Guizhou Province, known as the "Asian Phosphorus Capital," harbors the third-largest phosphate rock reserve in China. This mountainous region faces significant engineering challenges during mining, particularly with inclined thick ore bodies. The need to prevent ecological damage and safeguard lives during phosphate mining operations has led researchers to develop intelligent systems for predicting surface movement and deformation.
This study focuses on the subsequent filling mining of inclined thick ore bodies, employing advanced methodologies to anticipate geological shifts. Using the probability integral method, the researchers developed a comprehensive system within MATLAB App Designer aimed at calculating surface movements resulting from mining activities. The initial findings have industrial applications, indicating promising uses for artificial neural networks (ANNs) to accurately predict surface changes.
Industrial validations revealed astonishing results—predictions indicated maximum surface settlement values of -38.37 mm and -39.73 mm for the strike and inclined main sections, respectively, with the system achieving prediction accuracy of 90.56%. The authors suggest implementing correction coefficients of 1.26 and 1.40 to fine-tune the predictions and align them closer to real-world measurements, which stood at -48.03 mm and -55.49 mm for the same sections.
Researchers have highlighted the significant geological variances presented by Guizhou's complex mining environments. Previous studies have examined movements through various techniques, including extensive field monitoring and simulations. The probability integral method, foundational to this research, integrates dynamic theoretical frameworks and superior mathematical rigor, making it fitting for both horizontal and inclined seams.
The Intelligent Prediction System (IPS) comprises three functional modules: the initial parameter setting, expected parameter calculation, and visualization of movement predictions. The modular design is aimed at allowing precise adjustments for real-time data application to the specific characteristics of the mining areas being studied.
Within the initial parameter setting module, users must enter specific mining conditions, including geological parameters, mining depth, and ore body attributes. This customization is pivotal for the accurate analysis of the parameters influencing surface movement.
Once initial parameters are defined, the IPS allows access to the expected parameter calculation module, which utilizes existing field data to determine movement predictions through the integration of the ANN. This neural network is pivotal for analyzing extensive datasets, predicting subsidence patterns, and informing operations before physical changes occur.
The visualization module enables operators to observe the predicted outcomes of surface movements, providing real-time three-dimensional images and characteristic curves for the mining sections. This visibility is invaluable for engineering teams on the ground, facilitating informed decision-making and enhancing safety measures considerably.
Field trials conducted at the Daxin Beidoushan Mine resulted in substantial findings. Validation against actual measurements indicated an excellent predictive capability with ROC curve analyses yielding AUC values of 0.9259 for strike sections and 0.8852 for inclined sections. Such figures strongly affirm the system’s reliability, enhancing confidence among mining engineers as they tackle the inherent risks beset by mining operations.
Despite the robustness of this intelligent system, limitations persist, particularly related to environmental conditions influencing terrain during operations—effectively, changes to geological landscapes during the mining process can lead to varying prediction accuracies based on the parameters set. The study founders encourage research teams to develop improved methods for obtaining training data and sample accuracy, fostering broader applications of the IPS across diverse mining conditions.
Overall, the study illuminates the continuing advancements within mining technology, emphasizing the importance of developing predictive methods amid increasingly complex geological challenges and the pressing necessity for ecological preservation. By incorporating modern technologies such as ANNs alongside traditional mathematical models, the researchers aim to usher in safer mining practices, mitigate risks to properties, and uphold ecological integrity across mining operations.