Abstract: To address the industrial challenges of relying on manual inspection and cleaning of filter cloth fouling during filter press cake discharge in coal preparation plants—such as low operational efficiency, high labor intensity, and potential safety hazards—a systematic study was conducted to achieve fully automated and intelligent filter press operations. Taking the filter-pressed coal slime from Nanliang Coal Preparation Plant as the research object, this study focused on four aspects: fouling adhesion mechanism, intelligent identification methods, cleaning device development, and coordinated system control. First, Fourier-transform infrared spectroscopy (FTIR) and X-ray diffraction (XRD) were employed to characterize the fundamental properties and mineral composition of the coal samples. Single-factor experiments using a self-developed detachment-force testing apparatus were conducted to investigate the effects of ash content, moisture content, particle size, and mineral composition on the adhesion force between coal slime and filter cloth, and the weighted contributions of these factors were quantitatively determined. Subsequently, a rail-mounted inspection system synchronized with the filter plate shifting trolley was designed. An image-segmentation model integrating a convolutional neural network (CNN) with a recurrent neural network (RNN) was developed for filter cloth fouling identification. A supervisory computer monitoring platform featuring real-time display, graded warning, and data traceability was also developed. Furthermore, considering the distinct characteristics of two typical fouling types—namely, blocky and pasty fouling—a composite cleaning device combining high-frequency vibration excitation with flexible scraping and brushing was developed. The operating parameters, including vibration frequency, brush thickness, and speed, were optimized experimentally. Finally, a closed-loop identification and cleaning system was established based on a programmable logic controller (PLC) to enable coordinated operation between the identification and cleaning processes. The results indicate that mineral composition and moisture content are the dominant factors affecting the adhesion force between coal slime and filter cloth, with weighted contributions of 58.48% and 31.30%, respectively. The proposed intelligent identification method achieves a pixel-level identification accuracy of over 99% for fouling and can effectively adapt to complex lighting and dusty conditions in filter press workshops. Under the recommended conditions of a vibration frequency of 60 Hz, brush thickness of 10 mm, and speed of 30 mm/s, the composite cleaning device can efficiently and nondestructively remove both blocky and pasty fouling. The cleaning rate for fully formed filter cakes reaches 100%. Additionally, the device can avoid the raised structures of filter plates and prevent damage to the filter cloth. The integrated system, consisting of synchronized positioning, intelligent identification, dynamic monitoring, and real-time cleaning, can effectively replace manual inspection and cleaning of filter cloth fouling, thereby reducing labor intensity and mitigating safety risks. The findings provide technical support and a practical engineering reference for the unmanned and intelligent upgrading of filter press processes in coal preparation plants.