Method for predicting combustion instability based on pressure fluctuation change

Abstract

The invention discloses a method for predicting combustion instability based on pressure fluctuation change, and relates to a method for predicting combustion instability. The problem that combustion instability cannot be judged through short-time sampling pressure fluctuation data in the prior art is solved. The method comprises the steps that a pressure pulsation time sequence in a combustion chamber is monitored in real time; a rolling window mode is adopted, and a pressure fluctuation arrangement mode with the delay time being 1 and the embedding dimension being 4 in a window is obtained; the arrangement models are further divided into three classes, namely, the class A (arrangement has no turning points), the class B (arrangement has one turning point) and the class C (arrangement has two turning points); and calculating the occurrence probabilities of the A, B and C classes in the rolling window, and predicting the time sequence change development of the pressure pulsation in the combustion chamber based on the probabilities of the A, B and C classes. The invention belongs to the technical field of aerospace propulsion systems.

Description

Technical Field

[0001] This invention relates to a method for predicting combustion instability, belonging to the field of aerospace propulsion system technology. Background Technology

[0002] During the development of liquid propulsion systems, non-periodic, severe pressure pulsations frequently occur within the thrust chamber, manifested as the irregular generation of high-intensity oscillations during combustion. A significant characteristic of this transient abnormal combustion is its non-reproducibility—even when repeated tests are conducted under identical experimental conditions, the dynamic pressure spectrum still exhibits significant random differences. This unpredictability of operating conditions leads to potential hidden failures in the propulsion device, making the development of real-time monitoring technology a crucial issue for ensuring the reliable operation of the thrust chamber, possessing both academic research value and promising engineering applications.

[0003] Pressure changes during combustion often exhibit unpredictable and dramatic oscillations and complex fluctuations, which traditional analytical methods based on linear assumptions cannot accurately capture. For example, when combustion approaches an unstable state, the pressure signal may suddenly experience abnormal fluctuations with exponential growth, while conventional models often underestimate the destructive power of this nonlinear effect. This limitation makes the development of new monitoring technologies capable of tracking dynamic characteristics in real time particularly important. Only by dynamically analyzing the multidimensional changes in combustion states can critical risks be effectively identified, providing early warning and assurance for engineering safety.

[0004] Most existing methods are based on linear models to monitor and warn of combustion stability. However, these methods are difficult to accurately predict random combustion instability in the combustion chamber and are unable to cope with the safety hazards caused by combustion instability. Summary of the Invention

[0005] To address the problem that existing methods cannot determine combustion instability by sampling pressure fluctuation data over a short period of time, this invention proposes a method for predicting combustion instability based on pressure fluctuation changes.

[0006] The technical solution adopted by the present invention to solve the above problems is as follows: The steps of the present invention include: Step 1: Obtain the pressure pulsation time series of the combustion chamber. ; Step 2: Within the scrolling calculation window 1, extract a vector with an embedding dimension of 4 from the window data based on parameters with an embedding dimension of 4 and a delay time of 1: , In the formula, the window length is 60, the embedding dimension is 4, the delay time is 1, and each window generates 57 4-dimensional vectors. Step 3: For each vector... The four data points within are sorted by size, and the index sequence of their original positions is recorded. Step 4: Count the number of groups A, B, and C in window one. , , Calculate the probability of occurrence of groups A, B, and C; Step 5: Determine the status based on the threshold values ​​of the three groups (A, B, and C).

[0007] Furthermore, in step 1, pressure pulsation data of the combustion chamber is obtained through a dynamic pressure sensor. The length of the sliding window 1 is set to 60, and the window interval is 10. When the amount of time series data of pressure pulsation meets the length of the sliding window, step 2 is executed.

[0008] Furthermore, in step 3, the measured pressure pulsation time series are divided into three groups, A, B, and C, based on the number of inflection points in different pressure fluctuation arrangement patterns. Group A refers to groups with no turning points in the pressure fluctuation pattern; Group B refers to a group where there is a turning point in the arrangement of pressure fluctuations; Group C refers to a group where there are two turning points in the pressure fluctuation pattern.

[0009] Furthermore, the formula for calculating the probabilities of occurrence of groups A, B, and C in step 4 is as follows: .

[0010] Furthermore, in step 5, the probabilities of the three categories A, B, and C are monitored respectively. When the probability of A is greater than the threshold of 0.89, combustion instability is considered to have occurred in the combustion chamber; when the probability of B is less than the threshold of 0.08, combustion instability is considered to have occurred in the combustion chamber; and when the probability of C is less than the threshold of 0.07, combustion instability is considered to have occurred in the combustion chamber.

[0011] The beneficial effects of this invention are: 1. This invention can predict combustion instability phenomena that are about to occur in the combustion chamber and provide early warning, thereby controlling the combustion before instability occurs; 2. This invention is based on the changing pattern of pressure pulsation signals when combustion instability occurs. When combustion is merely a noise-driven, stable state, the pressure pulsation exhibits disordered fluctuations, meaning that types B and C have a high probability of occurring. However, when combustion instability occurs, the pressure fluctuations show periodic changes. At this time, the probability of type A increases significantly, while the probability of types B and C decreases significantly. Therefore, the probability changes of types A, B, and C can be used to determine the likelihood of combustion instability occurring. Attached Figure Description

[0012] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0013] Specific implementation method one: as follows Figure 1 As shown, a method for predicting combustion instability based on pressure fluctuations includes the following steps: Step 1: Obtain the pressure pulsation time series of the combustion chamber based on a dynamic pressure sensor. High-frequency pressure pulsation signals can usually reflect the "precursors" of combustion instability. Based on the analysis of these high-frequency pressure pulsation signals, it is often possible to curb and regulate combustion instability before it occurs. Step 2: Set the length of the sliding window 1 to 60. When the amount of time-series data on pressure pulsations meets the length of the sliding window, start executing S03; otherwise, continue to acquire the pressure pulsation time-series data in the combustion chamber from S01. Setting a shorter data sliding window helps us to determine the combustion state in a shorter time, which is faster than performing Fourier transform after long-term sampling. Step 3: Within the scrolling calculation window 1, extract a vector with an embedding dimension of 4 from the window data based on parameters with an embedding dimension of 4 and a delay time of 1: , Here, the window length is 60, the embedding dimension is 4, and the delay time is 1, resulting in 57 4-dimensional vectors generated per window. When the embedding dimension is too large, the extracted vector dimension increases, and the subsequent arrangement patterns also increase, which increases the computational load. At the same time, it is difficult to determine the combustion state by pressure fluctuation changes. Based on previous research, parameters with an embedding dimension of 4 and a delay time of 1 can better carry out subsequent determination of the combustion state. And for each vector The four data points within are sorted by size, and the index sequence of their original positions is recorded; for example, if The sorted index is (2,1,4,3), corresponding to the sorting pattern. Following these steps will yield different arrangement patterns; Step 4: Based on the number of turning points in different pressure fluctuation patterns, the measured pressure pulsation time series are divided into three groups: A, B, and C. There are a total of 24 different arrangement patterns for S03, but it is difficult to monitor the changes in each arrangement pattern. Therefore, classifying and grouping them for monitoring is a better approach. Group A refers to a pressure fluctuation arrangement without a turning point, such as a monotonically increasing or decreasing arrangement pattern (e.g., (1,2,3,4) or (4,3,2,1)). Group B refers to a situation where there is a turning point in the pressure fluctuation arrangement, such as (1,2,4,3)(1,2,4,3)(1,2,4,3) (increase first, then decrease). Group C refers to a pressure fluctuation pattern with two turning points, such as (1,3,2,4)(1,3,2,4)(1,3,2,4) (alternating increases and decreases); Step 5: Count the number of groups A, B, and C in window one. Calculate the probability of occurrence of groups A, B, and C: ; Step 6: Determine the status based on the thresholds set for each of the three groups (A, B, and C): The probability of monitoring class A exceeds the class A threshold of 0.89; the probability of monitoring class B is lower than the class B threshold of 0.08; and the probability of monitoring class C is lower than the class C threshold of 0.07.

[0014] If we rely solely on a set of trigger thresholds, there will be a significant number of misjudgments and issues with unstable combustion. Therefore, in this case study, the combustion status is determined by combining these three types of overall triggering conditions, which greatly reduces the likelihood of misjudgments and failures to identify the combustion status.

[0015] If two out of these three groups trigger their respective thresholds, then the last 50 data sets in the scrolling window are retained. If the number of data sets in the scrolling window in step 2 is not met, then the capture of pressure pulsation timing data based on the dynamic pressure sensor in step 1 continues.

[0016] If two of these three groups trigger their respective thresholds, combustion instability is considered to have occurred. Subsequent control measures will then be implemented to manage this combustion instability.

[0017] Working principle Based on the variation of pressure pulsation signals when combustion instability occurs, when combustion is merely a noise-driven stable state, pressure pulsation exhibits disordered fluctuations, meaning that types B and C are highly probable. However, when combustion instability occurs, pressure fluctuations show periodic changes. In this case, the probability of type A increases significantly, while the probability of types B and C decreases significantly. Therefore, the probability changes of types A, B, and C can be used to determine the likelihood of combustion instability.

[0018] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent substitutions, and improvements made to the above embodiments without departing from the scope of the present invention, based on the technical essence of the present invention and within the spirit and principles of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. A method for predicting combustion instability based on pressure fluctuation changes, characterized in that, The specific steps include: Step 1: Obtain the pressure pulsation time series of the combustion chamber. ; Step 2: Within the scrolling calculation window 1, extract a vector with an embedding dimension of 4 from the window data based on parameters with an embedding dimension of 4 and a delay time of 1: ， In the formula, the window length is 60, the embedding dimension is 4, the delay time is 1, and each window generates 57 4-dimensional vectors. Step 3: For each vector... The four data points within are sorted by size, and the index sequence of their original positions is recorded. Step 4: Count the number of groups A, B, and C in window one. , , Calculate the probability of occurrence of groups A, B, and C; Step 5: Determine the status based on the threshold values ​​of the three groups (A, B, and C).

2. The method for predicting combustion instability based on pressure fluctuations according to claim 1, characterized in that, In step 1, pressure pulsation data of the combustion chamber is obtained through a dynamic pressure sensor. The length of the sliding window 1 is set to 60, and the window interval is 10. When the amount of time series data of pressure pulsation meets the length of the sliding window, step 2 is executed.

3. The method for predicting combustion instability based on pressure fluctuation changes according to claim 1, characterized in that, In step 3, the measured pressure pulsation time series are divided into three groups, A, B, and C, based on the number of inflection points in different pressure fluctuation arrangement patterns. Group A refers to groups with no turning points in the pressure fluctuation pattern; Group B refers to a group where there is a turning point in the arrangement of pressure fluctuations; Group C refers to a group where there are two turning points in the pressure fluctuation pattern.

4. The method for predicting combustion instability based on pressure fluctuation changes according to claim 1, characterized in that, The formula for calculating the probabilities of the three groups A, B, and C in step 4 is: 。 5. The method for predicting combustion instability based on pressure fluctuation changes according to claim 1, characterized in that, In step 5, the probabilities of the three categories A, B, and C are monitored respectively. When the probability of A is greater than the threshold of 0.89, combustion instability is considered to have occurred in the combustion chamber; when the probability of B is less than the threshold of 0.08, combustion instability is considered to have occurred in the combustion chamber; when the probability of C is less than the threshold of 0.07, combustion instability is considered to have occurred in the combustion chamber.

Images