Research On Loose State Detection And Position Recognition Method Of Bolt Connection Group

Bolts are the most commonly used standardized fasteners in mechanical systems.A bolt connection group is a combination of multiple bolts on the same cross-section.The combination of the cross-section geometry and layout will make the connection structure have better connection strength,Carrying capacity and service life.However,the bolted connection group often suffers from many severe working conditions(such as cyclical load,mechanical shock,chemical corrosion,and improper operation of personnel,etc.).These unfavorable conditions may lead to the loosening and failure of the bolted connection group,and even the equipment Shutdown will cause the machine to crash or kill people.Therefore,it is urgent to carry out research on the detection of the loosening state of the bolted connection group and the position identification.Aiming at the difficulty of extracting the characteristics of the vibration and shock response signal of the bolted connection group,this paper uses the sparse representation theory to mine the potential dynamic feature changes in the signal.Firstly,the sparse representation and reconstruction of the noise-containing vibration response signal through the third-order simulation verify the feasibility of the sparse representation theory in the application of this type of signal.A bolted rotor experimental platform was built,and the loosening state of bolted connection groups in six different states was collected,and a method for detecting the loosening state of bolted connection groups based on sparse representation reconstruction error was proposed,which was constructed by the K-SVD dictionary learning algorithm Redundant dictionaries for different loosening states,based on the principle of sparse representation of the smallest single-class reconstruction error,realize high-accuracy detection of the loosening state of bolted connection groups.Since mechanical vibration signals are usually mixed with interference caused by objective factors such as system noise and load changes,the accuracy of state detection is reduced.In this paper,a method for detecting the loosening state of bolted connection groups based on the combination of sparse representation and random forest algorithm in a strong noise environment is proposed.The sparse representation coefficient is used as the feature input of the random forest algorithm.The results show that the proposed method has good recognition accuracy in the detection of the loose state of bolted connection groups,and the comparison proves that the method has strong anti-interference ability.The vibration signal is the overall response signal of the system,and it is not sensitive to local structural changes.In the experiment,the location of the acceleration sensor is reasonably arranged so that it can also detect the response changes caused by the local looseness of the structure.Aiming at the difficulty in identifying the loose position of bolted connection groups,a method for identifying the loose position of bolted connection groups based on sparse wavelet packet energy characteristics is proposed.The wavelet packet decomposition of the local looseness and individual looseness state signals of the bolted connection group after sparse reconstruction is carried out by the db10 wavelet basis function,and the sparse energy characteristics and energy information entropy of each sub-band are calculated.The results show that the energy characteristics of the sub-bands and the energy information entropy can reflect the current distribution of signal energy,so that the loose position of the bolted connection group can be preliminarily inferred.