An automobile engine exhaust temperature sensor data fusion processing method

Abstract

The application discloses a kind of automobile engine exhaust temperature sensor data fusion processing methods, first by multiple exhaust temperature sensors of different positions to collect engine exhaust temperature, and to abnormal value data is carried out abnormal value correction;Then the corrected data is decomposed into multiple intrinsic mode functions by variational mode decomposition, is divided into different time scale components based on the center frequency of mode component, and extracts customized feature set. Next, an improved attention mechanism time convolution network with physical constraints is used to fuse sensor distance and time delay parameters, achieving multi-scale spatio-temporal feature weighted fusion. Finally, the final fused temperature value is output through weighted pooling. The application significantly improves the accuracy and robustness of exhaust temperature monitoring, solving the fusion bias problem caused by multi-sensor data spatio-temporal asynchrony, especially suitable for engine exhaust temperature state real-time monitoring system.

Description

Technical Field

[0001] This invention relates to the field of automotive engine exhaust temperature detection technology, and specifically to a method for data fusion processing of automotive engine exhaust temperature sensors. Background Technology

[0002] To meet the stringent China VI emission standards, a three-way catalytic converter and a particulate filter are added to the exhaust system of automobile engines. The main function of the three-way catalytic converter is to convert harmful gases such as carbon monoxide, hydrocarbons, and nitrogen oxides emitted from vehicle exhaust into harmless carbon dioxide, water, and nitrogen through oxidation and reduction reactions under the action of a catalyst adjusted according to exhaust temperature and engine operating conditions, thereby meeting emission standards. The particulate filter is mainly used to capture particulate matter in the exhaust gas and control the combustion process of particles based on exhaust temperature to achieve optimal combustion efficiency and reduce particulate emissions. Since the operation of both the three-way catalytic converter and the particulate filter depends on exhaust temperature data, the accuracy and real-time nature of the exhaust temperature data greatly affect the performance of both.

[0003] In actual operating conditions, exhaust temperature data collected by a single exhaust temperature sensor is often subject to significant noise due to various interferences. This noise may originate from engine vibrations, severe fluctuations in exhaust flow, or minor interference from the sensor's own electronic components, causing the collected temperature values ​​to exhibit irregular jumps and spikes. Secondly, the accuracy of exhaust temperature data collected by a single sensor is difficult to guarantee. For example, after long-term use, exhaust temperature sensors may develop calibration deviations, resulting in a fixed offset between the detected value and the actual temperature; in high-temperature environments, the sensitivity of the exhaust temperature sensor decreases, failing to accurately reflect the true temperature. Furthermore, exhaust temperature data collected by a single sensor is unstable; the temperature at the same location can fluctuate significantly within a short period due to sensor response delays, external electromagnetic interference, and other factors, complicating temperature monitoring. Therefore, to obtain accurate exhaust temperature data, multiple exhaust temperature sensors need to be installed within the exhaust system, with these sensors installed at both ends of the three-way catalytic converter and the particulate filter. However, how to fuse the exhaust temperature data collected by these exhaust temperature sensors, which may be subject to noise interference, insufficient accuracy, or lack of stability, in order to obtain real-time and accurate exhaust temperature data for the entire exhaust system is a problem that urgently needs to be solved. Summary of the Invention

[0004] The present invention addresses the problem that it is difficult to obtain real-time and accurate exhaust temperature data of current automobile engines, and provides a method for fusion processing of exhaust temperature sensor data of automobile engines, which has the characteristics of fast processing speed, good fusion effect and high accuracy.

[0005] To solve the above problems, the present invention is achieved through the following technical solution:

[0006] A method for fusing and processing data from automotive engine exhaust temperature sensors includes the following steps:

[0007] Step 1: Use two or more exhaust temperature sensors installed at different locations on the car engine to collect the exhaust temperature of the car engine over a period of time.

[0008] Step 2: Calculate the average exhaust temperature of all temperature sensors at each sampling point, and based on this, screen out the abnormal exhaust temperatures of the temperature sensors at the corresponding sampling points and calibrate them.

[0009] Step 3: Use variational mode decomposition algorithm to perform variational mode decomposition on the exhaust temperature collected by each exhaust temperature sensor to obtain more than two IMF components and their corresponding center frequencies for each exhaust temperature sensor.

[0010] Step 4: Assign each IMF component of each exhaust temperature sensor to the corresponding time scale according to its corresponding center frequency, and extract features within the time window corresponding to each time scale to obtain the multi-scale feature set of each exhaust temperature sensor.

[0011] Step 5: Feed the multi-scale feature sets of each exhaust temperature sensor into the temporal convolutional network to obtain the temperature features of each exhaust temperature sensor at different time scales.

[0012] Step 6: Input the temperature features of each exhaust temperature sensor at different time scales into the spatiotemporal attention mechanism network to obtain the temperature enhancement features of each exhaust temperature sensor at different time scales;

[0013] Step 7: Use a linear interpolation function to dimensionally align the temperature enhancement features of the exhaust temperature sensor at different time scales to obtain the temperature alignment features of each exhaust temperature sensor at different time scales.

[0014] Step 8: Dynamically assign time scale weights according to the importance of features at different time scales, and use the time scale weights to perform weighted summation of the temperature alignment features of each exhaust temperature sensor at different time scales to obtain the temperature fusion features of each exhaust temperature sensor.

[0015] Step 9: Perform average pooling to integrate the temperature fusion features of all exhaust temperature sensors to obtain multi-sensor temperature integration features, and map the multi-sensor temperature integration features to the final exhaust temperature value.

[0016] In step 1 above, the sampling frequency of the exhaust temperature sensor is an integer multiple of 1 Hz and greater than or equal to 1 Hz, and the sampling duration is an integer multiple of 10 minutes and greater than or equal to 10 minutes.

[0017] In step 2 above, the specific process of screening out the abnormal exhaust temperatures of each temperature sensor at the corresponding sampling point and calibrating them is as follows:

[0018] When the temperature sensor detects the exhaust temperature x at sampling point t m (t) and the average exhaust temperature at sampling point t If the difference is within the preset allowable range, then the exhaust temperature x is retained. m (t);

[0019] When the temperature sensor detects the exhaust temperature x at sampling point t m (t) and the average exhaust temperature at sampling point t If the temperature exceeds the preset allowable range, the average exhaust temperature at the previous sampling point t-1 will be used. Replace this exhaust temperature x m (t);

[0020] In the above, m = 1, 2, ..., M, where M is the number of exhaust temperature sensors, and t = 1, 2, ..., T, where T is the sampling duration.

[0021] The specific process of step 4 above is as follows:

[0022] Step 4.1: Convert the individual IMF components u of each exhaust temperature sensor m,k According to its corresponding center frequency ω m,k Assigned to the corresponding time scale, that is:

[0023] When ω m,k When the frequency is ≥1Hz, the IMF component of the exhaust temperature sensor will be u m,k Assign to a second scale; within the second scale, use a time window of size 1 second to assign the IMF component u. m,k The time windows are divided into segments, with the number of segments being T. s ;

[0024] When 1 / 600Hz < ω m,k When <1Hz, the IMF component u of the exhaust temperature sensor will be... m,k Assign to a minute scale; within the minute scale, use a time window with a window size of 1 minute for the IMF component u. m,kThe time windows are divided into segments, with the number of segments being T. min ;

[0025] When ω m,k When the frequency is ≤1 / 600Hz, the IMF component u of the exhaust temperature sensor will be... m,k Assigned to a 10-minute scale; within the 10-minute scale, IMF components are assigned using a time window with a window size of 10 minutes. m,k The time windows are divided into segments, with the number of segments being T. 10min ;

[0026] Step 4.2: For each IMF component u of each exhaust temperature sensor m,k Based on the allocated time scale, features within the corresponding time window are extracted;

[0027] For IMF components allocated to the second scale u m,k Utilizing each time window i s The exhaust temperature of all sampling points within the range is calculated, and the mean value of the four time-domain features of each time window is obtained. variance Skewness and peak amplitude And obtain the IMF component u m,k The second feature F m,k,s ;

[0028] For IMF components allocated to the minute scale u m,k Utilizing each time window i min The exhaust temperature of all sampling points within the range is calculated, and the mean value of five time-domain features for each time window is obtained. variance energy Skewness and kurtosis And two derived features, namely the correlation coefficient of the mean values ​​at adjacent time points. and mean series trend factor And obtain the IMF component u m,k minute feature F m,k,min ;

[0029] For IMF components assigned to a 10-minute scale, u m,k Utilizing each time window i 10min The exhaust temperature of all sampling points within the range is calculated, and the mean value of the four time-domain features of each time window is obtained. variance energy and kurtosis And two derived long-term trend characteristics, namely long-term standard deviation. and sequence cumulative sum And obtain the IMF component u m,k 10-minute feature F m,k,10min ;

[0030] Step 4.3: For each exhaust temperature sensor, calculate the second characteristic F of all IMF components assigned to the second scale. m,k,s Substitute into the second feature set F m,s In the middle, the minute features F of all IMF components assigned to the minute scale m,k,min Included in the minute feature set F m,min In the middle, the 10-minute features F of all IMF components assigned to the 10-minute scale m,k,10min Included in the 10-minute feature set F m,10min Thus, the multi-scale feature set F of each exhaust temperature sensor is obtained. m ={F m,s ,F m,min ,F m,10min};

[0031] In the above, m = 1, 2, ..., M, where M is the number of exhaust temperature sensors; k = 1, 2, ..., K, where K is the number of IMF components for each exhaust temperature sensor; i s =1,2,…,T s T s i is the number of time windows on a second scale. min =1,2,…,T min T min i represents the number of time windows on a minute scale. 10min =1,2,…,T 10min T 10min This represents the number of time windows on a 10-minute scale.

[0032] The specific process of step 6 above is as follows:

[0033] Step 6.1: Construct the current exhaust temperature sensor m and other exhaust temperature sensors m ′ Related features

[0034]

[0035] Step 6.2: Use 1×1 convolution to analyze the correlation features between the current exhaust temperature sensor m and other exhaust temperature sensors m′. Dimensional compression is performed to obtain the dimensional compression features of the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0036] Step 6.3: Compress the dimensionality of the current exhaust temperature sensor m with that of other exhaust temperature sensors m′. The data is fed into the tanh function activation layer to obtain the time attention score function between the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0037]

[0038] Step 6.4: Combine the time attention score function of the current exhaust temperature sensor m with that of other exhaust temperature sensors m′. The weights are fed into the softmax function activation layer to obtain the time attention weights of the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0039]

[0040] Step 6.5: Weight the time attention of the current exhaust temperature sensor m with that of other exhaust temperature sensors m′. The time delay characteristics of the current exhaust temperature sensor m and other exhaust temperature sensors m′ The data is fed into the weighted fusion layer to obtain the time attention features of the current exhaust temperature sensor m.

[0041]

[0042] Step 6.6: Construct the relative distance characteristics between the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0043]

[0044] Step 6.7: Connect the current exhaust temperature sensor m with other exhaust temperature sensors m ′ Relative distance features The input is fed into the tanh function activation layer to obtain the spatial attention score function between the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0045]

[0046] Step 6.8: Combine the spatial attention score function of the current exhaust temperature sensor m with that of other exhaust temperature sensors m′. The data is fed into the softmax function activation layer to obtain the spatial attention weights between the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0047]

[0048] Step 6.9: Assign spatial attention weights to the current exhaust temperature sensor m and other exhaust temperature sensors m′. The relative distance characteristics between the current exhaust temperature sensor m and other exhaust temperature sensors m′ The data are fed into the weighted fusion layer to obtain the spatial attention features of the current exhaust temperature sensor m.

[0049]

[0050] Step 6.10: Utilize 1×1 convolution to apply spatial attention features of the current exhaust temperature sensor m. Dimensional compression is performed to obtain the temperature enhancement characteristics of the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0051] In the above formula, This section describes the correlation characteristics between the current exhaust temperature sensor m and other exhaust temperature sensors m′. `concat(*)` indicates concatenation. The temperature characteristics of the current exhaust temperature sensor m. Let m represent the temperature characteristics of other exhaust temperature sensors m; m, m′=1,2,…,M, where M is the number of exhaust temperature sensors; Let tanh(*) be the time attention score function between the current exhaust temperature sensor m and other exhaust temperature sensors m′, and let τ be the tanh function. m,m′ τ is the signal transmission time delay between the current exhaust temperature sensor m and other exhaust temperature sensors m′. m,m′ =d m,m′ / v,d m,m′ W represents the relative distance between the current exhaust temperature sensor m and other exhaust temperature sensors m′, where v is the exhaust flow velocity; time1 Output the weight matrix U for the score. time1 V is the feature transformation weight matrix. time1 This is the time delay weight matrix; The time attention weights of the current exhaust temperature sensor m and other exhaust temperature sensors m′; The time attention characteristics of the current exhaust temperature sensor m, The time attention weights of the current exhaust temperature sensor m and other exhaust temperature sensors m′; The relative distance characteristics between the current exhaust temperature sensor m and other exhaust temperature sensors m′ are given. For the time attention characteristics of other exhaust temperature sensors m′; W is the spatial attention score function between the current exhaust temperature sensor m and other exhaust temperature sensors m′. sp1U is the spatial attention weight matrix. sp1 Input the weight matrix V for spatial attention. sp This is the spatial attention distance weight matrix; The spatial attention weights of the current exhaust temperature sensor m and other exhaust temperature sensors m′ are used. The spatial attention characteristics of the current exhaust temperature sensor m.

[0052] Compared with the prior art, the present invention has the following characteristics:

[0053] 1. This invention uses variational mode decomposition algorithm to separate sensor temperature signals into different frequencies, and accurately maps them to different time scales based on the center frequency. The second scale captures transient high-frequency fluctuations, the minute scale reflects medium-term trends, and the 10-minute scale represents long-term steady state. Combined with a temporal convolutional network, it solves the problem of poor adaptability of traditional methods to complex temperature dynamic changes.

[0054] 2. This invention employs a dynamic threshold calibration strategy, using the average temperature of the current or previous moment as a benchmark. Abnormal data with deviations exceeding 15% are replaced, avoiding measurement distortion caused by a single sensor failure. Compared to traditional filtering algorithms, this method does not rely on prior noise models and is more robust to random errors in the sensor array. Combined with VMD's ability to decompose nonlinear signals, the original temperature sequence is decomposed into multiple IMF components, effectively separating noise from the true signal and solving the problem of existing technologies relying on manual tuning of filtering parameters.

[0055] 3. This invention utilizes the parallel convolution operation characteristics of temporal convolutional networks. Compared with traditional Bayesian fusion algorithms, this invention achieves batch data processing by expanding causal convolution and residual connections, meeting the real-time sampling requirements of 10Hz for automobile engines.

[0056] 4. This invention introduces an attention mechanism after the temporal convolutional network, which includes a time lag factor τ and a sensor distance parameter d. The model can learn automatically, with higher weights for sensors that are closer, and more attention is given to transient temperature change points in the time series, without the need for manual preset of sensor importance.

[0057] 5. This invention completely eliminates the reliance on parameters such as sensor noise distribution and prior confidence in traditional methods. The model can automatically weaken the weight of abnormal sensors through multi-scale feature fusion, while traditional methods require manual recalibration of parameters, which improves the reliability of the system under complex working conditions and eliminates the need for regular maintenance of algorithm parameters. Attached Figure Description

[0058] Figure 1 This is a flowchart of a method for fusion processing of exhaust temperature sensor data from an automotive engine.

[0059] Figure 2 This is a structural diagram of a spatiotemporal attention mechanism network.

[0060] Figure 3 A comparison chart showing the fusion of total temperature data from multiple sensors for exhaust temperature.

[0061] Figure 4 This is a fusion diagram of total temperature data from multiple sensors for exhaust temperature.

[0062] Figure 5 This is a comparison chart of 10-minute temperature data fusion from multiple sensors for exhaust temperature.

[0063] Figure 6 This is a fusion graph of 10-minute temperature data from multiple sensors for exhaust temperature. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific examples and the accompanying drawings.

[0065] A method for fusion processing of exhaust temperature sensor data from an automotive engine, such as... Figure 1 As shown, the specific steps include the following:

[0066] Step 1: Using M (M>1) exhaust temperature sensors installed at different locations on the car engine, collect the exhaust temperature x of the car engine over a period of time T. m (t), where m = 1, 2, ..., M, t = 1, 2, ..., T.

[0067] In this embodiment, the number of exhaust temperature sensors in the automotive engine is M = 4. These 4 exhaust temperature sensors are respectively installed at both ends of the three-way catalytic converter and the particulate filter. Each exhaust temperature sensor has the same sampling frequency HZ and sampling duration T, where the sampling frequency HZ is an integer multiple of 1 Hz and greater than or equal to 1 Hz, and the sampling duration T is an integer multiple of 10 minutes and greater than or equal to 10 minutes. In this embodiment, the exhaust temperature sensor has a sampling frequency of 10 Hz, a sampling duration of 60 minutes, and 36,000 sampling points.

[0068] Step 2: Calculate the average exhaust temperature of all temperature sensors at each sampling point. Based on this, the exhaust temperature x at the corresponding sampling point of the temperature sensor was selected. m The abnormal exhaust temperature (t) was measured and calibrated.

[0069] Step 2.1: Calculate the average exhaust temperature of M temperature sensors at sampling point t.

[0070]

[0071] Step 2.2: Set the exhaust temperature x at sampling point t for each temperature sensor. m (t) and the average exhaust temperature at sampling point t The data is compared, and then abnormal temperature data is filtered out and calibrated.

[0072] When the temperature sensor detects the exhaust temperature x at sampling point t m (t) and the average exhaust temperature at sampling point t The difference is within the preset allowable range, such as This indicates the exhaust temperature x at sampling point t. m (t) is normal and does not require calibration; retain this exhaust temperature x. m (t), i.e., x m (t)=x m (t);

[0073] When the temperature sensor detects the exhaust temperature x at sampling point t m (t) and the average exhaust temperature at sampling point t Exceeding the preset allowable range, such as This indicates the exhaust temperature x at sampling point t. m (t) is an anomaly and needs to be corrected by taking the average exhaust temperature from the previous sampling point t-1. Replace this exhaust temperature x m (t), that is

[0074] Step 3: After processing the abnormal exhaust temperature data, use the variational mode decomposition algorithm to perform variational mode decomposition on the exhaust temperature collected by each exhaust temperature sensor, and obtain the K IMF components u of each exhaust temperature sensor. m,k and its corresponding center frequency ω m,k , where k = 1, 2, ..., K.

[0075] Variational Mode Decomposition (VMD) is a signal processing method based on a variational framework, aiming to decompose complex signals into multiple modal components with different center frequencies. After completing the calibration of anomalous data at all sampling times, variational mode decomposition is performed on the complete time series of each sensor to extract the modal components and center frequencies. In this embodiment, for each temperature sensor m, the calibrated time series {x} is taken. m (1),x m (2),…,x mUsing (T)} as the input signal, variational mode decomposition is applied to obtain K (K>1) intrinsic mode functions (IMF) components, each IMF component being a time series {u m,k (1),u m,k (2),…,u m,k (T)}. In this embodiment, K = 4. Then, each IMF component u of each exhaust temperature sensor is automatically solved through iterative optimization using the VMD algorithm. m,k The center frequency is ω m,k .

[0076] Step 4: Convert the K IMF components u of each exhaust temperature sensor m,k According to its corresponding center frequency ω m,k The features are assigned to corresponding time scales, and features are extracted within the time window corresponding to each time scale to obtain the multi-scale feature set F for each exhaust temperature sensor. m .

[0077] Multi-scale feature extraction is a method based on the frequency characteristics of a signal. It assigns the signal to different time scales, such as seconds, minutes, and 10-minute scales, and extracts features that characterize the signal at each scale. These features are then fused to form a feature set. Its core principle is to leverage the differences in signal sensitivity at different time scales to comprehensively capture transient, medium-term, and long-term features.

[0078] Step 4.1, Time Scale Allocation and Time Window Division: Based on IMF component u m,k center frequency ω m,k Dividing time scales and windows:

[0079] (1) When ω m,k When ≥1Hz, the IMF component u m,k Assigned to a second scale. Within the second scale, a time window of size 1 second is used to assign the IMF component u. m,k The time windows are divided into segments, with the number of segments being T. s The number of exhaust temperature sampling points included in each time window is N. s The second-scale time window maintains one time step per second to ensure that the time resolution is consistent with the output frequency.

[0080] In this embodiment, the sampling frequency is 10 Hz and the sampling duration is 60 minutes, so the IMF components u are divided into seconds. m,k Classified as T s = 3600 time windows, each containing N s = 10 sampling points.

[0081] (2) When 1 / 600Hz < ω m,kWhen <1Hz, the IMF component u m,k Assigned to a minute scale. Within the minute scale, IMF components are assigned using a 1-minute time window. m,k The time windows are divided into segments, with the number of segments being T. min The number of exhaust temperature sampling points included in each time window is N. min The minute-scale time window is aligned with the minute-level scale, enhancing the time granularity of medium-term trend characteristics.

[0082] In this embodiment, the sampling frequency is 10 Hz and the sampling duration is 60 minutes, so the IMF components u are divided into minute-scale values. m,k Classified as T min = 60 time windows, each containing N min = 600 sampling points.

[0083] (3) When ω m,k When ≤1 / 600Hz, the IMF component u m,k Assigned to a 10-minute scale. Within the 10-minute scale, IMF components u are assigned using a time window with a window size of 10 minutes. m,k The time windows are divided into segments, with the number of segments being T. 10min The number of exhaust temperature sampling points included in each time window is N. 10min The 10-minute time window is adapted to a 10-minute time scale, covering the long-term characteristics of the entire monitoring cycle.

[0084] In this embodiment, the sampling frequency is 10 Hz and the sampling duration is 60 minutes, so the IMF components u are divided into 10-minute scales. m,k Classified as T 10min = 6 time windows, each containing N 10min = 6000 sampling points.

[0085] Step 4.2, Feature Extraction within the Window: For each IMF component u of each exhaust temperature sensor m,k Based on the allocated time scale, features within the corresponding time window are extracted;

[0086] (1) For IMF components u allocated to the second scale m,k Utilizing each time window i s N inside s The exhaust temperature at the sampling point is used to calculate four time-domain features for each time window, and the IMF component u is obtained. m,k Second characteristics Where D s=4 represents the number of second features, which are used to enhance the ability to extract high-frequency features and support accurate modeling of transient temperature changes.

[0087] ① The four time-domain features include:

[0088] Time window i s mean

[0089]

[0090] Time window i s variance

[0091]

[0092] Time window i s skewness

[0093]

[0094] Time window i s peak amplitude

[0095]

[0096] In the formula, i s =1,2,…,T s T s N is the number of time windows on a second scale. s Let i be the number of sampling points per time window on a second scale. s -1)×N s For time window i s The starting position, where n is the offset.

[0097] (2) For IMF components allocated to the minute scale u m,k Utilizing each time window i min N inside min Five time-domain features and two derived features of the exhaust temperature at the sampling points for each time window were calculated, and the IMF component u was obtained. m,k minute features Where D min =7 represents the number of minute features, which are used to balance the computational cost and representation accuracy of intermediate features.

[0098] ① The five time-domain features include:

[0099] Time window i min mean

[0100]

[0101] Time window i min variance

[0102]

[0103] Time window i min energy

[0104]

[0105] Time window i min skewness

[0106]

[0107] Time window i min peak

[0108]

[0109] ②The two derived features include:

[0110] Time window i min The correlation coefficient of the mean values ​​of adjacent time points

[0111]

[0112] Time window i min mean series trend factor

[0113]

[0114] In the formula, i min =1,2,…,T min T min N represents the number of time windows on a minute scale. min The number of sampling points per time window on a minute scale, (i min -1)×N min For time window i min The starting position, n is the offset; μ f The mean of all time windows on the minute scale The average value, For time window i min -1 is the mean.

[0115] (3) For the IMF component u of the exhaust temperature sensor assigned to a 10-minute scalem,k Utilizing each time window i 10min N inside 10min The exhaust temperature at the sampling points was used to calculate four time-domain features and two derived long-term trend features for each time window, and the IMF component u was obtained. m,k 10-minute characteristics Where D 10min =6 represents the number of 10-minute features. These 10-minute features are used to preserve key statistics of long-term steady-state features, thereby improving the stability of the fusion results.

[0116] ① The four time-domain features include:

[0117] Time window i 10min mean

[0118]

[0119] Time window i 10min variance

[0120]

[0121] Time window i 10min energy

[0122]

[0123] Time window i 10min peak

[0124]

[0125] ② Two derived long-term trend characteristics include:

[0126] Time window i 10min long-term standard deviation

[0127]

[0128] Time window i 10min Sequence accumulation and

[0129]

[0130] In the formula, i 10min =1,2,…,T 10min T 10min N represents the number of time windows on a 10-minute scale. 10min The number of sampling points per time window on a 10-minute scale, (i10min -1)×N 10min For time window i 10min The starting position, n is the offset; μ m,k,n This represents the average of the nth 10-minute time window; From the first time window to the i-th time window 10min Mean of a 10-minute window sequence of means

[0131]

[0132] It is important to note that if an exhaust temperature sensor has two IMF components assigned to the same time scale, the average of the characteristic values ​​for the same feature should be used as the characteristic value for that feature. For example, if an exhaust temperature sensor has two IMF components assigned to the second scale, the mean, variance, skewness, and peak amplitude of each IMF component should be calculated separately. Then, the average of the means of the two IMF components should be used as the mean of the second feature, the average of the variances of the two IMF components should be used as the variance of the second feature, the average of the skewnesses of the two IMF components should be used as the skewness of the second feature, and the average of the peak amplitudes of the two IMF components should be used as the peak amplitude of the second feature. If an exhaust temperature sensor has no IMF components assigned to a certain time scale, all features corresponding to that time window should be set to 0. For example, if an exhaust temperature sensor has 0 IMF components assigned to the minute scale, then all 5 time-domain features and 2 derived features should be set to 0.

[0133] Step 4.3, Construction of Multi-Scale Feature Set: For each exhaust temperature sensor, construct the second feature F of all IMF components assigned to the second scale. m,k,s Substitute into the second feature set F m,s In the middle, the minute features F of all IMF components assigned to the minute scale m,k,min Included in the minute feature set F m,min In the middle, the 10-minute features F of all IMF components assigned to the 10-minute scale m,k,10min Included in the 10-minute feature set F m,10min Thus, the multi-scale feature set F of each exhaust temperature sensor is obtained. m ={F m,s ,F m,min ,F m,10min}

[0134] Step 5: Assemble the multi-scale feature set F for each exhaust temperature sensor. m The feature sets F at various time scales m,s ,F m,min ,F m,10minThe data are fed into a temporal convolutional network to obtain the temperature features X3 of each exhaust temperature sensor at different time scales.

[0135] The multi-scale feature set F of the exhaust temperature sensor m m ={F m,s ,F m,min ,F m,10mim The input is used as the input to the temporal convolutional network. A temporal convolutional network (TCN) is a convolutional neural network specifically designed for temporal data. It consists of dilated, causal 1D convolutional layers with the same input and output lengths, designed to capture temporal dependencies through convolutional operations. In this embodiment, the temporal convolutional network consists of two dilated causal convolutions and a residual layer.

[0136] Step 5.1, the first dilated causal convolution output X1 is:

[0137]

[0138] In the formula, b1 is the weight matrix when the expansion factor d1 = 1, b1 is the bias vector, BatchNorm is the batch normalization operation, tanh is the hyperbolic tangent activation function, Dropout is the random deactivation operation to prevent overfitting, C1 = 32 output channels, and p1 = 0.1 dropout rate.

[0139] For the second-scale feature set F m,s After applying the first dilated causal convolution, we obtain Its channel count has expanded from 4 dimensions to 32 dimensions.

[0140] For the minute-scale feature set F m,min After applying the first dilated causal convolution, we obtain Its channel count has expanded from 7 dimensions to 32 dimensions.

[0141] For the 10-minute scale feature set F m,10min After applying the first dilated causal convolution, we obtain Its channel count has expanded from 6 dimensions to 32 dimensions.

[0142] Step 5.2, the output X2 of the second dilated causal convolution is:

[0143]

[0144] In the formula, X1 is the output tensor of the first dilated causal convolution. b2 is the weight matrix, b2 is the bias vector, BatchNorm is the batch normalization operation, tanh is the hyperbolic tangent activation function, Dropout is the random deactivation operation to prevent overfitting, the output channels C2 = 64, and the dropout rate p2 = 0.2.

[0145] For the second-scale feature set F m,s After applying the second dilated causal convolution, we obtain Its channel count has expanded from 32 dimensions to 64 dimensions.

[0146] For the minute-scale feature set F m,min After applying the second dilated causal convolution, we obtain Its channel count has expanded from 32 dimensions to 64 dimensions.

[0147] For the 10-minute scale feature set F m,10min After applying the second dilated causal convolution, we obtain Its channel count has expanded from 32 dimensions to 64 dimensions.

[0148] Step 5.3, the residual calculation output, i.e., temperature feature X3, is:

[0149] X3=tanh(X2+X res )

[0150] In the formula, T * T represents the number of windows corresponding to the time scale. s T is the number of time windows on a second scale. min T represents the number of time windows on a minute scale. 10min This represents the number of time windows on a 10-minute scale.

[0151] In the second scale:

[0152] X res,s =W s1×1 *F m,s

[0153] In the formula, This is a 1×1 convolution kernel used to adjust the feature dimensions, * represents the convolution operation, F m,s Let be the input tensor corresponding to the second-scale feature set.

[0154] In minute scale:

[0155] X res,min =W min×1 *F m,min

[0156] In the formula, This is a 1×1 convolution kernel used to adjust the feature dimensions, * represents the convolution operation, F m,minThis is the input tensor corresponding to the minute-scale feature set.

[0157] On a 10-minute scale:

[0158] X res,10min =W 10min1×1 *F m,10min

[0159] In the formula, This is a 1×1 convolution kernel used to adjust the feature dimensions, * represents the convolution operation, F m,10min This is the input tensor corresponding to the 10-minute scale feature set.

[0160] Step 6: Input the temperature features X3 of each exhaust temperature sensor at different time scales into the spatiotemporal attention mechanism network to obtain the temperature enhancement features of each exhaust temperature sensor at different time scales.

[0161] The basic principle of attention mechanisms is to calculate the similarity or relevance between queries and keys, obtain the weight coefficient of the value corresponding to each key, and then sum the weighted values ​​to obtain the final attention value. Spatiotemporal attention mechanism networks combine time delay and spatial distance. By simultaneously introducing bidirectional temporal context information into the spatial attention mechanism and incorporating the relative distance between sensors, they more comprehensively explore the spatiotemporal dependencies between sensors, thereby enhancing the focus on key spatiotemporal information. The principle is as follows: Figure 2 As shown, the specific process is as follows:

[0162] Step 6.1: Construct the correlation features between the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0163]

[0164] In the formula, concat(*) represents concatenation. The temperature characteristics of the current exhaust temperature sensor m. Let m' be the temperature characteristic of other exhaust temperature sensors, where m and m' are exhaust temperature sensor index values, m, m' = 1, 2, ..., M, and M is the number of exhaust temperature sensors.

[0165] Step 6.2: Use 1×1 convolution to analyze the correlation features between the current exhaust temperature sensor m and other exhaust temperature sensors m′. Dimensional compression is performed to obtain the dimensional compression features of the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0166]

[0167] In the formula, conv 1×1 (*) represents a 1×1 convolution, out channels =C2 indicates that the number of output channels is C2.

[0168] Step 6.3: Compress the dimensionality of the current exhaust temperature sensor m with that of other exhaust temperature sensors m′. The data is fed into the tanh function activation layer to obtain the time attention score function between the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0169]

[0170] In the formula, τ m,m′ τ is the signal transmission time delay between the current exhaust temperature sensor m and other exhaust temperature sensors m′. m,m′ =d m,m′ / v,d m,m′ The relative distance between the current exhaust temperature sensor m and other exhaust temperature sensors m′ is given by , v is the exhaust flow velocity; tanh(*) represents the tanh function, W time1 Output a weight matrix for the score. U time1 The feature transformation weight matrix, V time1 This is the time delay weight matrix.

[0171] Step 6.4: Combine the time attention score function of the current exhaust temperature sensor m with that of other exhaust temperature sensors m′. The weights are fed into the softmax function activation layer to obtain the time attention weights of the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0172]

[0173] Step 6.5: Weight the time attention of the current exhaust temperature sensor m with that of other exhaust temperature sensors m′. The time delay characteristics of the current exhaust temperature sensor m and other exhaust temperature sensors m′ The data is fed into the weighted fusion layer to obtain the time attention features of the current exhaust temperature sensor m.

[0174]

[0175] Step 6.6: Construct the current exhaust temperature sensor m and other exhaust temperature sensors m v Relative distance features

[0176]

[0177] In the formula, concat(*) represents concatenation, and d m,m′ The distance between the current exhaust temperature sensor m and other exhaust temperature sensors m′ is given.

[0178] Step 6.7: Determine the relative distance characteristics between the current exhaust temperature sensor m and other exhaust temperature sensors m′. The input is fed into the tanh function activation layer to obtain the spatial attention score function between the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0179]

[0180] In the formula, W sp1 Here is the spatial attention weight matrix. U sp1 Input a weight matrix for spatial attention. V sp1 This is the spatial attention distance weight matrix.

[0181] Step 6.8: Combine the spatial attention score function of the current exhaust temperature sensor m with that of other exhaust temperature sensors m′. The data is fed into the softmax function activation layer to obtain the spatial attention weights between the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0182]

[0183] Step 6.9: Assign spatial attention weights to the current exhaust temperature sensor m and other exhaust temperature sensors m′. The relative distance characteristics between the current exhaust temperature sensor m and other exhaust temperature sensors m′ The data are fed into the weighted fusion layer to obtain the spatial attention features of the current exhaust temperature sensor m.

[0184]

[0185] Step 6.10: Utilize 1×1 convolution to apply the spatial attention features of the current exhaust temperature sensor m, i.e., the exhaust temperature enhancement features. Dimensional compression is performed to obtain the temperature enhancement characteristics of the current exhaust temperature sensor m and other exhaust temperature sensors m′.

[0186]

[0187] Temperature enhancement features on a second-scale Temperature enhancement features on a minute scale Temperature enhancement features on a 10-minute timescale

[0188] Step 7: Use a linear interpolation function to enhance the temperature characteristics of each exhaust temperature sensor at different time scales. Dimensional alignment was performed to obtain the temperature alignment features H of each exhaust temperature sensor at different time scales. s,unsampled H min,unsampled H 10min,unsampled .

[0189] Linear interpolation refers to an interpolation method where the interpolation function is a first-order polynomial, and its interpolation error at the interpolation nodes is zero. In this invention, features at minute and 10-minute scales are unified to a second-scale temporal resolution, ensuring that features at different scales can be directly fused in the time dimension, i.e.:

[0190]

[0191] In the formula, LinearInterp(*) represents the linear interpolation function.

[0192] Step 8: Dynamically assign time scale weights based on the importance of features at different time scales, and use these weights to perform a weighted summation of the temperature alignment features of each exhaust temperature sensor at different time scales to obtain the temperature fusion feature of each exhaust temperature sensor.

[0193] Step 8.1: Based on the importance of features at different time scales, dynamically allocate weights for fusion to improve the model's ability to represent temperature changes.

[0194] α = Softmax(w)

[0195] In the formula, Softmax(*) represents the Softmax function. w = [w s ,w min ,w 10min [[α] is a learnable parameter vector, α = [α] s ,α min ,α 10min ] represents the learnable scale weights.

[0196] Step 8.2: Weight the samples based on the assigned weights to obtain the temperature fusion features.

[0197] In the formula, α s +α min +α 10min =1,

[0198] Step 9: Fuse the temperature characteristics of all exhaust temperature sensors. The data is fed into an average pooling layer for integration to obtain the integrated temperature characteristics H of the multiple sensors. pool and integrates the temperature characteristics of multiple sensors H pool The input is fed into the fully connected layer and mapped to the final temperature value Y. out .

[0199] Step 9.1: Temperature fusion characteristics of all exhaust temperature sensors Integration:

[0200]

[0201] Step 9.2: Integrate the temperature characteristics H from multiple sensors pool Perform mapping output:

[0202] Y out =H pool ·W out +b out

[0203] in, As the weight, b out ∈R 1 For the bias term, output

[0204] By integrating information from multiple sensors, a fused temperature value is finally output, ensuring complementary information across multiple scales and improving the accuracy of the results.

[0205] This invention first acquires engine exhaust temperature signals from multiple exhaust temperature sensors located at different positions and uses an averaging method to correct outlier data. Then, variational mode decomposition (VMD) decomposes the corrected data into multiple intrinsic mode functions (EMFs), classifying them into components at different time scales (second, minute, and 10-minute scales) based on the center frequency of each modal component, and extracting customized feature sets (temporal features and derived features). Next, a physically constrained improved attention mechanism temporal convolutional network is used to fuse sensor distance and time delay parameters, achieving multi-scale spatiotemporal feature weighted fusion. Finally, weighted pooling and fully connected layers are used to output the final fused temperature value. This method, employing a multi-scale spatiotemporal feature fusion based on temporal convolutional networks to weight the data acquired and processed by multiple temperature sensors, can adapt to random variations in sensor measurement data and capture long-term dependencies in time-series data, better understanding the importance of sensor data at different time points. This method significantly improves the accuracy and robustness of exhaust temperature monitoring, solves the fusion bias problem caused by the spatiotemporal asynchrony of multi-sensor data, and is particularly suitable for real-time monitoring systems of engine exhaust temperature status.

[0206] The four temperature sensors collected using the method of this invention are processed, and the processing results are as follows: Figure 3 and Figure 4 As shown in the figure, the measurement results of each temperature sensor exhibit significant fluctuations. The engine exhaust temperature, after undergoing multi-scale spatiotemporal feature fusion processing via a temporal convolutional network, becomes more stable and accurate, eliminating random noise interference and random fluctuations in the measurement results from sensors at different locations, thus more accurately reflecting changes in engine exhaust temperature. To illustrate the results at different time scales, the results from the four temperature sensors and the fusion processing over a 10-minute period are plotted as follows. Figure 5 and Figure 6 As shown, the temperature change is more stable on a 10-minute timescale, which can more accurately reflect the trend of engine exhaust temperature change.

[0207] It should be noted that although the embodiments described above are illustrative, they are not intended to limit the invention. Therefore, the invention is not limited to the specific embodiments described above. Any other embodiments obtained by those skilled in the art under the guidance of this invention without departing from its principles are considered to be within the protection scope of this invention.

Claims

1. A method for data fusion processing of automotive engine exhaust temperature sensor data, characterized in that, The steps include the following: Step 1: Use two or more exhaust temperature sensors installed at different locations on the car engine to collect the exhaust temperature of the car engine over a period of time. Step 2: Calculate the average exhaust temperature of all temperature sensors at each sampling point, and based on this, screen out the abnormal exhaust temperatures of the temperature sensors at the corresponding sampling points and calibrate them. Step 3: Use variational mode decomposition algorithm to perform variational mode decomposition on the exhaust temperature collected by each exhaust temperature sensor to obtain more than two IMF components and their corresponding center frequencies for each exhaust temperature sensor. Step 4: Assign each IMF component of each exhaust temperature sensor to the corresponding time scale according to its corresponding center frequency, and extract features within the time window corresponding to each time scale to obtain the multi-scale feature set of each exhaust temperature sensor. The specific process is as follows: Step 4.1: Calculate the individual IMF components of each exhaust temperature sensor. Based on its corresponding center frequency Assigned to the corresponding time scale, that is: when At that time, the IMF component of the exhaust temperature sensor will be... Allocation to a second scale; In the second scale, IMF components are paired using a time window of size 1 second. The time windows are divided into segments, and the number of time windows is... ; when At that time, the IMF component of the exhaust temperature sensor will be... Assign to a minute scale; within the minute scale, use a 1-minute time window for IMF components. The time windows are divided into segments, and the number of time windows is... ; when At that time, the IMF component of the exhaust temperature sensor will be... Assigned to a 10-minute scale; within the 10-minute scale, IMF components are assigned using a time window with a window size of 10 minutes. The time windows are divided into segments, and the number of time windows is... ; Step 4.2: For each IMF component of each exhaust temperature sensor Based on its assigned time scale, features within the corresponding time window are extracted to obtain the IMF component. Second characteristics Minute features and 10-minute features ; Step 4.3: For each exhaust temperature sensor, calculate the second characteristic of all IMF components assigned to the second scale. Included in the second feature set In the middle, the minute features of all IMF components assigned to the minute scale. Included in minute feature set In the middle, the 10-minute features of all IMF components assigned to the 10-minute scale. Included in the 10-minute feature set Thus, the multi-scale feature set of each exhaust temperature sensor is obtained. ; in , This refers to the number of exhaust temperature sensors; , The number of IMF components for each exhaust temperature sensor; , The number of time windows on a 1-second scale. , The number of time windows on a 1-minute scale. , The number of time windows on a 10-minute scale; Step 5: Feed the multi-scale feature sets of each exhaust temperature sensor into the temporal convolutional network to obtain the temperature features of each exhaust temperature sensor at different time scales. Step 6: Input the temperature features of each exhaust temperature sensor at different time scales into the spatiotemporal attention mechanism network to obtain the temperature enhancement features of each exhaust temperature sensor at different time scales; Step 7: Use a linear interpolation function to dimensionally align the temperature enhancement features of the exhaust temperature sensor at different time scales to obtain the temperature alignment features of each exhaust temperature sensor at different time scales. Step 8: Dynamically assign time scale weights according to the importance of features at different time scales, and use the time scale weights to perform weighted summation of the temperature alignment features of each exhaust temperature sensor at different time scales to obtain the temperature fusion features of each exhaust temperature sensor. Step 9: Perform average pooling to integrate the temperature fusion features of all exhaust temperature sensors to obtain multi-sensor temperature integration features, and map the multi-sensor temperature integration features to the final exhaust temperature value.

2. The method for data fusion processing of automotive engine exhaust temperature sensor data according to claim 1, characterized in that, In step 1, the sampling frequency of the exhaust temperature sensor is an integer multiple of 1 Hz and greater than or equal to 1 Hz, and the sampling duration is an integer multiple of 10 minutes and greater than or equal to 10 minutes.

3. The method for data fusion processing of automotive engine exhaust temperature sensor data according to claim 1, characterized in that, In step 2, the specific process of screening out the abnormal exhaust temperatures of each temperature sensor at the corresponding sampling point and calibrating them is as follows: When the temperature sensor is at the sampling point exhaust temperature With sampling points Average exhaust temperature If the difference is within the preset allowable range, then the exhaust temperature is retained. ; When the temperature sensor is at the sampling point exhaust temperature With sampling points Average exhaust temperature If the value exceeds the preset allowable range, then the previous sampling point will be... Average exhaust temperature Replace this exhaust temperature ; in , This refers to the sampling duration.

4. The method for data fusion processing of automotive engine exhaust temperature sensor data according to claim 1, characterized in that, In step 4.2: For IMF components assigned to the second scale Utilize each time window The exhaust temperature of all sampling points within the range is calculated, and the mean value of the four time-domain features of each time window is obtained. ,variance skewness and peak amplitude And obtain the IMF component. Second characteristics ; For IMF components allocated to the minute scale Utilize each time window The exhaust temperature of all sampling points within the range is calculated, and the mean value of five time-domain features for each time window is obtained. ,variance ,energy skewness and kurtosis And two derived features, namely the correlation coefficient of the mean values ​​at adjacent time points. and mean series trend factor And obtain the IMF component. minute features ; For IMF components assigned to a 10-minute scale Utilize each time window The exhaust temperature of all sampling points within the range is calculated, and the mean value of the four time-domain features of each time window is obtained. ,variance ,energy and kurtosis And two derived long-term trend characteristics, namely long-term standard deviation. and sequence cumulative sum And obtain the IMF component. 10-minute characteristics .

5. The method for data fusion processing of automotive engine exhaust temperature sensor data according to claim 1, characterized in that, The specific process of step 6 is as follows: Step 6.1: Construct the current exhaust temperature sensor Other exhaust temperature sensors Related features : in, for Indicates splicing, Current exhaust temperature sensor Temperature characteristics, For other exhaust temperature sensors Temperature characteristics; , This refers to the number of exhaust temperature sensors. Step 6.2: Use 1×1 convolution on the current exhaust temperature sensor. Other exhaust temperature sensors Related features Dimensional compression is performed to obtain the current exhaust temperature sensor. Other exhaust temperature sensors Dimensional compression features ; Step 6.3: Set the current exhaust temperature sensor Other exhaust temperature sensors Dimensional compression features Send in In the function activation layer, the current exhaust temperature sensor is obtained. Other exhaust temperature sensors Time attention score function : in, express function, Current exhaust temperature sensor Other exhaust temperature sensors Signal transmission time delay, , Current exhaust temperature sensor Other exhaust temperature sensors The relative distance, The exhaust flow rate is... Output a weight matrix for the score. The feature transformation weight matrix, This is the time delay weight matrix; Step 6.4: Set the current exhaust temperature sensor Other exhaust temperature sensors Time attention score function Send in In the function activation layer, the current exhaust temperature sensor is obtained. Other exhaust temperature sensors Time attention weight : Step 6.5: Set the current exhaust temperature sensor Other exhaust temperature sensors Time attention weight and current exhaust temperature sensor Other exhaust temperature sensors Related features The data is fed into the weighted fusion layer to obtain the current exhaust temperature sensor data. Time attention characteristics : Step 6.6: Construct the current exhaust temperature sensor Other exhaust temperature sensors Relative distance features : in, For other exhaust temperature sensors Temporal attention characteristics; Step 6.7: Set the current exhaust temperature sensor Other exhaust temperature sensors Relative distance features Send in In the function activation layer, the current exhaust temperature sensor is obtained. Other exhaust temperature sensors Spatial attention scoring function : in, Here is the spatial attention weight matrix. Input the weight matrix for spatial attention. This is the spatial attention distance weight matrix; Step 6.8: Set the current exhaust temperature sensor Other exhaust temperature sensors Spatial attention scoring function Send in In the function activation layer, the current exhaust temperature sensor is obtained. Other exhaust temperature sensors Spatial attention weights : Step 6.9: Set the current exhaust temperature sensor Other exhaust temperature sensors Spatial attention weights and current exhaust temperature sensor Other exhaust temperature sensors Relative distance features The data is fed into the weighted fusion layer to obtain the current exhaust temperature sensor data. Spatial attention characteristics : Step 6.10: Use 1×1 convolution on the current exhaust temperature sensor. Spatial attention characteristics Dimensional compression is performed to obtain the current exhaust temperature sensor. Other exhaust temperature sensors Temperature enhancement features .

Images