount of cavitation data and the insufficient feature extraction ability of the models used, there are still serious issues with the accuracy of fault state judgment for these methods. To improve the effectiveness of diagnose cavitation faults, this paper proposes a method entitled SGST which consists of Short-time Fourier Transform (STFT), Generative Adversarial Network (GAN) and Swin Transformer. In this method, first, the vibration signals of hydraulic mechanical equipment are pre-emphasized to compensate for attenuation of high-frequency components. Then STFT is used to transform the signals into spectrograms to capture abundant features based on time-frequency domain. Next, GAN is adopted to perform data augmentation on the spectrograms, increasing the scale and diversity of the dataset. Ultimately, the augmented spectrograms are input into the Swin Transformer network for classification training. The experimental results show that this method achieves an accuracy of 98.6% on the validation set, which is better than the existing cavitation fault diagnosis methods, proving its feasibility and effectiveness.

Based on the Prisoner's Dilemma game, we study the effect of the diverse strategy-updating time scale on the evolution of cooperation under the normalized payoff framework. Agents can adjust their strategy-updating speed according to their fitness and collective influence, and this mechanism promotes the emergence of cooperation on Barabási-Albert scale- free networks. Moreover, agents with higher values of collective influence may have longer persistence-cooperation duration and diffuse their cooperative behaviors among neighbors efficiently. Through investigating the game-learning skeleton, we find that the heavy-tailed in-degree distribution emerges and influencers with proper depth length play an important role in maintaining cooperation.

In this paper, we study the evolution of cooperation in a stochastic game model where agents evolve in terms of some famous and important memory-two strategies, such as AllC, AllD, TFT and W SLS strategies. Comparing with memory-one strategies, it is found that the cooperation is easier to emerge in stochastic games with memory-two strategies. When agents remember more rounds of outcomes, cooperation can boost in noisy environment when some efficient transition rules work to reward mutual cooperation and punish defection. This may provide some potential clues to seek for efficient methods that maintain cooperation in multi-agent games.

Based on the celebrated public goods game with group interaction, we study the evolution of cooperation on two kinds of scale-free networks with similar degree distribution and clustering coefficient. It is showed that there are different evolution routes in the structured population. The metric clusters existing on the popularity-similarity network let cooperation diffuse in local regions. Whereas, cooperators on the clustered scale-free network tend to invade hubs firstly, and then spread from hubs to leaves with a top-down pattern, which leads to the higher cooperation level on the clustered scale-free network than that on the popularity-similarity network.