The Development Of Dynamic Loading Identification System For Super-heavy Intelligent Rock Roadheaders

The content of this paper is an important part of the manufacture ofintelligent super-heavy rock roadheaders (No:2012AA06A405), the fifthsubtopic of Intelligent coal mining technology and equipment(one),which ismajor project in national863program, resources and environment technologyareas.It is proposed on the issue of rock roadheaders such as complicatedoperating conditions, changing loads, difficult dynamic load real-timeidentification.As a part of automatic control system of the rock roadheader, theidentification of dynamic load is of great significance to improve the intelligentlevel and increase lifetime of roadheaders. Recently, rock roadheaders have beenmore widely used in coal mine, but intelligent tunneling technology is still in itsinfancy. Cutting speed of roadheaders is adjusted slowly and is in the stage ofmanual operation. Manual operation not only requires great labour intensity, butalso leads to a serious loss for cutting picks because of the difficulty ofjudging cutting state in time. It is important to adjust cutting speed automaticallyaccording to loading variation and reliable dynamic loading identificationtechnology is necessary. Therefore, it has very important practical significanceto develop rock dynamic load identification system for super-heavy intelligentrock roadheaders.Based on the analysis of the cutting mechanism’s dynamic characteristicsfor super-heavy intelligent rock roadheaders and combined with advanced signalanalysis techniques, intelligent recognition technology and the practicaloperation of rock roadheaders, the dynamic load identification system wasdeveloped through a series of dynamic load simulation experiments. The mainresearch contents are as follows: Refered to a large number of relevant literature, the current situation andtrends of rock roadheaders have been expound domestic and abroad. Afteranalyzing cutting mechanism’s loads distribution of different working status,physics parameters such as cylindrical cantilever vibration, cutting motor current,rotary cylinder and lifting cylinder pressure were defined which could reflectdynamic load availably.According to limited space of cutting mechanism, various sensors whichapplyed to atrocious environment of mine were selected to finish monitoringaccurately. Combined with system requirements, the overall framework ofdynamic loading identification system had ultimately been designed, which wasbased on data acquisition card and industrial computer.Because of the signals containing a lot of interference and comparing theadvantages of Fourier transform, wavelet transform and wavelet packettransform, feature extraction method was determined. The steps of featureextraction were introduced in detail and the range of characteristic frequency ofvibration signals, the current and hydraulic cylinder pressure signals wasconfirmed through analyzing a large amount experimental data. Combined withfeature extraction method, RBF neural network and evidence theory wereselected as intelligent identification methods of dynamic load. Tructuralalgorithm of RBF neural network and the steps of evidence theory wereintroduced simply, but recognition principle of combined neural network,dynamic load recognition principle of D-S evidence and fusion principle ofcombined RBF neural network with D-S evidence theory were the emphasis.Dynamic load recognition software programs were developed under theenvironment of LabVIEW, including MATLAB Script node invoked byLabVIEW, multiple parameter synchronization acquisition program, featureextraction of wavelet packet, dynamic loading identification programs of neuralnetwork, fusion programs of evidence theory, database management programs,as well as man-machine interface. Debugging results showed validity ofsoftwareprograms.According to working status of rock roadheaders, the scheme of dynamic load recognition was designed after dividing operating mode with pressuresignals of rotation cylinders and lift cylinders. Based on three categories ofunderholing, horizontal cutting, longitudinal cutting, operating diagnosis networkwas made up of each operation diagnosis network and one network withoutconsidering operation was replaced by three networks. The model of one-leveland two-level RBF neural network fusion recognition was builted and arecognition method using multi-neural network and evidence theory for rockroadheaders was presented, combining vibration data, the current and hydrauliccylinder data, integrating two data fusion methods by using their superiority andavoiding their disadvantages. Finally date was used to train, test and analyse.