Research And Experiment Of Drilling State Identification Method For Hydraulic Rock Drill

Drill-blasting method currently accounts for 70% to 75% of the ore-rock excavation engineering and remains dominant in quite a long period of time. Hydraulic rock drill as a representative of drill-tunneling equipment, and the drilling efficiency obtain growing attention. But complicated geological structure of the rock layer, and switching hydraulic rock drill gear dependent on engineering experience to achieve impact energy changs, which lead to poor coupling performance between output parameters of hydraulic rock drill with working medium and seriously reduce the drilling efficiency. Therefore, it is necessary for hydraulic rock drill to identify rock properties during drilling process and the output of hydraulic rock drill operating parameters match with the drilling state, which effectively improves the drilling efficiency and promotes hydraulic rock drill to automation and intelligent development.In drilling process, hydraulic rock drill and rock has force and counterforce at the same time. The hydraulic rock drill working parameters to characterize the drill state. This paper studied working state parameters of hydraulic rock drill and obtained the key characteristic parameters associated with drilling state. According to wave dynamics theory of impact machine, establishing the mathematical model of piston rebound in impact system, and applying co-simulation and analysis based on AMESim and MATLAB. Meanwhile, completing the dynamics modeling and analysis of rotary system, thrust system and buffer device. Finally, confirming six drilling identification key parameters. These parameters are working flow rate of impact system, terminal impact velocity and rebound velocity, working pressure of hydraulic motor, working pressure of thrust hydraulic cylinder and peak oil pressure of buffer chamber.In order to provide data source for drilling state identification, this paper proposed the overall scheme that measuring drilling state identification key parameters of hydraulic rock drill, designed test experimental system based on Lab VIEW, and built experimental platform of drilling state identification system of hydraulic rock drill. In addition, taking into account the simplicity and feasibility of measurement, introduced three-point method, which could obtained indirectly the velocity of impact piston by measuring nitrogen chamber pressure. Under different working flow rate of impact system, acquiring and studying the data of drilling state identification key parameters in drilling three different rocks, research shows the variation of the parameters is consistent with the theoretical analysis.For the status of traditional precise mathematical model is difficult to identify the drilling state, this paper introduced artificial intelligence recognition technology, proposed BP neural network optimized by GA-LM algorithm and one versus one multi-class SVMs optimized by PSO algorithm to realize the identification of drilling state. Considered drilling state identification key parameters as input, and carried out simulation study applying above algorithm. The results shows that comparing with standard BP neural network, momentum BP neural network, variable learning rate BP neural network, LM-BP neural network, BP neural network optimized by GA-LM algorithm has higher drilling state identification accuracy; comparing with non-optimized algorithm, cross validation optimization algorithm, one versus one multi-class SVMs optimized by PSO algorithm has higher identification accuracy of drilling state. In addition, in order to avoid one versus one multi-class SVMs appears inseparable area, this paper proposed sub-divide decision method based on the combination of secondary classification and Euclidian distance, which effectively improve classification performance.This paper proposed a modified method of conflict evidences combination, which introduced credibility and credibility threshold, credibility as evidence weight to weight average and replacing the evidences that are less than credibility threshold, then using the evidences to combine based on D-S evidence theory, the results shows that it effectively improves combination problem of conflict evidence. Based on the improved combination of conflict evidences, this paper proposed a new data fusion method based on the combination of BP neural network(multi-class SVMs) and D-S evidence theory, compared with the simulation and analysis of BP neural network optimized by GA-LM algorithm and one versus one multi-class SVMs optimized by PSO algorithm, the results shows that adopted the data fusion method based on the combination of BP neural network(multi-class SVMs) and D-S evidence theory to identify drilling state. This method has better fault-tolerant and higher identification accuracy, and proves more practical and more effective.