Font Size: a A A

On Description Of Combustion Process Based On Vibration Acceleration Signal Measured From Cylinder Head

Posted on:2009-12-08Degree:DoctorType:Dissertation
Country:ChinaCandidate:S B JiFull Text:PDF
GTID:1102360245994910Subject:Thermal Engineering
Abstract/Summary:PDF Full Text Request
Combustion of internal combustion engine (ICE) is reflection of energy conversion process, which quality determines the power, economy efficiency and emission behavior of ICE. It is widely used in laboratory investigation and some special engine that quality of combustion estimated by pressure analysis, because pressure trace is direct reflection of combustion. But it is inconvenient to diagnose or monitor combustion due to installation difficulties of cylinder pressure sensor. It is well known that surface vibration of cylinder head tie up with combustion and vibration pickup is economy and easy mounting, so studying combustion using Vibration signal has significant theoretical and practical worth.From the research of domestic and abroad, it can be found that the relationship between combustion excitation and vibration of cylinder head is ambiguous, but which is the foundation of combustion estimated using vibration acceleration signal of cylinder head. In view of this, the relationship is studied using finite element analysis, blind source separation, signal analysis and neural network in this thesis. The main research efforts and conclusions are as follows.1. Study on vibration acceleration signal acquisition and processing. 40 kHz, which is sampling frequency for signals, can satisfy sampling requirement for all the response signals on cylinder head. Vibration sensor is installed on cylinder head by glue connection and optimal location is confirmed by calculation and experiment. The deviation result from zero deviation is slaked by ten order polynomial based on least-squares method. Three different type of Filters are compared and the optical filter is FIR filter based on hamming window. It is studied that principle and main characteristics of different time-frequency domain analysis, and these methods are used to analyze vibration acceleration signal of cylinder head. Comparison of different time-domain and frequency-domain analysis shows that S-transform has better performance in analyzing low frequency and continuous wavelet transform has better performance in analyzing high frequency.2. Theoretical analysis for response of combustion excitation. Displacement and acceleration of cylinder head are studied using a model based on finite element analysis. Results indicate that there is affinity between displacement of cylinder head and pressure in cylinder; the two signals have almost same trend before peak pressure. Displacement of cylinder head oscillates after peak pressure and the oscillation frequency is related with the natural frequency of the holistic system including cylinder head and block. It can be found that acceleration of cylinder head and pressure rise acceleration signal have similar trend before peak pressure. Displacement and acceleration of cylinder head in different combustion behaviors are compared, it suggests that amplitude of displacement and fluctuation of acceleration are in correspondence with combustion process in same speed, while there is no correlation between fluctuation of acceleration and combustion process in different speeds. The effect of pretightening forces and material stiffness is studied, and it can be found that displacement and fluctuation of vibration acceleration will increase result form stiffness reduction of the system when the pretightening force of attachment bolt or material stiffness decrease. Results also indicate oscillation frequency will reduce along with decrease of stiffness.3. Partition main phase of combustion based on time-domain waveform of vibration acceleration signal. Theoretical analysis and experimental results indicate that the position of the first peak appear in waveform of acceleration tie up with the inflection point in curve of pressure increase rate which point is usually considered as start of combustion timing. A new method used to estimate combustion timing based on vibration acceleration signal is proposed. It also can be found that the position of trough appear in curve of vibration acceleration signal is adjacent with the time of peak rate of pressure rise, and the second wave crest of vibration acceleration signal is adjacent with position of peak pressure. Based on these conclusions, a new method used to divide combustion phase is proposed.4. Frequency analysis for response of piston slap. The certainty of piston slap in reciprocating dynamic machinery and the influencing factors of piston slap are analyzed in this thesis. Response signals of piston slap that obtained from the research ZH195 diesel engine driven by motor and the research 495T diesel engine cut off one cylinder fuel supply are analyzed by continuous wavelet transform and results indicate energy distribution of response signals are all around 5kHz and from 10kHz to 20kHz. Blind source separation method coule be used to separate the vibration acceleration on cylinder head and FastICA algorithm based on Independent component analysis is used to separate cylinder head vibration acceleration signal. Response of piston slap separated by FastICA are analyzed by continuous wavelet transform and results show the response contains high frequency components even in normal working condition, also it can be found energy of low frequency components increase.5. Relationship between vibration acceleration signal and combustion process. Time frequency domain analysis is used to obtain the frequency information of short interval vibration to enhance frequency resolution and the new method is verified by contrast with FFT. This method is used to analyze pressure rise acceleration signal. Results show that the main energy distributes under 2000Hz and the energy has the same trend with combustion process. The method is also used to analyze vibration acceleration signal and results show that the frequency information is affected by both combustion process and system performance.6. Combustion estimation based on vibration acceleration signal in the interval from start of combustion to peak pressure. The parameters of vibration acceleration signal such as fluctuation from the first peak to the first valley and frequency energy are compared with combustion process. Results show that there is irrespective between the parameters of vibration acceleration signal and combustion process in different speed, while there is linear relationship between the parameters and the maximum pressure increase rate in the same speed. A new method used to evaluate combustion difference of multi-cylinder engine is proposed based on the parameters. The method is verified in the research 495T diesel engine and results show the method is feasible.7. Study on reconstruction of pressure in cylinder based on neural network. Different training functions are compared and conjugate gradient method is considered to be the optimal function for the application. Reconstruction of pressure in cylinder is firstly studied using the entire vibration acceleration signal and results show that the reconstructed pressure may fluctuate obviously sometimes due to vibration acceleration signal after the peak pressure position contains less useful information related with combustion process. In order to improve reconstruction precision, vibration data from start of combustion to peak pressure are used to reconstruct pressure curve in cylinder. The trained neural network is examined by four untrained group data from different work conditions, and results suggest the method has better performance. Key point of reconstructed pressure and target pressure are almost coincident and the minimum error of amplitude is 1.4% and the maximum error of amplitude is 7.9%, according to the results, author considers that the method can achieve better application effect if there have more training datas.
Keywords/Search Tags:Vibration acceleration signal, Estimation of combustion process, Finite element analysis, Blind source separation, Time frequency analysis, Neural network
PDF Full Text Request
Related items