The drilling robot is the key equipment for pressure relief in rockburst mines, and the accurate recognition of a pressure relief hole is the premise for optimizing the layout of pressure relief holes and intelligent drilling. In view of this, a pressure relief hole recognition method for a drilling robot, based on single-image generative adversarial network (SinGAN) and improved faster region convolution neural network (Faster R-CNN), is proposed. Aiming at the problem of insufficient sample generation diversity and poor performance of the traditional SinGAN model, some improvement measures including image size adjustment, multi-stage training, and dynamically changing iteration times are designed as an improved SinGAN for the generation of pressure relief hole images. In addition, to solve the problem that the traditional depth neural network is not ideal for small-size target recognition, an improved Faster R-CNN based on multi-scale image input and multi-layer feature fusion is designed with the improved SqueezeNet as the framework, and the sample data collected from ground experiments are used for comparative analysis. The results indicate that the improved SinGAN model can improve the diversity of generated images on the premise of ensuring the quality of image samples, and can greatly improve the training speed of the model. The accuracy and recall rate of the improved Faster R-CNN model were able to reach 90.09% and 98.32%, respectively, and the average detection time was 0.19 s, which verifies the superiority of the improved Faster R-CNN model. To further verify the practicability of the proposed method, some field images were collected from the underground rockburst relief area in the coal mine, and a corresponding test analysis was carried out. Compared with three YOLO models, the accuracy and recall rate of improved Faster R-CNN model improved significantly, although the training time and recognition time increased to a certain extent, which proves the feasibility and effectiveness of the proposed method.

The anti-punch drilling robot is a core piece of equipment used to realize unmanned drilling and pressure relief operations in underground coal mines. Adaptive drilling, conducted according to the coal rock properties encountered during drilling, is essential to improve the safety and efficiency of the pressure relief working face. This paper analyzed the composition of the anti-punch drilling robot drilling system and workflow of the drilling system and then calculated the optimal rotary speed and the optimal feed speed for different Platts hardness coefficients of the coal rock through the analysis of the drilling rod force. Based on the characteristics of the drilling electrohydraulic control system, a rotary adaptive controller based on a self-resistant control algorithm and a feed adaptive controller based on sliding mode variable structure control were designed. A joint simulation was carried out using AMESim 2020.1 and Simulink 2020b software to analyze the control performance of each controller. Finally, an experimental platform for the drilling robot electrohydraulic control system was constructed; different hardness coefficients of concrete specimens were used to simulate the hardness of the coal rock with different traits. Single coal rock hardness experiments and drilling experiments with sudden changes in coal rock hardness were carried out. The experimental results showed that the adaptive control strategy proposed in this paper satisfies the requirements of drilling system control.

The recognition of shearer cutting state is the key technology to realize the intelligent control of the shearer, which has become a highly difficult subject concerned by the world. This paper takes the sound signal as analytic objects and proposes a novel recognition method based on the combination of variational mode decomposition (VMD), principal component analysis method (PCA), and least square support vector machine (LSSVM). VMD can decompose a signal into various modes by using calculus of variation and effectively avoid the false component and mode mixing problems. On this basis, an improved gravitational search algorithm (IGSA) is designed by using the position update mechanism of Levy flight strategy to find the optimal parameter combination of VMD. Then, the feature extraction is achieved by calculating the envelope entropy and kurtosis of the decomposed intrinsic mode functions (IMFs). To avoid dimensional disasters and reinforce the classification performance, PCA is introduced to choose useful features, and the LSSVM-based classifier is reasonably constructed. Finally, the experimental results indicate that the proposed method is more feasible and superior in the recognition of shearer cutting states.

Reliable coal–rock identification during drilling is challenged by limited labeled data and non-stationary, strongly coupled multi-source signals. This laboratory case study evaluates a recognition-oriented adaptive framework, IMGO–VMD–WT, that uses torque-guided improved moss growth optimization (IMGO) to select variational mode decomposition (VMD) parameters, grades intrinsic mode functions by information content, applies differentiated wavelet-threshold processing, and reconstructs the signal with residual preservation. A self-built laboratory drilling rig provided data from six material categories and eleven pressure, displacement, torque, and rotational-speed channels. Under five-fold contiguous-block validation, raw features, wavelet thresholding, fixed VMD–WT, PSO–VMD–WT, GWO–VMD–WT, and IMGO–VMD–WT achieved 90.26%, 91.94%, 92.17%, 92.13%, 92.13%, and 92.57% accuracy, respectively, with corresponding macro-F1 scores of 78.34%, 80.37%, 81.73%, 82.17%, 82.17%, and 82.22%. The exploratory paired comparison between IMGO–VMD–WT and fixed VMD–WT gave accuracy and macro-F1 differences of 0.40 and 0.49 percentage points, with exact sign-permutation p-values of 0.6875 and 0.5625 and bootstrap 95% intervals that included zero. In a controlled 10 dB noise-injection test on 12 measured torque windows, IMGO–VMD–WT reached 19.19 dB output SNR and an RMSE of 1.379. These results support decomposition-aware processing as a feasible offline strategy for limited-data laboratory drilling analysis.