| Vibrating screen is an important material screening processing machinery.It is used in industrial production for the classification and dehydration of raw materials.Because of its indispensable role,it plays an important role in the chemical industry such as coal and stone mines.In recent years,the demand for vibrating screens has increased year by year,and the market prospect is excellent.However,in actual industrial production,because of continuous working for a long time and the harsh working environment,the vibrating screen is prone to a series of failures,which causes the production line to stop production and brings direct economic losses to the enterprise.Therefore,it is extremely necessary to deeply study the fault diagnosis of the vibrating screen and develop a set of vibrating screen health monitoring methods,which has important academic value and practical engineering significance.This article takes the vibrating screen as the research object,analyzed the causes of the vibrating screen failure and the common fault types,and designed a failure experiment model to build a vibrating screen failure experimental platform,and completed the selection of hardware and software platform;on this basis,three types of failure experiments were designed,namely,imbalance of excitation force,change in spring stiffness,and change in spring height.For the original acceleration signal data collected from the fault experimental platform,the available acceleration signal was obtained after the pre-processing steps,and then one and two integrations,and after removing the trend term,the available speed signal and displacement signal were obtained respectively.The X and Y dimensional displacement images were synthesized using Lissajous images to obtain the two-dimensional vibration trajectory information,realizing real-time processing of the vibrating screen fault experimental data.The design uses a combination of variational mode decomposition and multi-scale permutation entropy to extract the fault feature of the vibrating screen.The vibrational acceleration signal was subjected to variational mode decomposition,and the frequency spectrum estimation method was used to select the sensitive component containing the fault information.Calculating the permutation entropy value of the acceleration signal at different scales,selected the permutation entropy value corresponding to the fault-sensitive scale factors to form the fault feature vector group,and completed the fault feature extraction.Aiming at the feature vector group after feature extraction,the variable predictive model based class discriminate,BP neural network,support vector machine,and variable predictive model based RBF network model were used to study the recognition algorithm of vibrating screen faults,and to test and analyze the recognition rates of different faults recognition algorithm.The comparison found that the variable predictive model based class discriminate classification recognition rate was 93.50%,the BP neural network recognition rate was 87.00%,the support vector machine diagnosis rate was 91.50%,and the variable predictive model based RBF network model classification recognition rate was 96.00%.It is verified that the fourth recognition algorithm has better applicability and higher classification effect.After the system development is completed,the vibrating screen fault monitoring system is tested and analyzed on the industrial site to obtain measurement data under different fault conditions.Testing the experiment by changing the magnitude of the exciting force,the spring stiffness and the spring height.The results show that the test system can identify these three types of faults,with an average recognition rate of 94.27%,which needs to be improved in the subsequent work.The method for monitoring the health status of the vibrating screen has high accuracy and practicability,and can be used for fault diagnosis of the vibrating screen.It provides a new idea for the monitoring of the health status of the vibrating screen and has certain reference significance. |
