Oil mist separator control system and control method

Abstract

The application discloses an oil mist separator control system and a control method. The oil mist separator control system comprises an oil mist property monitoring module, a fixed information acquisition module, an oil mist information input module, an information collection monitoring module and an information integration module. The oil mist property monitoring module is used for acquiring an illumination scattering image of an oil mist separation area. The fixed information acquisition module is used for acquiring oil mist viscosity information. The oil mist information input module is used for acquiring oil mist input rate information. The information collection monitoring module is used for acquiring waste oil outflow rate information. The information integration module is used for integrating the oil mist viscosity information, the oil mist input rate information and the waste oil outflow rate information into fusion features. An image information segmentation module is used for extracting oil mist distribution features from the illumination scattering image. A control adjustment module is used for generating control parameters according to the oil mist distribution features and the fusion features. The technical scheme can accurately analyze the distribution of the oil mist in the oil mist separation area by acquiring the illumination scattering image of the oil mist separation area, and effectively determine the oil mist separation effect.

Description

Technical Field

[0001] This application relates to the field of oil-gas separation technology, and more specifically, to an oil mist separator control system and control method. Background Technology

[0002] Oil mist separation technology, as a key means of industrial environmental management, plays an important role in fields such as machining, metal manufacturing, and lubrication systems. This technology effectively captures suspended oil mist particles in the air through physical or chemical methods, achieving oil mist recovery and exhaust gas purification, and has significant environmental value.

[0003] Centrifugal separation, one of the current mainstream oil mist separation technologies, works by injecting oil mist particles into a cavity using a jet. Within the cavity, a rotating airflow is created, and the oil mist particles move towards the inner wall of the separator under inertia, eventually settling due to gravity. In this process, the jet velocity and the tangential angle of incidence are two key parameters affecting separation efficiency. Appropriately increasing the tangential angle of incidence lengthens the airflow path and increases the frequency of inertial collisions; simultaneously, the enhanced centrifugal force (related to the rotation radius and velocity) promotes the migration of oil mist particles towards the outer wall, thus improving separation efficiency. However, an excessively large tangential angle of incidence can induce turbulence, leading to secondary suspension of separated particles and reducing separation efficiency. Similarly, the airflow velocity must be controlled within a reasonable range.

[0004] Existing technologies generally employ a fixed-parameter operating mode, maintaining a constant tangential incident angle and airflow velocity for separation operations, with parameter adjustments only made when the separation efficiency drops to a threshold. This passive adjustment method not only suffers from response lag but also requires a long processing time, making it difficult to achieve optimal overall separation efficiency. Summary of the Invention

[0005] The summary section of this application is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] To address the technical problems mentioned in the background section above, some embodiments of this application provide an oil mist separator control system including:

[0007] The oil mist property monitoring module is used to acquire light scattering images of the oil mist separation area;

[0008] A fixed information acquisition module is used to acquire oil mist viscosity information;

[0009] The oil mist information input module is used to acquire oil mist input rate information;

[0010] The information collection and monitoring module is used to acquire information on the rate of waste oil discharge.

[0011] The information integration module is used to fuse oil mist viscosity information, oil mist input rate information, and waste oil output rate information into a fusion feature.

[0012] The image information segmentation module is used to extract oil mist distribution features from light scattering images;

[0013] The control adjustment module is used to generate control parameters based on the oil mist distribution characteristics and fusion characteristics. The control parameters include airflow rate, airflow tangential incident angle, and oil mist input rate.

[0014] This technical solution acquires light scattering images of the oil mist separation area, enabling precise analysis of the three-dimensional distribution of oil mist within that area. Analysis of oil droplet distribution characteristics allows for effective assessment of the oil mist separation performance. Since the fused features include key parameters such as the physicochemical properties of the oil mist, the oil mist input rate, and the waste oil output rate, they can reflect the dynamic characteristics of the oil mist to be separated within the current separation area over time. By co-processing the spatial distribution analysis results with the fused features, and subsequently optimizing and controlling the system parameters, reasonable adjustments can be made to airflow parameters (including velocity and tangential angle of incidence) and the oil mist input rate, thereby improving oil mist separation efficiency. Through this closed-loop feedback control mechanism, the system can automatically adapt to changes in oil mist characteristics under different operating conditions, reducing manual intervention costs and significantly improving the overall operating efficiency of the oil mist separation system while ensuring separation accuracy.

[0015] The oil mist property monitoring module includes:

[0016] A laser light source is used to emit a laser beam into the oil mist separation area;

[0017] The photosensitive sensor is used to acquire image information of the oil mist separation area;

[0018] An information collection device is used to collect image information and laser wavelength information, and to use the image information and laser wavelength information as a light scattering image.

[0019] This application utilizes a laser source to irradiate the oil mist separation region, stimulating the Tyndall effect and causing laser scattering within the oil mist medium. The oil mist concentration is negatively correlated with the laser scattering characteristics; that is, the higher the oil mist concentration in the region, the more significant the attenuation of laser scattering energy, and the weaker the corresponding diffusion effect. Based on this physical characteristic, the oil mist distribution within the oil mist separation region can be accurately mapped by analyzing the diffusion characteristics of the laser source.

[0020] Furthermore, the photosensitive sensor is oriented vertically downwards; the irradiation direction of the laser light source is perpendicular to the arrangement direction of the photosensitive sensor.

[0021] This design employs a layout where the laser light source's illumination direction is orthogonal to the sensor's orientation, effectively preventing interference from the incident laser on lens imaging. Given that oil mist typically exhibits a distribution characteristic of spreading outwards from the center within the oil mist separation area, the sensor uses a vertical top-down view to acquire image information, enabling a complete and accurate reflection of the spatial diffusion pattern of the oil mist.

[0022] Furthermore, the information integration module includes:

[0023] The information collector is used to collect information on oil mist viscosity, waste oil discharge rate, and oil mist input rate.

[0024] The information preprocessor calculates the cumulative oil mist value in the oil mist separation area per unit time based on the waste oil discharge rate information and the oil mist input rate information, and generates an accumulation value sequence. Each element in the accumulation value sequence is the corresponding cumulative oil mist value per unit time.

[0025] An information extraction unit is used to extract short-term features from the accumulated value sequence;

[0026] The information fusion unit converts oil mist viscosity information into weighted information and multiplies it with short-term features to obtain fused features.

[0027] Based on the short-term characteristics calculated from the oil mist accumulation value, these short-term characteristics can reflect the oil mist accumulation in a short period of time and accurately describe the intrinsic relationship between waste oil discharge and oil mist input under the current environment.

[0028] Furthermore, the short-term features are extracted as follows:

[0029] S1: Obtain the accumulated value sequence N; N = {n1, n2, ..., nn} i …}, where n i This represents the value of the i-th element in the accumulated value sequence, where i represents the index of the oil mist accumulated value;

[0030] S2: Predefine a smoothing factor α, α∈(0,1), and generate a mean sequence ND based on the accumulated value sequence N; ND={d1、d2、…d i …},d i d represents the value of the i-th element in the mean sequence ND; i =α·n i +(1-α)·d i-1 Where i ≥ 1, d1 = n1;

[0031] S3: Calculate the first difference Δd of the mean sequence ND. iGenerate a first-order difference sequence ND 1 ;

[0032] Δd i =d i -d i-1 Where i ≥ 2;

[0033] S4: Calculate the second difference Δ of the mean sequence ND. 2 d i Generate a second-order difference sequence ND 2 ;

[0034] Δ 2 d i =d i -2d i-1 +d i-2 Where i ≥ 3;

[0035] S5: Divide the first-order difference sequence ND 1 and second-order difference sequence ND 2 As a short-term feature.

[0036] In this scheme, differential data is used to obtain short-term features, which can accurately obtain the short-term changes of the accumulated value sequence in a short period of time, understand the amplitude of its numerical fluctuation in a short time interval, and thus accurately describe the efficiency change information in the oil mist separation process.

[0037] Existing methods for extracting short-term features can only capture features within a relatively short period or those near the current time point. They cannot capture short-term features that are comprehensive and considered from the perspective of the entire sequence. Therefore, this application provides the following method for extracting short-term features:

[0038] The short-time features also include residual features. The residual characteristic is the variance of the residual difference sequence R;

[0039] R = N - ND; where N represents the cumulative value sequence and ND represents the mean sequence;

[0040]

[0041] k represents the window size, ΔR i-g μ represents the value of the ig-th element in the residual difference sequence. ΔR ΔR represents the mean of the residual difference sequence within the window. i Let ΔR represent the i-th value in the residual difference sequence. i =R i -R i-1 .

[0042] This technical solution introduces residual features into short-time feature modeling to quantify the stability of the system's operating state. Specifically, the variance characteristics of the residual features can effectively distinguish the system's operating modes: low variance represents a steady-state operating state, while high variance indicates a transient process state. This short-time feature system not only includes time-domain variation information such as first-order / second-order difference sequences, enabling it to characterize dynamic features within a short timescale, but also uses the global statistical properties of the residual features to determine whether the system is in a steady-state or transient state, thereby achieving a multi-dimensional and accurate description of the oil mist separation process.

[0043] Furthermore, the image information segmentation module includes:

[0044] Image information input device, used to input images of light-induced oil fog;

[0045] The feature extraction unit is used to segment the oil mist image under illumination and generate several feature regions related to the oil mist concentration.

[0046] The rendering unit renders the feature regions.

[0047] Furthermore, the feature extraction unit includes:

[0048] An encoder is used to extract attention features from illuminated oil fog images at different resolutions;

[0049] The decoder is used to segment the illuminated oil fog image based on attention features at different resolutions, generating several feature regions related to the oil fog concentration.

[0050] In the technical solution provided in this application, attention features at different resolutions are obtained by an encoder, and then the attention features at different resolutions are extracted by a decoder. This enables the segmentation of the oil fog image without supervision, and accurately describes the distribution of oil fog in the oil fog separation area.

[0051] Furthermore, the encoder includes:

[0052] Each of the m resolution downsampling blocks is used to downsample the illumination and oil fog image, generating m downsampled feature maps S. q q represents the index of the downsampled feature map, where S q The resolution is greater than S q+1 The resolution;

[0053] m-1 hybrid attention fusion blocks will be used to combine S q and S q+1 The input is fed into the corresponding hybrid attention fusion block to generate the encoded feature W. q ;

[0054] The decoder consists of m-1 decoding blocks, each receiving an input encoded feature W.q and W q+1 The system generates fused information and merges the fused information with each other based on pre-set weight parameters to generate several feature regions related to oil mist concentration.

[0055] This scheme extracts global spatial axial features at each resolution level using a compressed attention mechanism, and inputs the generated multi-resolution feature maps into the decoder module. After receiving feature maps of different resolutions, the decoder downsamples the feature maps to different levels of sharpness and fuses them element-wise with the high-resolution feature maps processed at the same level. This mechanism effectively enhances the decoder's ability to acquire contextual semantic information in multi-scale feature spaces, enabling accurate image segmentation.

[0056] As a second aspect of this application, this application provides an oil mist separator control method, which uses the aforementioned oil mist separator control system to generate control parameters. Attached Figure Description

[0057] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application.

[0058] Furthermore, throughout the accompanying drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the elements are not necessarily drawn to scale.

[0059] In the attached diagram:

[0060] Figure 1 This is a schematic diagram of the control system for an oil mist separator.

[0061] Figure 2 This is a schematic diagram of the image information segmentation module. Detailed Implementation

[0062] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While some embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this application. It should be understood that the drawings and embodiments of this application are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0063] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0064] The present application will now be described in detail with reference to the accompanying drawings and embodiments.

[0065] Centrifugal oil mist separators are typically barrel-shaped structures with an air inlet. The oil-gas mixture enters the oil mist separator tangentially at a certain speed through the air inlet, flowing in a vortex pattern. Oil droplets are thrown against the barrel wall by centrifugal force, captured by the wall, and then fall into the oil collection tank at the bottom of the oil mist separator under the influence of gravity.

[0066] like Figure 1 As shown, the oil mist separator control system of this application includes an oil mist property monitoring module, a fixed information acquisition module, an oil mist information input module, an information collection and monitoring module, an information integration module, an image information segmentation module, and a control adjustment module.

[0067] The oil mist property monitoring module is used to acquire light scattering images of the oil mist separation area. The oil mist separation area refers to the working area in the oil mist separator where oil mist separation takes place. By injecting airflow into the oil mist separation area at a certain incident angle, a swirling flow field is formed within the oil mist separation area, and the oil mist particles are separated from the air under the action of centrifugal force.

[0068] Light scattering images refer to the image information formed after the oil mist separation area is illuminated. The oil mist property monitoring module includes a laser source, a photosensor, and an information collection device. The photosensor is used to receive and detect light intensity. The photosensitive area of ​​the photosensor includes multiple individual sensing units, each capable of individually sensing the intensity of the light signal illuminating its surface. Thus, the photosensor can acquire light intensity matrix information. This light intensity matrix information is the image information of the oil mist separation area.

[0069] Furthermore, in this solution, the photosensor uses a light intensity sensor array composed of multiple independent light intensity receivers to acquire light intensity matrix information. As an optical sensor, the light intensity receiver can output corresponding electrical signal values ​​by quantizing incident light signals of different intensities. Compared to traditional image sensors, this design has a significant advantage in resisting oil mist interference: when the surface of the photosensitive area is covered with slight oil mist, the incident light signal can still effectively penetrate the oil mist layer and be detected by the light intensity receiver, thus accurately sensing changes in light intensity. Even in the presence of oil mist, the system can still reliably acquire light intensity distribution data for each area, ensuring the stability of the detection function. This light intensity detection-based solution effectively overcomes the problem of traditional vision lenses being susceptible to contamination interference in oil mist environments.

[0070] The separator contains a cylindrical cavity, which can also be understood as the oil mist separation area. A photosensitive sensor is positioned at the top of the separator, its signal receiving range sufficient to cover the entire cross-section of the separator. The laser source is located on the inner wall of the separator. The photosensitive sensor faces vertically downwards, and the laser source emits laser light parallel to the horizontal plane into the oil mist separation area; that is, the direction of the laser source's illumination is perpendicular to the orientation of the photosensitive sensor. Thus, when the laser light emitted by the laser source illuminates the oil mist separation area, the photosensitive sensor can capture the light signal and obtain the light intensity matrix information of the oil mist separation area.

[0071] The information collection device is used to collect image information and laser wavelength information, and to use the image information and laser wavelength information as a light scattering image. The information collection device includes an image sensor and a spectrometer / photodetector.

[0072] The wavelength of the laser emitted by the laser source is related to the properties of the oil mist. In practice, lasers with the strongest absorption capacity for waste oil are preferred. In the technical solution of this application, an infrared laser is used as the laser source. When the photosensor receives light signals, it only receives light signals with the same wavelength as the laser.

[0073] As mentioned earlier, the photosensor can obtain light intensity matrix information. In the photosensor matrix, the lower the light signal intensity, the higher the oil mist concentration at the corresponding location in the oil mist separation area, and vice versa. Therefore, the distribution of oil mist within the oil mist separation area can be reflected through the oil mist illumination image. The fixed information acquisition module is used to acquire oil mist viscosity information. In one optional embodiment, the fixed information acquisition module can use a light scattering spectrometer to measure the particle size distribution and concentration of oil mist particles, indirectly reflecting the oil mist viscosity information; alternatively, a viscometer or viscosity meter can be used to directly acquire the viscosity information of the oil mist.

[0074] The information collection and monitoring module is used to acquire waste oil discharge rate information. Waste oil discharge rate information refers to the amount of waste oil discharged from the oil mist separator per unit time after the oil-gas mixture has been processed. The waste oil volume can be obtained through a flow meter.

[0075] The oil mist information input module is used to acquire oil mist input rate information. A gas flow meter can be used as the oil mist information input module. The information integration module is used to fuse oil mist viscosity information, waste oil discharge rate information, and oil mist input rate information into a fused feature.

[0076] A higher waste oil discharge rate indicates a faster oil mist separation speed. The oil mist input rate is the rate at which oil mist is input into the oil mist separation zone; both the oil mist input rate and the waste oil discharge rate are measured in terms of the pure volume of waste oil. Therefore, the accumulation of oil mist within the oil mist separation zone can be determined based on the oil mist input rate and the waste oil discharge rate. Specifically, the amount of waste oil accumulated in the oil mist separation zone during that time period is calculated by multiplying the difference between the oil mist input rate and the waste oil discharge rate by the time interval.

[0077] The information integration module includes an information collector, an information preprocessor, an information extraction unit, and an information fusion unit. The information collector integrates oil mist viscosity information, waste oil discharge rate information, and oil mist input rate information.

[0078] The information preprocessor calculates the cumulative oil mist value within the oil mist separation region per unit time based on the waste oil discharge rate and the oil mist input rate, generating an accumulation value sequence. Each element in the accumulation value sequence represents the cumulative oil mist value within the corresponding unit time. The information extraction unit extracts short-time features from the accumulation value sequence. The information fusion unit converts the oil mist viscosity information into weighted information and multiplies it with the short-time features to generate fused features.

[0079] There is a certain relationship between oil mist viscosity and airflow velocity. Generally speaking, the higher the oil mist viscosity, the greater the airflow velocity required for oil mist separation. If the amount of oil mist accumulated in the oil mist separation area per unit time increases, the airflow velocity needs to be increased to accelerate the oil mist separation rate.

[0080] When performing oil mist separation operations on the same batch of oil mist, the type of oil mist usually does not change. Therefore, the viscosity of the oil mist does not change, and its viscosity value is fixed. Thus, it can be regarded as a fixed weighting parameter and weighted and fused with short-term characteristics.

[0081] Thus, by fusing oil mist viscosity information and short-term characteristics, the oil mist viscosity information can be incorporated into the short-term characteristics, allowing the oil mist viscosity information to be taken into account when adjusting operating parameters based on the short-term characteristics.

[0082] The accumulated value sequence refers to the amount of oil mist accumulated over different time periods, reflecting the dynamic changes in oil mist discharge from the oil mist separation zone. Short-term features extracted from the accumulated value sequence can reflect oil mist accumulation information within a nearby time node. If the oil mist accumulation is too high, the oil mist separation rate needs to be increased; conversely, if it is too low, the oil mist input rate needs to be increased.

[0083] Short-term characteristics can reflect changes in oil mist accumulation information within a nearby time period. Compared to the original values ​​of the accumulation value sequence, they pay more attention to changes in oil mist accumulation information, contain more implicit information, and facilitate accurate adjustment of control parameters.

[0084] The extraction method for short-term features is as follows:

[0085] S1: Obtain the accumulated value sequence N; N = {n1, n2, ..., nn} i …}, where n i This represents the value of the i-th element in the accumulated value sequence, where i represents the index of the oil mist accumulated value;

[0086] S2: Predefine a smoothing factor α, α∈(0,1), and generate a mean sequence ND based on the accumulated value sequence N; ND={d1、d2、…d i …},d i d represents the value of the i-th element in the mean sequence ND; i =α·n i +(1-α)·d i-1 Where i ≥ 1, d1 = n1;

[0087] S3: Calculate the first difference Δd of the mean sequence ND. i Generate a first-order difference sequence ND 1 ;

[0088] Δd i =d i -d i-1 Where i ≥ 2;

[0089] S4: Calculate the second difference Δ of the mean sequence ND. 2 d i Generate a second-order difference sequence ND 2 ;

[0090] Δ 2 d i =d i -2d i-1 +d i-2 Where i ≥ 3;

[0091] S5: Divide the first-order difference sequence ND 1 and second-order difference sequence ND 2 As a short-term feature.

[0092] In this scheme, the first-order difference sequence and the second-order difference sequence are used as short-time features. The first-order difference sequence and the second-order difference sequence can reflect the changes in the oil mist accumulation value in a short period of time. However, this change is obtained based on the changes in the entire sequence and cannot reflect the intrinsic relationship between the changes in the oil mist accumulation value at the current time point and the changes in the overall oil mist accumulation value.

[0093] Therefore, short-time features also include residual features. The residual characteristic is the variance of the residual difference sequence R. R = N - ND, where N represents the cumulative value sequence and ND represents the mean sequence. k represents the window size, ΔR i-g μ represents the value of the i-th ig in the residual difference sequence. ΔR ΔR represents the mean of the residual difference sequence within the window. i Let ΔR represent the i-th value in the residual difference sequence. i =R i -R i-1 The value of k affects the smoothing effect and noise sensitivity. A smaller value of k makes the calculation more sensitive to short-term fluctuations and is suitable for capturing transient anomalies. In one embodiment of the invention, k∈[5,10].

[0094] residual characteristic variance To quantify the volatility of oil mist accumulation, a larger variance indicates poorer system stability, suggesting potential system fluctuations. Control parameters need to be adjusted based on the direction of these fluctuations. Residual features are introduced into the short-time characteristics to quantify the stability of oil mist accumulation. Specifically, the variance of the residual features effectively distinguishes changes in oil mist accumulation information; low variance represents a steady-state operation, while high variance indicates a transitional state. This short-time feature system not only includes time-domain variation information such as first- / second-order difference sequences, capable of characterizing dynamic features within a short timescale, but also uses the global statistical properties of the residual features to determine whether the system is in a steady-state or transitional state, thus achieving a multi-dimensional and accurate description of the oil mist separation process. For example, if the first- / second-order difference sequence indicates a current oil mist accumulation value of 10, but the residual features indicate a transitional state, then the reliability of the current oil mist accumulation value of 10 is low, and the oil mist accumulation value may continue to increase.

[0095] The image information segmentation module is used to extract oil fog distribution features from light scattering images. The image information segmentation module includes an image information input unit, a feature extraction unit, and a rendering unit.

[0096] The image information input unit is used to input an image of oil fog under illumination. The feature extraction unit is used to segment the image of oil fog under illumination, generating several feature regions related to the oil fog concentration. The rendering unit is used to render the feature regions; the higher the oil fog concentration, the higher the rendered value. The final rendered image is the oil fog distribution feature.

[0097] like Figure 2 As shown, the feature extraction unit includes an encoder and a decoder. The encoder is used to extract attention features of the illuminated oil fog image at different resolutions. The encoder includes m downsampling blocks at different resolutions and m-1 hybrid attention fusion blocks.

[0098] The illumination and oil fog image is downsampled using m resolution downsampling blocks, generating m downsampled feature maps S. q q represents the index of the downsampled feature map. Where S q The resolution is greater than S q+1 The resolution. In one optional embodiment of this application, m = 4.

[0099] Fast downsampling primarily involves downsampling the input illumination and oil fog image to obtain different resolutions. For example, it merges nine pixel values ​​within a 3x3 window into a single pixel value, with the merged pixel value being the average of the original nine pixel values. Different downsampling modules use different window sizes; the lower the resolution, the larger the window size.

[0100] The core objective of downsampling blocks is to progressively reduce the spatial resolution of an image while extracting multi-level semantic features. Downsampling blocks achieve feature extraction through multiple convolutional operations. Each convolutional layer uses a small-sized kernel to capture local texture patterns.

[0101] The hybrid attention fusion block corresponds to each of the other downsampling blocks except the first downsampling block. Sq and Sq+1 are input to the corresponding hybrid attention fusion block to generate encoded features W. q The number of hybrid attention fusion blocks is one less than the number of downsampling blocks. Therefore, the first hybrid attention fusion block is connected to the second downsampling block, and so on.

[0102] Hybrid attention fusion block combines channel attention and spatial attention for S q and S q+1 Weighting:

[0103] Channel attention: A channel description vector is generated by compressing the spatial dimension through global average pooling, and then the importance weights of each channel are learned through a fully connected network.

[0104] Spatial attention: Using large-size convolutional kernels to scan spatial dimensions, a spatial weight map is generated to focus on areas with high oil mist concentration.

[0105] The decoder is used to segment the illuminated oil fog image based on attention features at different resolutions, generating several feature regions related to the oil fog concentration. The decoder consists of m-1 decoding blocks, each receiving the encoded feature W as input. q and W q+1 The system generates fused information and merges the fused information with each other based on pre-set weight parameters to generate several feature regions related to oil mist concentration.

[0106] After the encoded features enter the decoding block, the core objective of the decoder is to gradually restore the spatial resolution and fuse high-level semantic information with low-level detail information to generate a refined oil mist segmentation result.

[0107] Low-level encoded features are concatenated with the input features of the current decoded block via skip connections, supplementing the details lost during encoding. The concatenated feature map is then further refined through convolution operations, eliminating semantic gaps between features at different scales.

[0108] The decoding block progressively increases the resolution of the feature map through upsampling operations. After each upsampling layer, a convolutional layer and activation function are typically applied to further optimize the feature representation. Ultimately, the decoding block outputs a feature map with the same resolution as the input image, where the value at each location represents the probability that the corresponding pixel belongs to different oil mist concentration levels.

[0109] The control adjustment module generates control parameters based on oil mist distribution and fusion characteristics. These control parameters include airflow rate, airflow tangential incident angle, and oil mist input rate. The core function of this control adjustment module is to generate corresponding control parameters based on the input oil mist distribution and fusion characteristics, thereby achieving effective regulation of the oil mist separation process. During oil mist separation, the light scattering image contains oil mist distribution information. If the concentration in the oil mist separation area is high, it indicates that the airflow rate is too low and the centrifugal force is insufficient. The fusion characteristics reflect the oil mist accumulation characteristics in the oil mist separation area. If the fusion characteristics show that the current oil mist accumulation value is gradually increasing, it indicates that the oil mist accumulation in the current oil mist separation area is increasing, and the oil mist input rate needs to be appropriately reduced. Therefore, the control parameters specifically cover three key dimensions: airflow rate, tangential incident angle, and oil mist input rate.

[0110] Limitations and classifications of control parameters:

[0111] Although airflow velocity, tangential incident angle, and oil mist input rate all affect the oil mist separation process, the process is also limited by factors such as equipment performance, process requirements, and safety regulations in actual production. Therefore, the adjustable range of each parameter has clear boundaries. Based on extensive practical data and analysis of equipment operating characteristics, reasonable range settings have been established for each parameter, as follows:

[0112] Airflow rate: Taking into account the power range of the ventilation equipment, the size of the duct, and the capacity of the separation device, it is divided into 5 levels. The difference between each level is set according to the gradient of the influence of airflow on oil mist separation efficiency, so as to ensure that the movement state of oil mist can be effectively controlled by adjusting the airflow rate under different operating conditions.

[0113] Tangential incident angle: Based on the separator's structural design, the physical characteristics of oil mist particles, and the separation principle, it is divided into 5 levels, each corresponding to a different airflow tangential incident angle. The aim is to optimize the movement trajectory of oil mist in the separation space by adjusting the tangential incident angle, thereby improving separation efficiency.

[0114] Oil mist input rate: Considering the output characteristics of the oil mist generating equipment, the processing capacity of the separator, and the stability requirements of the system, it is divided into three levels. The rate range of each level has been rigorously verified through testing to ensure that the separation system can maintain reliable operation under different input rates. The above-mentioned level division of airflow rate, tangential incident angle, and oil mist input rate is a preferred implementation method obtained from experiments. In practice, the airflow rate level, tangential incident angle level, and oil mist input rate level can also be set to other values.

[0115] Through the above-mentioned gear division, a total of 75 (5×5×3) control parameter combinations are formed, which not only ensures the flexibility of parameter adjustment, but also avoids the problems of increased control complexity and system instability caused by excessively wide parameter range.

[0116] To achieve rapid and accurate selection of control parameters, the oil mist separator control system of this invention incorporates a pre-built, efficient, and reliable database. This database stores a large amount of oil mist distribution characteristics and fusion characteristics under different operating conditions, along with corresponding optimal control parameter selection methods. The specific database construction process is as follows:

[0117] Data acquisition: Real-time acquisition of oil mist distribution characteristics and fusion characteristics under different operating conditions.

[0118] Data processing and annotation: The collected fusion features and oil mist distribution features are preprocessed, including noise filtering and data normalization. Then, through separation efficiency experiments, the optimal control parameters for each feature combination are determined to form accurate annotation data.

[0119] When the control adjustment module needs to adjust the control parameters, it first extracts the oil mist distribution features and fusion features of the current system, and then uses a similarity matching algorithm (such as a matching algorithm based on Euclidean distance or cosine similarity) to search for the record most similar to the current features in the database, thereby quickly matching the corresponding control parameter combination.

[0120] The database contains multiple labeled data sets, each consisting of one feature combination and one optimal control parameter. The feature combination includes one fusion feature and one oil mist distribution feature. When the control adjustment module receives one fusion feature and one oil mist distribution feature, it calculates the similarity between the received fusion feature and oil mist distribution feature and the fusion features and oil mist distribution features of each labeled data set in the database. It then finds the labeled data set with the highest similarity and outputs the optimal control parameter of that labeled data set as the current control parameter.

[0121] The present invention also provides a control method for an oil mist separator, wherein the control method uses the oil mist separator control system to generate control parameters.

[0122] The above description is merely a selection of preferred embodiments of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this application.

Claims

1. An oil mist separator control system, characterized in that, include: The oil mist property monitoring module is used to acquire light scattering images of the oil mist separation area; A fixed information acquisition module is used to acquire oil mist viscosity information; The oil mist information input module is used to acquire oil mist input rate information; The information collection and monitoring module is used to acquire information on the rate of waste oil discharge. The information integration module is used to fuse oil mist viscosity information, oil mist input rate information, and waste oil output rate information into a fusion feature. The image information segmentation module is used to extract oil mist distribution features from light scattering images; The control adjustment module is used to match control parameters based on the oil mist distribution characteristics and fusion characteristics. The control parameters include airflow rate, airflow tangential incident angle and oil mist input rate. The photosensitive sensor is oriented vertically downwards; the direction of the laser light source is perpendicular to the orientation of the photosensitive sensor. The information integration module includes: The information collector is used to collect information on oil mist viscosity, waste oil discharge rate, and oil mist input rate. The information preprocessor calculates the cumulative oil mist value in the oil mist separation area per unit time based on the waste oil discharge rate information and the oil mist input rate information, and generates an accumulation value sequence. Each element in the accumulation value sequence is the corresponding cumulative oil mist value per unit time. An information extraction unit is used to extract short-term features from the accumulated value sequence; The information fusion unit converts oil mist viscosity information into weighted information and multiplies it with short-term features to obtain fused features.

2. The oil mist separator control system according to claim 1, characterized in that, The oil mist property monitoring module includes: A laser light source is used to emit a laser beam into the oil mist separation area; The photosensitive sensor is used to acquire image information of the oil mist separation area; An information collection device is used to collect image information and laser wavelength information, and to use the image information and laser wavelength information as a light scattering image.

3. The oil mist separator control system according to claim 1, characterized in that, The extraction method for the short-term features is as follows: S1: Obtain the accumulated value sequence N; N = {n1, n2, ..., nn} i …}, where n i This represents the value of the i-th element in the accumulated value sequence, where i represents the index of the oil mist accumulated value; S2: Predefine a smoothing factor α, α∈(0,1), and generate a mean sequence ND based on the accumulated value sequence N; ND={d1, d2, ... d i …},d i This represents the value of the i-th element in the mean sequence ND; Where i≥1, d1=n1; S3: Calculate the first difference of the mean sequence ND. Generate a first-order difference sequence ND 1 ; Where i ≥ 2; S4: Calculate the second difference of the mean sequence ND. Generate a second-order difference sequence ND 2 ; Where i ≥ 3; S5: Divide the first-order difference sequence ND 1 and second-order difference sequence ND 2 As a short-term feature.

4. The oil mist separator control system according to claim 3, characterized in that, The short-time features also include residual features. The residual characteristic is the variance of the residual difference sequence R; R = N - ND; where N represents the cumulative value sequence and ND represents the mean sequence; ； k represents the window size. This represents the value of the ig-th element in the residual difference sequence. This represents the mean of the residual difference sequence within the window. Let ΔR represent the i-th value in the residual difference sequence. i =R i -R i−1 .

5. The oil mist separator control system according to claim 4, characterized in that, The image information segmentation module includes: Image information input device, used to input images of light-induced oil fog; The feature extraction unit is used to segment the oil mist image under illumination and generate several feature regions related to the oil mist concentration. The rendering unit renders the feature regions.

6. The oil mist separator control system according to claim 5, characterized in that, The feature extraction unit includes: An encoder is used to extract attention features from illuminated oil fog images at different resolutions; The decoder is used to segment the illuminated oil fog image based on attention features at different resolutions, generating several feature regions related to the oil fog concentration.

7. The oil mist separator control system according to claim 6, characterized in that the encoder... include: Each of the m resolution downsampling blocks is used to downsample the illumination and oil fog image, generating m downsampled feature maps S. q q represents the index of the downsampled feature map, where S q The resolution is greater than S q+1 The resolution; m-1 hybrid attention fusion blocks will be used to combine S q and S q+1 The input is fed into the corresponding hybrid attention fusion block to generate the encoded feature W. q ; The decoder consists of m-1 decoding blocks, each receiving an input encoded feature W. q and W q+1 The system generates fused information and merges the fused information with each other based on pre-set weight parameters to generate several feature regions related to oil mist concentration.

8. A control method for an oil mist separator, wherein control parameters are generated using the oil mist separator control system according to any one of claims 1-7.

Images