Engine combustion state prediction system and method based on multi-source spectral information fusion

Abstract

The engine combustion state prediction system and method based on multi-source spectral information fusion relates to the field of energy and power device perception technology. The timing control module ensures the synchronization of timing, the flame spontaneous radiation intensity perception module and the multi-source information detection module detect the same flame area of ​​the engine's combustion chamber, and the training set of the state prediction model uses the radiation intensity perception data to divide the image features and spectral information into data sets, including three sets corresponding to the three states of stable combustion, transition state and flameout state of the engine. The state prediction model uses spectral information as input and outputs the predicted combustion state to the engine in real time. Make full use of the image features and spectral information obtained by multi-source information detection, and use the state prediction model to predict the engine combustion state in real time, providing a reference for the engine operation state control, with higher accuracy.

Description

Technical Field

[0001] The present invention relates to the field of energy and power device sensing technology, and specifically to an engine combustion state prediction system and method based on multi-source spectral information fusion. Background Art

[0002] The design focus of the engine is to improve the performance of the engine itself, so that the fuel in the engine combustion chamber burns more fully, so as to improve the combustion efficiency of the fuel while reducing pollutant emissions. Identifying and predicting the combustion state of the engine helps to make real-time adjustments to the engine's operating state, or even make adjustments in advance. It can also detect and solve potential faults in a timely manner, avoid engine losses caused by abnormal flame conditions, and thus improve engine performance and safety. In addition, the combustion process of the engine can be optimized through data analysis and comparison, and combustion efficiency can be improved and energy saving and consumption reduction can be reduced. The transition state between the engine's combustion state from stable combustion to the flameout state is often very short, generally in the order of ms.

[0003] In engineering applications, there are many optical means for engine measurement, such as PLIF technology, spontaneous radiation detection technology, PMT technology, TDLAS and Schlieren. Compared with general electronic detectors that can only measure a single parameter, optical measurement can also obtain multiple parameters at the same time. In addition, optical measurement has the characteristics of non-destructive testing and will not interfere with the engine's operating status. Most importantly, the engine pressure data obtained by traditional pressure sensors has a lag compared to the actual engine status, and the pressure data lags behind the spontaneous radiation intensity data by about 200ms, and cannot respond instantly. In comparison, optical measurement methods have good real-time performance.

[0004] At present, when measuring the actual engine, multiple optical measurement methods are often deployed at the same time, and different optical measurement methods are independent of each other. How to obtain accurate information about the flame combustion state from the large amount of information obtained by optical measurement requires a state prediction model with high accuracy. In addition, if the combustion state of the engine needs to be adjusted in real time, this requires the state prediction model to have the characteristics of instant response. Summary of the invention

[0005] In order to address the shortcomings of the background technology, the present invention provides an engine combustion state prediction system and method based on multi-source spectral information fusion, which makes full use of the image features and spectral information obtained by multi-source information detection, and predicts the engine combustion state in real time through a state prediction model, providing a reference for engine operation state control with higher accuracy.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An engine combustion state prediction system based on multi-source spectral information fusion includes a flame spontaneous radiation intensity perception module, a multi-source information detection module, a timing control module and a state prediction model. The timing control module connects the flame spontaneous radiation intensity perception module and the multi-source information detection module to ensure the timing synchronization of the two in the detection process. The flame spontaneous radiation intensity perception module and the multi-source information detection module detect the same flame area of ​​the engine's combustion chamber to obtain radiation intensity perception data and multi-source information data of the flame area respectively. The state prediction model is trained through a training set. The training set uses the radiation intensity perception data to divide the image features and spectral information into data sets. The image features and spectral information are obtained by extracting the features of the multi-source information data, and the training set should include three sets corresponding to the three states of stable combustion, transition state and flameout state of the engine. The trained state prediction model uses the subsequently given spectral information as input and outputs the predicted combustion state to the engine in real time.

[0008] The engine combustion state prediction method based on multi-source spectral information fusion includes the following steps:

[0009] Step 1: Acquisition of detection data from the optical synchronous detection system

[0010] The same flame area of ​​the engine's combustion chamber is detected using an optical synchronous detection system to obtain radiation intensity perception data and multi-source information data of the flame area. The timing control module is used to ensure the timing synchronization of the two, and the radiation intensity perception data is used to characterize the combustion state of the engine.

[0011] Step 2: Extraction of state features

[0012] Extracting the multi-source information detection module obtains the image features and spectral information in the multi-source information data, including the flame area, perimeter, flame center position and Zernike moment, and obtains the data set;

[0013] Step 3: Information fusion and data set division

[0014] The data obtained by the flame spontaneous radiation intensity sensing module and the multi-source information detection module are fused, and the nodes of different combustion states of the engine are determined using the radiation intensity sensing data. The combustion state of the engine is divided into three states: stable combustion, transition state and flameout state, and the data set obtained in step 2 is divided into the three states to form a training set;

[0015] Step 4: State prediction model training and combustion state prediction

[0016] The state prediction model is trained using the training set obtained in step three. The training set is first normalized by mean variance. The state prediction model is based on partial least squares and artificial neural network machine learning algorithms. It is tested using a pre-prepared prediction set. The model parameters are continuously adjusted to determine whether the state prediction model has the optimal prediction accuracy for the prediction set. The trained state prediction model uses the given spectral information as input to perform real-time prediction of the combustion state and transmit it to the engine as a reference for its control.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention utilizes a multi-source spectral information fusion method to fuse the data obtained by the front-end optical synchronous detection system, which can better reflect the real combustion process and thus has higher accuracy. At the same time, the present invention can achieve real-time state prediction output for a given engine spectral information input, achieving real-time control or even early warning, and has better real-time performance than non-optical detection methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the principle of the engine combustion state prediction system of the present invention;

[0019] Figure 2 It is a comparison chart of the flame spontaneous radiation intensity perception data curve and the pressure curve;

[0020] Figure 3 It is a flow chart of the engine combustion state prediction method of the present invention.

[0021] In the figure: 1-flame spontaneous radiation intensity perception module, 2-multi-source information detection module, 3-timing control module, 4-state prediction model, 5-engine. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0023] like Figure 1As shown, the engine combustion state prediction system based on multi-source spectral information fusion includes an optical synchronous detection system at the front end and a state prediction model 4 at the back end. The optical synchronous detection system is composed of a flame spontaneous radiation intensity perception module 1, a multi-source information detection module 2 and a timing control module 3. The flame spontaneous radiation intensity perception module 1 and the multi-source information detection module 2 detect the same flame area of ​​the combustion chamber of the engine 5, and the timing control module 3 ensures the synchronization of the timing of the two detection processes, which is convenient for the subsequent data set division. The flame spontaneous radiation intensity perception module 1 obtains the radiation intensity perception data of the flame area, and the multi-source information detection module 2 obtains the multi-source information data of the flame area, and obtains image features and spectral information through feature extraction. The state prediction model 4 is trained by a training set, and the training set is obtained by dividing the extracted image features and spectral information into data sets using the radiation intensity perception data, and should include three sets corresponding to the three states of stable combustion, transition state and flameout state of the engine 5. The trained state prediction model 4 uses the subsequent given spectral information as input and outputs the predicted combustion state to the engine 5 in real time, providing a reference for the control of the engine 5.

[0024] like Figure 1 to Figure 3 As shown in the figure, the engine combustion state prediction method based on multi-source spectral information fusion is combined with Figure 3 As shown, the following steps are included:

[0025] Step 1: Acquisition of detection data from the optical synchronous detection system

[0026] The same flame area of ​​the combustion chamber of the engine 5 is detected by using an optical synchronous detection system. The detection technologies used include flame spontaneous radiation intensity sensing technology and multi-source information detection technology. The flame spontaneous radiation intensity sensing module 1 is used to obtain the radiation intensity sensing data of the flame area, and the multi-source information detection module 2 is used to obtain the multi-source information data of the flame area. The timing control module 3 is used to ensure the timing synchronization of the two. Figure 2 As shown, it can be seen that the pressure curve is highly consistent with the flame spontaneous radiation intensity sensing data except for a certain hysteresis. Therefore, the flame spontaneous radiation intensity sensing data can also characterize the combustion state of the engine 5. The flame spontaneous radiation intensity sensing data in the present invention is mainly used for the division of subsequent data sets.

[0027] Step 2: Extraction of state features

[0028] The multi-source information detection module 2 extracts image features and spectral information from the multi-source information data, and these feature information are closely related to the combustion state of the engine 5, including flame area, perimeter, flame center position and Zernike moment, etc., to obtain a data set;

[0029] Step 3: Information fusion and data set division

[0030] The information fusion process mainly integrates the data obtained by the flame spontaneous radiation intensity perception module 1 and the multi-source information detection module 2 to obtain a training set for the state prediction model 4. The training set uses the radiation intensity perception data to divide the extracted image features and spectral information into data sets. The training set used to train the state prediction model 4 should include the complete process of the engine 5 from stable combustion through the transition state to the flameout state. The radiation intensity perception data is used to determine the nodes of the different combustion states of the engine 5, and the combustion state of the engine 5 is divided into three states: stable combustion, transition state and flameout state. The data set obtained in step 2 is divided into the three states to form a training set;

[0031] Step 4: State prediction model training and combustion state prediction

[0032] The state prediction model 4 is trained using the training set obtained in step three. The training set is first normalized in terms of mean variance. The state prediction model 4 is based on partial least squares and artificial neural network machine learning algorithms and is tested using a pre-prepared prediction set. The model parameters are continuously adjusted to determine whether the state prediction model 4 has the optimal prediction accuracy for the prediction set. The trained state prediction model 4 does not need to be trained again in subsequent applications and can use the given spectral information as input to perform real-time prediction of the combustion state and pass it to the engine 5 as a reference for its control.

[0033] In summary, the present invention applies the correlation law between the spectral information characteristics of the combustion flow field and the engine performance and working mode. Based on the principles of monotonicity, sensitivity, stability, measurability and controllability, the spectral characteristic parameters closely related to the engine working mode are selected as perception parameters, and multi-source data fusion is performed using a multi-source spectral information fusion method. Combined with the information fusion method, an engine combustion state prediction machine learning model is trained to construct an engine combustion state prediction system and method.

[0034] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other forms of assembly without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations within the meaning and range of equivalents of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0035] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. The engine combustion state prediction system based on multi-source spectral information fusion is characterized by: The invention comprises a flame spontaneous radiation intensity perception module (1), a multi-source information detection module (2), a timing control module (3) and a state prediction model (4), wherein the timing control module (3) connects the flame spontaneous radiation intensity perception module (1) and the multi-source information detection module (2) to ensure the timing synchronization of the two during the detection process, the flame spontaneous radiation intensity perception module (1) and the multi-source information detection module (2) detect the same flame area of ​​the combustion chamber of the engine (5) to obtain radiation intensity perception data and multi-source information data of the flame area respectively, and the state prediction model (4) is based on partial least squares and artificial neural network. A network machine learning algorithm, a state prediction model (4) is trained by a training set, wherein the training set is obtained by dividing the image features and spectral information into data sets using the radiation intensity perception data, wherein the image features and spectral information are obtained by extracting the features of the multi-source information data, including flame area, perimeter, flame center position and Zernike moment, and the training set should include three sets corresponding to the three states of stable combustion, transition state and flameout state of the engine (5). The trained state prediction model (4) takes the subsequently given spectral information as input and outputs the predicted combustion state to the engine (5) in real time.

2. The engine combustion state prediction method based on multi-source spectral information fusion is characterized by: According to the engine combustion state prediction system of claim 1, the prediction method thereof comprises the following steps: Step 1: Acquisition of detection data from the optical synchronous detection system Using an optical synchronous detection system to detect the same flame area of ​​the combustion chamber of the engine (5), obtaining radiation intensity perception data and multi-source information data of the flame area, ensuring the timing synchronization of the two through a timing control module (3), and using the radiation intensity perception data to characterize the combustion state of the engine (5); Step 2: Extraction of state features Extracting the multi-source information detection module (2) to obtain image features and spectral information in the multi-source information data, including flame area, perimeter, flame center position and Zernike moment, to obtain a data set; Step 3: Information fusion and data set division The data obtained by the flame spontaneous radiation intensity sensing module (1) and the multi-source information detection module (2) are fused, and the nodes of different combustion states of the engine (5) are determined using the radiation intensity sensing data, and the combustion state of the engine (5) is divided into three states: stable combustion, transition state and flameout state, and the data set obtained in step 2 is respectively divided into the three states to form a training set; Step 4: State prediction model training and combustion state prediction The training set obtained in step 3 is used to train the state prediction model (4). The training set is first normalized by mean variance. The state prediction model (4) is based on partial least squares and artificial neural network machine learning algorithms and is tested using a pre-prepared prediction set. The model parameters are continuously adjusted to determine whether the state prediction model (4) has the best prediction accuracy for the prediction set. The trained state prediction model (4) uses the given spectral information as input to perform real-time prediction of the combustion state and transmit it to the engine (5) as a reference for its control.

Images