Research On Decoking Status Detection Technology Based On Vibration Acoustics

Delayed coking is a thermal cracking process whose main purpose is to convert high residual carbon residual oil into light oil.The coke generated after the coking is attached to the tower wall,and the decoking operation is required before entering the next coke production.Hydraulic decoking is a commonly used decoking method.It mainly uses high pressure water to continuously impact the coke in the coke drum to fall off and flow out from the bottom chute.In the current decoking operation,the degree of coke removal depends entirely on the man-made observations of the operator.The labor intensity is high,the working environment is bad,and the subjective error of judgment is large.This thesis aims to establish a model and determine the cleanliness of decoking through the vibration signal acquisition and analysis of the coke drum under high pressure water impact.Through the analysis of the coke tower object,the kinetic model of the coke tower under hydraulic impact was established.Based on the model,simulation analysis was carried out to find the dependence between the thickness of the attached coke layer and the vibration signal of the coke drum.The thicker the coke layer is attached to the coke layer,the smaller the amplitude in the mode parameters and the lower the frequency.By extracting the characteristic parameters,pattern recognition can be used to establish the relationship between the decoupling state and the mode parameters.In this paper,the BP neural network is used to perform Fourier transform on the acquired vibration signal samples to obtain the amplitude-frequency curve of the vibration signal.The amplitude of different characteristic frequency bands is extracted as the characteristic parameter,and the sample learning training is carried out to establish a learning recognition network between the decoking state and the vibration characteristics of the coke tower.A trained and stable network can be used for the detection of decoking conditions.The hydraulic de-focus intelligent detection system consists of a vibration signal acquisition system,a signal pre-processing and analysis system,a neural network identification system and an output system.The main work flow: several vibration sensors are installed on the outer wall of the coke drum for hydraulic decoking,and the vibration signal is used to collect the vibration signal of the wall of the hydraulic decoking coke tower,but the signal collected at this time is the time domain signal,so it is required Fourier transform transforms the time domain signal into frequency domain signal,and then extracts the amplitude of different characteristic frequency bands as the characteristic parameters,carries out sample learning training,takes the extracted frequency feature value as input,and uses BP neural network to establish the characteristic value and The relationship between the degree of cleanliness of the coke drum decoking state.The experimental results show that the system can correctly identify the cleanliness of decoking and achieve the expected research goals.