The research on the recovery of low-grade thermal energy carried away by boiler flue gas is significant for sugar mills. This paper designs a waste heat recovery system based on sugar plant flue gas, integrating absorption refrigeration cycle and the organic Rankine cycle, and the effects of nine working fluids on the system are investigated. The aim is to realize the multi-form conversion of energy. The performance of the system is evaluated in terms of energy, exergy, and economic metrics. Multi-objective optimization is performed with the method of the NSGA-II genetic algorithm. The results show that Butane is the most suitable working fluid for ORC. The exergy efficiency of the system is 32.125% before optimisation, with an increased space cooling capacity of 15820.56 MW per year for the sugar mill. The exergy destruction analysis of the system reveals that the generator accounts for the highest proportion of exergy destruction (50.8%). The entire system shows the LCOE is as low as 0.0406$/kWh under the optimized condition. The optimized system can obtain an estimated annual electricity sales revenue of $136,300, and the sugar mill can save $308,600 in cooling costs. In addition, the payback period can be shortened to 5.79 years.

Surgical robots are increasingly utilized in medicine for their reliability and convenience. An accurate kinematic model is essential for precise robot control and enhanced surgical safety. However, the high nonlinearity and computational complexity of kinematics pose significant challenges to traditional numerical methods. This study designs a surgical robotic arm and establishes the motion mapping relationship between the joint space and the end-effector workspace. Subsequently, a hybrid kinematic estimation model based on deep pyramid convolutional neural network (DPCNN) is proposed, which integrates data sampling and an attention mechanism to improve computational accuracy. The Latin hypercube sampling technique is used to improve the uniformity of dataset sampling, and the triplet-fusion self-attention mechanism (TFSAM) is employed for multi-scale feature information. Experimental results show that the TFSAM-DPCNN model achieves coefficient of determination (R²) values exceeding 0.99 across all testing scenarios. Compared with other models, the proposed model reduced the root mean square error (RMSE) by up to 81.34%, exhibiting superior performance. Furthermore, the developed 3D simulation platform validates the effectiveness of the proposed model. This study offers a robust solution for multi-degree-of-freedom robot modeling, with potential applications across a range of robotic motion control systems.