High-temperature-resistant and high-pressure-resistant lubricating oil and preparation method thereof

Abstract

The invention relates to the field of intelligent preparation, and provides high-temperature-resistant and high-pressure-resistant lubricating oil and a preparation method thereof.The preparation method comprises the steps that a-olefin synthetic oil in a preset proportion is added into a blending kettle provided with a stirring device, then an antioxidant preservative, an extreme-pressure phosphorus-containing ash-free anti-wear agent and an amine high-temperature antioxidant are sequentially added into the blending kettle, stirring is conducted, and the high-temperature-resistant and high-pressure-resistant lubricating oil is obtained; the preparation method comprises the following steps: uniformly stirring at a first constant temperature to obtain a primary mixture, sequentially adding an antirust agent, a metal cleaning agent and a pour point depressant into the mixture, continuously stirring at a second constant temperature, intelligently controlling the stirring time by deeply analyzing the stirring state of the mixture in the stirring process, and finally preparing the high-temperature-resistant and high-pressure-resistant lubricating oil. Therefore, the quality and the performance of the high-temperature-resistant and high-pressure-resistant lubricating oil product can be effectively ensured.

Description

Technical Field

[0001] The present application relates to the field of intelligent preparation, and more specifically, to a high-temperature and high-pressure resistant lubricating oil and a preparation method thereof. Background Art

[0002] With the rapid development of modern industrial equipment towards high speed and heavy load, the lubrication protection of mechanical equipment under extreme working conditions faces severe challenges. In typical application scenarios such as aerospace power systems, deep-sea exploration equipment, and metallurgical rolling production lines, the lubricating oil not only needs to withstand continuous high temperatures above 200°C, but also maintain an effective lubricating film under contact pressures of several GPa.

[0003] A high-temperature and high-pressure resistant lubricating oil is a special lubricating oil designed specifically to provide effective lubrication protection for mechanical equipment under extreme working conditions such as high temperature and high pressure. Its preparation process involves mixing base oil with various additives (such as anti-wear agents, rust inhibitors, antioxidants, etc.) in a specific order and stirring evenly under certain temperature conditions.

[0004] However, the traditional fixed setting of the stirring duration cannot adapt to the fluctuations in the physical properties of raw materials and the interference of environmental variables. Operators can only make empirical state judgments through the observation window, resulting in frequent local agglomeration of functional components or premature reaction inactivation during the mixing process. Especially in the final mixing stage of adding rust inhibitors and metal detergents, the rheological properties of the mixture will show non-linear changes with the temperature field distribution. Mechanical stirring controlled by a traditional timer is prone to cause two typical problems - when the additives do not reach the critical dispersion degree and the stirring is terminated prematurely, it will lead to boundary lubrication failure of the lubricating oil under high-pressure shear; while over-stirring will cause unexpected cross-linking of amine antioxidants and phosphorus-containing components, reducing the high-temperature oxidation stability.

[0005] Therefore, an optimized preparation scheme for high-temperature and high-pressure resistant lubricating oil is desired. Summary of the Invention

[0006] The present application aims at the deficiencies in the prior art and provides a high-temperature and high-pressure resistant lubricating oil and a preparation method thereof.

[0007] According to one aspect of the present application, there is provided a preparation method of a high-temperature and high-pressure resistant lubricating oil, which includes: S1: adding a predetermined proportion of α-olefin synthetic oil into a blending kettle equipped with a stirring device; S2: sequentially adding a predetermined proportion of antioxidant preservatives, extreme pressure phosphorus-free ashless anti-wear agents, and amine high-temperature antioxidants into the blending kettle, and performing stirring treatment under a first constant temperature condition to obtain a mixture; S3: sequentially adding a predetermined proportion of rust inhibitors, metal detergents, and pour point depressants into the mixture, and performing stirring treatment under a second constant temperature condition to obtain a high-temperature and high-pressure resistant lubricating oil; wherein, the step S3 includes:

[0008] Obtain the monitoring video of the mixture stirring state collected by the camera;

[0009] Perform key frame sampling on the monitoring video of the mixture stirring state to obtain a time series of key frames of the mixture stirring state;

[0010] Perform timestamp annotation on the stirring state features of each key frame of the mixture stirring state in the time series of key frames of the mixture stirring state to obtain a time series of stirring state features with time information annotation;

[0011] Perform encoding and decoding on the time series of stirring state features with time information annotation for the stirring state time series context to obtain an estimated value of the remaining stirring time;

[0012] Display the estimated value of the remaining stirring time.

[0013] According to another aspect of the present application, a high-temperature and high-pressure resistant lubricating oil is provided, wherein the high-temperature and high-pressure resistant lubricating oil is prepared by the preparation method of the high-temperature and high-pressure resistant lubricating oil as described above.

[0014] Due to the adoption of the above technical solutions, the present application has significant technical effects:

[0015] The high-temperature and high-pressure resistant lubricating oil and its preparation method provided by the present application first add a predetermined proportion of α-olefin synthetic oil into a blending kettle equipped with a stirring device, and then sequentially add an antioxidant preservative, an extreme pressure phosphorus-containing ashless anti-wear agent, and an amine high-temperature antioxidant into the blending kettle, and stir evenly under the first constant temperature condition to obtain a preliminary mixture. Subsequently, a rust inhibitor, a metal cleaner, and a pour point depressant are sequentially added to the mixture, and stirring is continued under the second constant temperature condition. During the stirring process, the stirring time is intelligently controlled by deeply analyzing the stirring state of the mixture, and finally a high-temperature and high-pressure resistant lubricating oil is prepared. In this way, the quality and performance of the high-temperature and high-pressure resistant lubricating oil product can be effectively ensured. Description of the Drawings

[0016] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 It is a flowchart of the preparation method of the high-temperature and high-pressure resistant lubricating oil according to the embodiment of the present application.

[0018] Figure 2It is a flowchart of step S3 in the preparation method of the high-temperature and high-pressure resistant lubricating oil according to the embodiment of the present application.

[0019] Figure 3 It is a flowchart of step S33 in the preparation method of the high-temperature and high-pressure resistant lubricating oil according to the embodiment of the present application.

[0020] Figure 4 It is a flowchart of step S34 in the preparation method of the high-temperature and high-pressure resistant lubricating oil according to the embodiment of the present application.

[0021] Figure 5 It is a flowchart of step S341 in the preparation method of the high-temperature and high-pressure resistant lubricating oil according to the embodiment of the present application. Detailed implementation manners

[0022] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0023] With the development of modern industrial equipment towards high speed and heavy load, the lubrication protection of mechanical equipment under extreme working conditions faces severe challenges. For example, in scenarios such as aerospace power systems, deep-sea exploration equipment, and metallurgical rolling production lines, the lubricating oil needs to maintain effective lubrication at high temperatures above 200 °C and contact pressures of several GPa. The high-temperature and high-pressure resistant lubricating oil is a special lubricating oil designed specifically for such extreme conditions. Its preparation requires mixing the base oil and various additives in a specific order and stirring evenly at a certain temperature.

[0024] However, the traditional fixed stirring duration is difficult to cope with the fluctuations in raw material characteristics and the interference of environmental variables. Operators usually rely on experience to judge the mixing state, which easily leads to local agglomeration or premature inactivation of functional components. Especially in the final mixing stage of the rust inhibitor and the metal cleaner, the rheological properties of the mixture change non-linearly due to the temperature distribution. The traditional timed stirring method often causes two types of problems: one is insufficient stirring, resulting in the additive not reaching the critical dispersion degree, thus causing boundary lubrication failure under high-pressure shear; the other is over-stirring, which triggers the unexpected cross-linking of amine antioxidants and phosphorus-containing components, reducing the high-temperature oxidation stability of the lubricating oil.

[0025] Based on this, the present application proposes a preparation method of a high-temperature and high-pressure resistant lubricating oil according to an embodiment of the present application. Figure 1 It is a flowchart of the preparation method of the high-temperature and high-pressure resistant lubricating oil according to the embodiment of the present application. As Figure 1As shown in the figure, the preparation method of the high-temperature and high-pressure resistant lubricating oil according to the embodiment of the present application includes: S1: adding a-olefin synthetic oil in a predetermined proportion into a blending kettle equipped with a stirring device; S2: sequentially adding an antioxidant preservative, an extreme pressure phosphorus-containing ashless anti-wear agent, and an amine high-temperature antioxidant into the blending kettle, and performing stirring treatment under the first constant temperature condition to obtain a mixture; S3: sequentially adding a rust inhibitor, a metal detergent, and a pour point depressant into the mixture, and performing stirring treatment under the second constant temperature condition to obtain the high-temperature and high-pressure resistant lubricating oil.

[0026] In step S1, a-olefin synthetic oil in a predetermined proportion is added into a blending kettle equipped with a stirring device. It should be understood that a-olefin synthetic oil has excellent viscosity-temperature performance, low-temperature fluidity, oxidation stability and other characteristics, and is an ideal base oil for preparing high-temperature and high-pressure resistant lubricating oil. Adding it into the blending kettle can lay a foundation for subsequent mixing with various additives to form a lubricating oil with specific properties. Specifically, the a-olefin synthetic oil forms a uniform liquid phase system in the blending kettle, providing a stable environment for the subsequent addition of additives. When the additives are added, they can be gradually dispersed in this base oil system, avoiding uneven mixing caused by uneven distribution of the base oil, and helping to improve the consistency and stability of the final lubricating oil product.

[0027] In step S2, an antioxidant preservative, an extreme pressure phosphorus-containing ashless anti-wear agent, and an amine high-temperature antioxidant in a predetermined proportion are sequentially added into the blending kettle, and stirring treatment is performed under the first constant temperature condition to obtain a mixture. It should be understood that the antioxidant preservative can prevent the lubricating oil from oxidizing and corroding during use and extend its service life; the extreme pressure phosphorus-containing ashless anti-wear agent can form a tough protective film on the metal surface under extreme working conditions such as high pressure and high temperature, effectively reducing friction and wear; the amine high-temperature antioxidant can inhibit the oxidation reaction of the lubricating oil in a high-temperature environment and maintain its good performance. Adding these additives sequentially and stirring under the first constant temperature condition can make them fully mixed and interact with the a-olefin synthetic oil, enabling the lubricating oil to have multiple properties such as high-temperature resistance, high-pressure resistance, anti-wear, and antioxidant resistance to meet the use requirements under extreme working conditions. In particular, the first constant temperature condition here specifically refers to 50-60 °C. Stirring under the constant temperature condition of 50-60 °C helps to reduce the viscosity of the additives, and thus makes them more easily dispersed and flow in the a-olefin synthetic oil. This temperature range can not only ensure good solubility and activity of the additives, but also avoid the failure of the additives or unnecessary side reactions caused by too high temperature. In short, through stirring, the antioxidant preservative, the extreme pressure phosphorus-containing ashless anti-wear agent, and the amine high-temperature antioxidant can be evenly distributed in the base oil to form a stable mixture, ensuring that the lubricating oil can perform consistent properties in different parts and under different use conditions.

[0028] In step S3, a rust inhibitor, a metal cleaner, and a pour point depressant are sequentially added to the mixture, and stirring treatment is performed under a second constant temperature condition to obtain a high-temperature and high-pressure resistant lubricating oil. It should be understood that the rust inhibitor can form a protective film on the metal surface to prevent the metal from rusting due to contact with air, moisture, etc.; the metal cleaner can remove oil stains, impurities, etc. on the metal surface, enabling the lubricating oil to better play its role during equipment operation; the pour point depressant can lower the pour point of the lubricating oil, enabling it to maintain good fluidity in a low-temperature environment. Adding these additives and stirring at a specific temperature can further improve the performance of the lubricating oil, enabling it not only to effectively lubricate under high temperature and high pressure, but also to protect the equipment in different working environments and ensure the normal operation of the equipment. In particular, the second constant temperature condition here specifically refers to 70 - 80 °C. The constant temperature condition of 70 - 80 °C is conducive to the full reaction of the rust inhibitor, the metal cleaner, and the pour point depressant with the previously formed mixture, and at the same time enables them to be evenly distributed in the mixture. An appropriate temperature can increase the activity of the additives, accelerate the reaction rate, and enable the additives to better play their roles. Stirring helps to break up the agglomeration of the additives and promote their diffusion in the mixture, ensuring the uniform and stable performance of the finally obtained lubricating oil.

[0029] Generally speaking, due to the good thermal stability and oxidation stability of the a-olefin synthetic oil used, it can maintain a better molecular structure and physical properties in a high-temperature environment, is not prone to deterioration and decomposition, and can provide a stable lubrication basis for the lubricating oil at high temperature. At the same time, its molecular structure characteristics also endow the lubricating oil with a certain compressive capacity, which helps to maintain the integrity of the oil film under high-pressure conditions. Antioxidants such as amine high-temperature antioxidants prevent the oxidation reaction from proceeding by capturing free radicals, etc., thereby maintaining the stable performance of the lubricating oil and enabling it to be used for a long time without failure at high temperature, indirectly improving the high-temperature resistance of the lubricating oil. The extreme-pressure phosphorus-containing ashless anti-wear agent can form a tough chemical reaction film on the metal surface under high-pressure contact conditions. This film can effectively reduce the friction coefficient between metals, prevent direct contact between the metal surfaces, thereby withstanding a contact pressure of several GPa, avoiding problems such as wear and seizure, and ensuring the lubrication effect of the lubricating oil under high-pressure working conditions. In addition, during the preparation process, stirring treatment is performed under the first constant temperature condition and the second constant temperature condition respectively. Appropriate temperature control helps the additives to be fully mixed and evenly distributed with the base oil, enabling each component to better play a synergistic role. At the same time, precise temperature control also avoids problems such as a decrease in the performance of the additives or incomplete reaction caused by too high or too low temperature, ensuring the final high-temperature and high-pressure resistant performance of the lubricating oil.

[0030] Based on the technical problems in the above background art, the technical concept of this application is to construct a dynamic adaptive stirring process control system based on the collaborative mechanism of visual perception and time series modeling. Specifically, by deploying a high-frame-rate industrial camera at the final mixing stage of the rust inhibitor and metal cleaner, the microscopic rheological morphological evolution characteristics of the mixture are captured in real time. Combining with deep learning algorithms, spatio-temporal feature extraction and timestamp coding fusion are performed on the key frames of the stirring state, and a non-linear mapping model between the mixing uniformity and the stirring time is established. Then, by introducing context coding with dynamic constraints, the model can separate environmental noise and effective state features, and capture the temporal evolution law of the mixing dynamics to dynamically predict the remaining stirring time and guide the process termination time point in real time. This solution effectively overcomes the problems of functional component agglomeration or thermal decomposition caused by fixed stirring duration in traditional processes, maintains the molecular structure integrity of amine antioxidants, and ensures the uniform distribution of rust inhibitor nanoparticles in the oil phase, thus significantly improving the anti-micro-pitting ability and high-temperature oxidation stability of lubricating oil under extreme working conditions.

[0031] Specifically, Figure 2 is a flowchart of step S3 in the preparation method of the high-temperature and high-pressure resistant lubricating oil according to an embodiment of the present application. As Figure 2 shown, the step S3 includes: S31, obtaining a monitoring video of the mixture stirring state collected by a camera; S32, performing key frame sampling on the monitoring video of the mixture stirring state to obtain a time series of key frames of the mixture stirring state; S33, performing timestamp annotation on the stirring state features of each key frame of the mixture stirring state in the time series of key frames of the mixture stirring state to obtain a time series of mixture stirring state features with time information annotation; S34, performing encoding and decoding on the time series of mixture stirring state features with time information annotation for the stirring state time series context to obtain an estimated value of the remaining stirring time; S35, displaying the estimated value of the remaining stirring time.

[0032] In step S31, a monitoring video of the mixture stirring state collected by a camera is obtained. It should be understood that the collected monitoring video of the mixture stirring state contains the dynamic picture information of the mixture during the stirring process. Specifically, it includes the flow pattern information of the mixture, such as whether it is laminar flow or turbulent flow, and the change information of the flow velocity and direction; the appearance feature information of the mixture, such as the uniformity of color, whether there are local agglomeration phenomena or precipitation; the distribution information of different components in the mixture can also be observed, as well as the change process of these states over time. Generally speaking, by deeply analyzing the obtained monitoring video of the mixture stirring state, the stirring state of the mixture during the stirring process can be understood, and the stirring time can be accurately controlled according to the actual stirring state, which can ensure the precise matching of the additive dispersion and the reaction process, and thus avoid the quality problems of lubricating oil caused by insufficient or excessive stirring.

[0033] In step S32, key-frame sampling is performed on the monitoring video of the stirring state of the mixture to obtain a time series of key frames of the stirring state of the mixture. Accordingly, considering that the mixing uniformity in the stirring stage directly affects the realization of the functions of the additives. Traditional processes rely on fixed stirring durations or manual observation to judge the mixing state. However, due to the complex rheological characteristics of the fluid in the stirring kettle and the dynamics of the reaction of the additives, continuous video monitoring will generate a large amount of redundant data, and there is often a high degree of similarity between adjacent frames, resulting in both consuming computing resources and being difficult to capture key state transition points when directly processing the full amount of video. In addition, the phase change process of the mixture (such as the dispersion of rust inhibitor particles and the association of amine antioxidant molecules) has non-uniform time scale characteristics. Random sampling may miss the instantaneous state changes at the critical dispersion degree, while uniform sampling will weaken the representation weight of key time nodes. Based on this, the present application performs key-frame sampling on the monitoring video of the stirring state of the mixture to obtain a time series of key frames of the stirring state of the mixture. That is, through the key-frame sampling technology, representative frames that can characterize the mutation points of the mixing process are extracted from the continuous monitoring video stream to form a sequence of state snapshots with temporal correlation. This step designs an adaptive sampling strategy based on the phased characteristics of the mixing kinetics, preferentially capturing the video frames corresponding to landmark events such as the mutation of the stirring vortex morphology, the clarification of the phase interface, or the dissipation of particle agglomeration, and at the same time recording the occurrence time sequence of each key event through time stamps. This processing not only compresses the scale of the original data, but also constructs a time evolution map that reflects the essential laws of the dispersion and reaction processes of the additives by focusing on the key state transition points in the mixing process.

[0034] In step S33, time stamps are marked on the stirring state characteristics of each key frame of the stirring state of the mixture in the time series of key frames of the stirring state of the mixture to obtain a time series of stirring state characteristics with time information marked. Specifically, Figure 3 is a flowchart of step S33 in the preparation method of the high-temperature and high-pressure resistant lubricating oil according to an embodiment of the present application. As Figure 3As shown, step S33 includes: S331, respectively recording the timestamps of each key frame of the mixing state of the mixture in the time series of key frames of the mixing state of the mixture to obtain a time series of key frame timestamps; S332, extracting the mixing state features of each key frame of the mixing state of the mixture in the time series of key frames of the mixing state of the mixture to obtain a time series of mixing state feature vectors; S333, performing one-hot embedding encoding on each key frame timestamp in the time series of key frame timestamps to obtain a time series of one-hot embedding encoded vectors of key frame timestamps; S334, fusing the time series of the mixing state feature vectors and the time series of the one-hot embedding encoded vectors of key frame timestamps to obtain a time series of mixing state feature vectors annotated with time information as the time series of mixing state features annotated with time information.

[0035] In step S331, the timestamps of each key frame of the mixing state of the mixture in the time series of key frames of the mixing state of the mixture are respectively recorded to obtain a time series of key frame timestamps. Correspondingly, considering that the rheological properties of the mixture and the dispersion state of the additive have significant time-sequence dependence, but in the traditional process, manual observation or stirring control with a fixed duration cannot quantify the dynamic evolution law of the mixing kinetics. Since the dispersion process of the rust inhibitor and the metal cleaner involves non-steady-state phase changes (such as the hindered Brownian motion of nanoparticles, the instantaneous gradient change of the oil phase viscosity, etc.), it is difficult to distinguish the essential differences in the mixing states at different time nodes solely relying on the spatial features of the visual key frames. For example, under the same eddy current morphology, the particle aggregation in the early stage and the transient aggregation in the near-uniform dispersion stage may present similar apparent features. Without the annotation of time dimension information, it will lead to the subsequent model confusing the time-sequence correlation between physical mixing and chemical reactions. Therefore, in the technical solution of this application, the timestamps of each key frame of the mixing state of the mixture in the time series of key frames of the mixing state of the mixture are respectively recorded to obtain a time series of key frame timestamps. In this way, by recording the timestamps corresponding to each key frame and constructing a time series, the physical and chemical changes in the mixing process can be anchored on the continuous time axis. That is, this step accurately records the occurrence time sequence of key events such as the dispersion of the rust inhibitor and the association of amine antioxidants while extracting the spatial features of the mixing state, providing a spatio-temporal coupling data basis for the subsequent model to establish the mapping relationship between state features and time evolution. The serialized encoding of the timestamps not only marks the absolute time progress of the mixing process but also implies the relative time interval features between the state transition nodes.

[0036] In step S332, the stirring state features of each key frame of the mixture stirring state in the time series of the key frames of the mixture stirring state are extracted to obtain a time series of mixture stirring state feature vectors. Specifically, in the embodiment of the present application, step S332 includes: using a feature extractor based on the atrous feature pyramid network to extract the stirring state features of each key frame of the mixture stirring state in the time series of the key frames of the mixture stirring state to obtain the time series of the mixture stirring state feature vectors. Correspondingly, considering that the micro-rheological properties of the mixture (such as the dispersion uniformity of rust inhibitor particles and the molecular association state of amine antioxidants) directly determine the extreme working condition performance of the lubricating oil, but in the traditional process, manual observation or simple timing cannot quantify the dynamic evolution of these micro-states. Since the optical properties of the mixture in the stirring kettle are comprehensively affected by the temperature gradient, additive concentration distribution, and phase interface tension, the pixel-level data in the original monitoring video contains a large amount of environmental noise (such as light fluctuations, bubble interference) that has nothing to do with the mixing uniformity. If the original image data is directly used for time series modeling, it will be difficult for the model to distinguish the surface flow pattern changes from the substantial molecular-level dispersion process, thereby resulting in a prediction deviation of the remaining time. Based on this, in the present application, the stirring state features of each key frame of the mixture stirring state in the time series of the key frames of the mixture stirring state are extracted to obtain a time series of mixture stirring state feature vectors. In particular, in a specific example of the present application, a feature extractor based on the atrous feature pyramid network is used to extract the stirring state features of each key frame of the mixture stirring state in the time series of the key frames of the mixture stirring state to obtain the time series of the mixture stirring state feature vectors. Specifically, by introducing the atrous feature pyramid network to construct a multi-scale feature extractor, and using the dilation rate adjustment characteristic of atrous convolution, the feature capture of multiple receptive fields is realized in a single-layer network. This network synchronously extracts the local fine features (such as the gradient change of the particle distribution texture) and global rheological features (such as the topological deformation of the eddy current structure) of the key frame image through a parallel branch structure, and then through the cross-layer fusion mechanism of the feature pyramid, the stirring state representations of different scales are adaptively weighted and fused. The sparse sampling strategy of atrous convolution effectively expands the model's perception ability of the long-range spatial correlation features in the stirring kettle without increasing the number of parameters, for example, identifying the kinetic correlation between the diffusion path of rust inhibitor nanoparticles and the rotation direction of the macroscopic eddy current.

[0037] The following is a detailed elaboration of a specific implementation process of "using a feature extractor based on the atrous feature pyramid network to extract the stirring state features of each key frame of the mixture stirring state in the time series of the key frames of the mixture stirring state to obtain the time series of the mixture stirring state feature vectors":

[0038] First is the data preprocessing stage. After sampling key frames from the mixture stirring state monitoring video, the obtained key frame images need to undergo a series of processes to meet the requirements of subsequent feature extraction. Unifying the size is the primary task. Adjust all key frame images to a fixed specification, such as 224×224 pixels, which ensures the consistency of the images input to the feature extractor in the spatial dimension and provides a stable foundation for subsequent processing. Normalization is equally important. Normalizing the image pixel values to the range [0,1] or [-1,1] can effectively accelerate the convergence speed during model training and enable the model to learn image features faster. In some cases, to enhance the generalization ability of the model, data augmentation operations can also be selected. By using means such as rotation, flipping, and scaling to expand the diversity of training data, the model can accurately extract features when facing stirring state images of different angles and sizes.

[0039] Next, construct the Atrous Spatial Pyramid Pooling (ASPP) network. First, select a pre-trained convolutional neural network, such as the widely used ResNet, VGG, etc. These networks have powerful basic feature extraction capabilities after being trained on large-scale image datasets. Taking ResNet-50 as an example, remove its last fully connected layer and only retain the convolutional layer part to provide the basic architecture for subsequent construction of the ASPP module. On this basis, add the ASPP module, which consists of multiple atrous convolutional layers with different dilation rates and a global average pooling layer. Set different dilation rates such as 6, 12, 18, etc. Each atrous convolutional layer convolves the input feature map using its respective dilation rate to capture context information at different scales. Small dilation rates focus on local details, while large dilation rates focus on global features. The global average pooling layer then converts the input feature map into a global feature vector, and adjusts the number of channels through a 1×1 convolutional layer to make it consistent with the output channels of the atrous convolutional layers. Finally, concatenate the outputs of the atrous convolutional layers with different dilation rates and the output of the global average pooling layer in the channel dimension, and reduce the dimension through a 1×1 convolutional layer to form the final feature representation, which integrates multi-scale information and more comprehensively reflects the stirring state.

[0040] After completing the network construction, enter the feature extraction stage. Load the weights of the pre-trained basic convolutional network and initialize the parameters of the ASPP module. During training, according to the actual situation, some parameters of the basic convolutional network can be selected to be frozen, and only the ASPP module is trained, which can accelerate the training process. Input the preprocessed key frame images into the constructed network one by one. The images first go through the basic convolutional network to extract the underlying image features, and then these features are input into the ASPP module. The atrous convolutional layers with different dilation rates and the global average pooling layer in the module work together to fuse the features, and finally output the stirring state feature vectors containing context information at different scales, which accurately depict the stirring state of each key frame.

[0041] Finally, a time series of stirring state feature vectors is generated. The stirring state feature vectors corresponding to each key frame are collected in chronological order to form an ordered sequence. This sequence records the changes in the stirring state features of the mixture at different times during the stirring process, providing a data basis for subsequent analysis of the temporal evolution law of the stirring state.

[0042] In step S333, one-hot embedding encoding is performed on each key frame timestamp in the time series of the key frame timestamps to obtain a time series of key frame timestamp one-hot embedding encoding vectors. It should be understood that the key frame timestamp is essentially a discrete time point record. In time series analysis, the original timestamp data is usually numerical and difficult to be directly processed by some deep learning models. And in order to be able to convert discrete data into a vector representation, a common method is to convert each timestamp into a unique vector form so that the timestamp information can be effectively processed and learned by the model. In this application, one-hot embedding encoding is performed on each key frame timestamp in the time series of the key frame timestamps to obtain a time series of key frame timestamp one-hot embedding encoding vectors. That is to say, one-hot embedding encoding can clearly retain this uniqueness information. By mapping each timestamp to a vector with only one element being 1 and the rest being 0, the difference between different timestamps becomes more obvious, facilitating the subsequent model to accurately distinguish different time points during processing.

[0043] In step S334, fuse the time series of the mixture stirring state feature vectors and the time series of the one-hot embedding encoding vectors of the key frame timestamps to obtain the time series of the mixture stirring state feature vectors annotated with time information as the time series of the mixture stirring state features annotated with time information. Specifically, in the embodiment of the present application, step S334 includes: concatenating each corresponding mixture stirring state feature vector and key frame timestamp one-hot embedding encoding vector in the time series of the mixture stirring state feature vectors and the time series of the key frame timestamp one-hot embedding encoding vectors to obtain the time series of the mixture stirring state feature vectors annotated with time information. Correspondingly, considering that there is a complex non-linear coupling relationship between the state evolution of the mixture and the time dimension, but in traditional processes, visual observation and time control are often treated separately, resulting in difficulty in establishing a correlation model between the change rate of state features and the time process. Since the dispersion process of rust inhibitor particles involves a transition from rapid diffusion dominated by Brownian motion to slow suspension after surfactant adsorption, the time series relying solely on visual features may confuse the similar apparent states in different kinetic stages (such as the vortex morphology in the mid-term diffusion acceleration stage and the late stable stage is similar), and the isolated timestamp encoding cannot characterize the microscopic state characteristics corresponding to specific time nodes. This separation of spatio-temporal information will weaken the model's ability to judge the critical point of the mixing process. Therefore, in the present application, by fusing the time series of the mixture stirring state feature vectors and the time series of the key frame timestamp one-hot embedding encoding vectors, the time series of the mixture stirring state feature vectors annotated with time information is obtained as the time series of the mixture stirring state features annotated with time information. By fusing these two time series, a more accurate non-linear mapping relationship between the mixing uniformity and the stirring time can be established, so that the change in the rheological properties of the mixture (such as the fluctuation of the phase interface blur degree) can form a joint representation with the stage attributes of the time node (such as the rapid diffusion stage or the stable suspension stage). The obtained time series of the mixture stirring state feature vectors annotated with time information contains the stirring state and the corresponding time point at each moment, which helps the model learn how the change in the stirring state at different times affects the final mixing uniformity, thereby improving the accuracy of predicting the remaining stirring time.

[0044] In step S34, perform stirring state time series context sequence encoding and decoding on the time series of the mixture stirring state features annotated with time information to obtain an estimated value of the remaining stirring time. Specifically, Figure 4 FIG. is a flowchart of step S34 in the preparation method of the high-temperature and high-pressure resistant lubricating oil according to the embodiment of the present application. As Figure 4As shown, step S34 includes: S341, performing dynamic-constraint-based encoding on the time series of the time information-annotated mixture stirring state feature vectors to obtain a stirring state time series inference encoding vector; S342, performing feature decoding on the stirring state time series inference encoding vector to obtain an estimated value of the remaining stirring time.

[0045] In step S341, dynamic-constraint-based encoding is performed on the time series of the time information-annotated mixture stirring state feature vectors to obtain a stirring state time series inference encoding vector. Specifically, Figure 5 FIG. is a flowchart of step S341 in the preparation method of the high-temperature and high-pressure resistant lubricating oil according to an embodiment of the present application. As Figure 5 shown, step S341 includes: S341-1, calculating the time information-annotated mixture stirring state end information flow constraint parameter and the time information-annotated mixture stirring state axial information flow constraint parameter of each time information-annotated mixture stirring state feature vector in the time series of the time information-annotated mixture stirring state feature vectors; S341-2, calculating the time information-annotated mixture stirring state context propagation constraint parameter of each time information-annotated mixture stirring state feature vector based on the time information-annotated mixture stirring state end information flow constraint parameter and the time information-annotated mixture stirring state axial information flow constraint parameter of each time information-annotated mixture stirring state feature vector in the time series of the time information-annotated mixture stirring state feature vectors; S341-3, performing time series context dynamic constraint propagation inference on the sequence distribution of the time information-annotated mixture stirring state feature vectors based on the time information-annotated mixture stirring state context propagation constraint parameter of each time information-annotated mixture stirring state feature vector to obtain the stirring state time series inference encoding vector.

[0046] It should be understood that the state evolution of the mixture has complex spatio-temporal coupling characteristics, and traditional time-series modeling methods are difficult to effectively capture the dynamic non-linear laws of additive dispersion and reaction. Since the dispersion process of rust inhibitor nanoparticles involves a phase transition from rapid diffusion to surfactant adsorption, and the molecular association of amine antioxidants exhibits exponential decay reaction kinetics characteristics, simple sequence modeling (such as RNN or LSTM) is vulnerable to long-range dependence decay or local noise interference, resulting in misjudgment of critical phase transition nodes. For example, in the middle and late stages of mixing, the sudden drop in the additive dispersion rate may be misinterpreted as reaching a uniform state, while in fact it may be in a metastable state where the surfactant coating is not completed. At this time, if the stirring is terminated, it will lead to secondary aggregation of particles under subsequent high-pressure conditions. Therefore, in this application, a time-series of the mixture stirring state feature vectors annotated with the time information is encoded with a stirring state time-series context sequence based on dynamic constraints to obtain a stirring state time-series inference encoded vector.

[0047] Specifically, first, the end vector of the time-series of the mixture stirring state feature vectors annotated with the time information is used as the end constraint anchor encoding vector to represent the instantaneous state of the current mixing process, and axial constraint anchor encoding is extracted through cluster analysis to capture the global mixing pattern. Subsequently, the information flow constraint parameters of each feature vector with respect to the end and axial anchor encodings are calculated respectively to quantify the double influence weights of the local state and the global structure. By means of the transverse field-induced compactification method, the contradiction between local divergence suppression and global constraint maintenance is coordinated, and the topological consistency optimization of the constraint parameters is realized by using reversible metric rotation transformation. Finally, context dynamic propagation fuses the double constraint factors through weighting, enabling the model to focus on the microscopic details of the current mixing state and follow the macroscopic evolution law of the overall process during the information transmission process, thereby helping the model to more accurately infer the current and future stirring states and understand the progress of the stirring process.

[0048] Specifically, in the embodiment of this application, the step S341-1 includes: extracting the last time information-annotated mixture stirring state feature vector from the time-series of the mixture stirring state feature vectors annotated with the time information as the time information-annotated mixture stirring state propagation space end constraint anchor encoding vector, and this process can be expressed by the formula:

[0049] O={v1,v2,...,v i ,...,v t}

[0050] v tail =v t

[0051] where O is the time-series of the mixture stirring state feature vectors annotated with the time information, and v1, v2, v i and v tThey are the 1st, 2nd, ith, and tth time information - annotated mixture stirring state feature vectors in the time series of the mixture stirring state feature vectors respectively, v tail is the end - constraint anchoring encoding vector of the time information - annotated mixture stirring state propagation space;

[0052] Performing clustering analysis on the time series of the time information - annotated mixture stirring state feature vectors to obtain the axial - constraint anchoring encoding vector of the time information - annotated mixture stirring state propagation space. This process can be expressed by the formula:

[0053]

[0054] where Cluster is the clustering analysis operation, max(v i ) and min(v i ) are respectively taking the maximum and minimum values of v i , η is the adjustment parameter, a i is the ith time information - annotated mixture stirring state constraint anchoring value in the time series of the time information - annotated mixture stirring state constraint anchoring values, softmax is the normalization function, e i is the ith time information - annotated mixture stirring state constraint anchoring weight value in the time series of the time information - annotated mixture stirring state constraint anchoring weight values, t is the number of vectors in O, v axis is the axial - constraint anchoring encoding vector of the time information - annotated mixture stirring state propagation space;

[0055] Calculating the time information - annotated mixture stirring state end - information flow constraint parameter of each time information - annotated mixture stirring state feature vector in the time series of the time information - annotated mixture stirring state feature vectors relative to the end - constraint anchoring encoding vector of the time information - annotated mixture stirring state propagation space. This process can be expressed by the formula:

[0056]

[0057] where f(v i , v tail ) is to calculate the end - information flow constraint parameter between v i and v tail , is the eigenvalue at the jth position of v i , is the eigenvalue at the jth position of v tail , log2 is the logarithmic function value with base 2, n is the number of eigenvalues in v i , exp is the exponential function value with base e (natural constant), α i is v iCorresponding time information annotates the end - information - flow constraint parameter of the mixture stirring state;

[0058] Calculate the time - information - annotated mixture stirring state axial information - flow constraint parameter of each time - information - annotated mixture stirring state feature vector in the time series of the time - information - annotated mixture stirring state feature vector relative to the time - information - annotated mixture stirring state propagation space axial constraint anchoring coding vector. This process can be expressed by the formula:

[0059]

[0060] where, f(v i , v axis ) is the axial information - flow constraint parameter for calculating between v i and v axis , ||·|| 2 is the square of the Euclidean norm for calculating the vector, arccosh is the inverse hyperbolic cosine function, and β i is the time - information - annotated mixture stirring state axial information - flow constraint parameter corresponding to v i .

[0061] It should be understood that in the dynamic mixing process, the feature vector at the end of the time series represents the instantaneous state at the current moment (such as the adsorption saturation degree of rust inhibitor particles or the association rate of antioxidants). As a boundary condition, it is crucial for the convergence of information transmission. Traditional time - series models are prone to ignoring the correlation between the current state and the historical state due to the decay of long - range dependence. For example, a sudden drop in the dispersion rate in the middle and late stages of mixing may be misjudged as a steady state. By taking the last time - information - annotated mixture stirring state feature vector in the time series of the time - information - annotated mixture stirring state feature vector as the time - information - annotated mixture stirring state propagation space end - constraint anchoring coding vector, it is equivalent to introducing an "instantaneous observation point" of a dynamic system, which can constrain the information transmission direction (such as avoiding over - reliance on noise data in the early diffusion stage), and at the same time enhance the model's sensitivity to phase - change transitions (such as the critical point of surfactant adsorption completion) through time - series information. In this way, it not only suppresses the interference of local noise on the historical state, but also provides micro - state calibration for the model through the physical meaning of the end state (such as the current particle dispersion degree), avoiding misjudging the metastable state as the termination condition.

[0062] Accordingly, considering that the global pattern of the mixing process (such as the three stages of "rapid diffusion - adsorption phase transition - association attenuation") has a non - linear topological structure, simple time averaging will blur the phase transition characteristics. By clustering analysis, time information is extracted to label the axial constraint anchoring encoding vector of the propagation space of the mixture stirring state. Essentially, it is to perform manifold learning on the high - dimensional feature space to capture the core patterns of different mixing stages (such as the cluster center may correspond to the characteristic mutation point when the surfactant adsorption starts). For example, in the amine antioxidant system, clustering can separate the exponential decay kinetic characteristics from the linear diffusion characteristics. This axial constraint is equivalent to constructing an implicit mixing state phase diagram, enabling information transfer to follow the direction of the data principal components (such as transferring along the reaction rate gradient direction), which can prevent the model from deviating from the global evolution trajectory due to local stirring parameter fluctuations (such as temperature noise), thereby enhancing the prediction robustness against the risk of secondary agglomeration.

[0063] It should be understood that the end - information - flow constraint parameter of the time - information - labeled mixture stirring state quantifies the influence weight of the historical state on the current decision by measuring the similarity between the feature vectors at each moment and the current state. For example, in the surfactant coating stage, the high - similarity features during early rapid diffusion are assigned low weights (because they are irrelevant to the current metastable state), while the features near the phase transition point are strengthened. This dynamic weighting mechanism solves the defect of "uniform forgetting" in LSTM: the traditional gating mechanism is difficult to distinguish historical states with different physical meanings (such as misidentifying uncoated particles as coated). By introducing a constraint parameter in the form of conditional probability, the model can focus on the historical segments causally related to the current state (such as the change in adsorption rate in the last 10 seconds), thus accurately identifying special patterns such as "pseudo - steady - state dispersion rate".

[0064] Accordingly, in order to evaluate the degree of fit between each time - information - labeled mixture stirring state feature vector and the global pattern skeleton (such as the distance projected onto the principal component axis), and to force information transfer to follow the inherent kinetic law of the mixing process, it is necessary to calculate the axial information - flow constraint parameter of the time - information - labeled mixture stirring state in this application. For example, in the antioxidant association stage, the features conforming to the exponential decay law are enhanced, while the features deviating from this pattern (such as abnormal fluctuations caused by agitator failure) are suppressed. This is equivalent to adding a regularization term based on physical prior in the loss function, and its effects are reflected in two aspects: one is to prevent the model from overfitting local noise (such as instantaneous turbulent perturbations), and the other is to enhance the inference ability for unobserved states (such as predicting the possible secondary agglomeration path under high - pressure conditions) by maintaining consistency with the global structure. In particular, in the critical region of phase transition, this parameter can identify abnormal features deviating from the main pattern as phase - transition warning signals.

[0065] Specifically, in the embodiments of the present application, the step S341-2 includes: using a standard space constraint matrix to perform transverse field-induced compactification on the time information annotation mixture stirring state end information flow constraint parameter and the time information annotation mixture stirring state axial information flow constraint parameter corresponding to the time information annotation mixture stirring state feature vector to obtain the time information annotation mixture stirring state end information flow compact constraint parameter and the time information annotation mixture stirring state axial information flow compact constraint parameter; based on the time information annotation mixture stirring state end information flow compact constraint parameter and the time information annotation mixture stirring state axial information flow compact constraint parameter, performing constraint condition consistency on the time information annotation mixture stirring state end information flow constraint parameter and the time information annotation mixture stirring state axial information flow constraint parameter to obtain the optimized time information annotation mixture stirring state end information flow constraint parameter and the optimized time information annotation mixture stirring state axial information flow constraint parameter; performing non-linear weighting processing on the optimized time information annotation mixture stirring state end information flow constraint parameter and the optimized time information annotation mixture stirring state axial information flow constraint parameter to obtain the time information annotation mixture stirring state context propagation constraint parameter corresponding to the time information annotation mixture stirring state feature vector. The above process can be expressed by the formula:

[0066]

[0067] y i = Sigmoid(ω1·α i '+ ω2·β i ')

[0068] where δ i and ε i are respectively the time information annotation mixture stirring state end information flow compact constraint parameter and the time information annotation mixture stirring state axial information flow compact constraint parameter after transverse field-induced compactification of α i and β i , T represents the transpose operation, |·| is the absolute value operation, α i ' and β i ' are respectively the optimized time information annotation mixture stirring state end information flow constraint parameter and the optimized time information annotation mixture stirring state axial information flow constraint parameter corresponding to v i , ω1 and ω2 are respectively weighting parameters, Sigmoid is the weight mapping function, and y i is the time information annotation mixture stirring state context propagation constraint parameter corresponding to v i .

[0069] In particular, in the optimization of the temporal modeling of the dynamic mixing process, the collaborative optimization of the end - information - flow constraint parameters and the axial - information - flow constraint parameters for the time - information - annotated mixing state of the mixture needs to overcome the contradiction between the divergence of the feature space and the global dynamic law. For the current instantaneous state (such as the adsorption saturation of rust inhibitor particles) anchored by the end - information - flow constraint parameters of the time - information - annotated mixing state of the mixture and the main mode of the mixing stage (such as the adsorption phase - change characteristics of surfactants) characterized by the axial - information - flow constraint parameters, the transverse - field distribution of their parameter space is prone to generalization imbalance due to the interaction between local noise interference (such as stirring - temperature fluctuations) and the global manifold structure (such as the "diffusion - phase - change - decay" three - stage topology). Therefore, a transverse - field - induced compactification method based on differential - geometric metric transformation is adopted. First, the two types of parameters are projected into the covariant space through a standard - space constraint matrix, and a compact mapping relationship is constructed using the inherent dynamic law of the mixing process (for example, encoding the exponential - decay law of the antioxidant association stage as a covariant tensor). Then, a rotation transformation of the invertible metric component is introduced to achieve the covariant normalization of the parameter distribution in the finite - dimensional spinor space. This process essentially unifies the end - state calibration (such as the microscopic characterization of particle dispersion) and the axial - mode recognition (such as the principal - component projection of phase - change warning) in the topological manifold of the parameter space through a gauge - transformation framework, enabling information transfer to not only suppress noise interference in the early diffusion stage but also maintain dynamic synchronization with the global evolution trajectory (such as the path of the risk of secondary agglomeration) in the critical region. This two - stage optimization strategy, while maintaining the sensitivity to local features (such as the instantaneous capture of a sudden drop in the dispersion rate), strengthens the coupling between the physical mechanism of the mixing process and the data - driven model through the metric reconstruction of the parameter space, ultimately achieving multi - scale coordination of the microscopic - state evolution and the macroscopic - phase - change law.

[0070] It should be understood that during the mixing process, there is a dynamic coupling relationship between local states (such as the instantaneous dispersion rate of rust inhibitor particles) and global patterns (such as the overall association kinetics of amine antioxidants). Relying solely on one of them by traditional methods will lead to modeling deviations. The time information annotation of the end information flow constraint parameter of the mixture stirring state focuses on the microscopic state at the current moment (such as the real-time coverage rate of surfactant adsorption), while the time information annotation of the axial information flow constraint parameter of the mixture stirring state encodes the macroscopic evolution law of the mixing process (such as the phase transition sequence of the "diffusion - adsorption - association" three stages). By fusing the two, the model can adaptively adjust the constraint weights according to the mixing stage: in the initial stage of rapid diffusion, the axial constraint dominates to ensure compliance with Fick's diffusion law; when approaching the adsorption phase transition critical point, the end constraint is enhanced to capture the nonlinear sudden drop in the dispersion rate of nanoparticles. This dynamic balance solves the defect of traditional sequence models that cannot distinguish the characteristics of different mixing stages due to fixed weight allocation. For example, it avoids misjudging a temporary dispersion stagnation caused by local turbulence as a steady state of adsorption completion, and at the same time prevents over - focusing on instantaneous fluctuations during the antioxidant association stage and ignoring the overall exponential decay trend. The context propagation constraint parameter of the time - information - annotated mixture stirring state after fusion essentially constructs a physical - information - guided attention mechanism, enabling the model to not only respond to the mutation signals of microscopic states during information transmission but also maintain synchronous tracking of macroscopic kinetic laws.

[0071] Specifically, in the embodiment of the present application, the step S341 - 3 includes: based on the context propagation constraint parameter of the time - information - annotated mixture stirring state of each time - information - annotated mixture stirring state feature vector, performing temporal context - dynamic - constraint propagation inference on the sequence distribution of the time - information - annotated mixture stirring state feature vector to obtain the stirring - state temporal inference encoding vector, which can be expressed by the following formula:

[0072]

[0073] where z is the stirring - state temporal inference encoding vector.

[0074] It should be understood that the spatio-temporal coupling characteristics of the mixed state require the encoding vector to simultaneously contain local transient details and global evolution paths. By injecting the context propagation constraint parameters of the time information annotation of the mixture stirring state into the feature propagation process, the model achieves the topological alignment of multi-scale features when reconstructing the encoding vector: the end constraint ensures that the physical quantities at the current moment (such as the adsorption energy on the particle surface) are accurately retained, while the axial constraint forces the encoding space to be homeomorphic to the backbone structure of the mixed phase diagram. For example, in the middle stage of nanoparticle dispersion, this encoding will highlight the spatial gradient distribution of the surfactant adsorption rate and simultaneously map it to the metastable region of "uncompleted coating" along the axial constraint direction; when approaching the phase transition point, the encoding vector will simultaneously reflect the mutation of the local adsorption entropy and the migration trend of the global phase diagram towards the "stable dispersion" state. This encoding mechanism enables the model to analyze implicit associations that are difficult to capture by traditional methods, such as identifying the causal relationship between the local under-adsorption of the surfactant (microscopic anomaly) caused by the temperature gradient and the overall mixing entropy not reaching the threshold (macroscopic constraint), so as to accurately judge whether the uniform state of terminating stirring is truly achieved, rather than remaining in the metastable state. That is, the finally generated temporal inference encoding vector of the stirring state is essentially a state tensor of the mixing process, and its multi-dimensional components respectively correspond to the evolution trajectories of key physical parameters at different spatio-temporal scales, which can provide a feature representation with both resolution and robustness for subsequent working condition prediction.

[0075] In step S342, feature decoding is performed on the temporal inference encoding vector of the stirring state to obtain an estimated value of the remaining stirring time. Specifically, in the embodiment of the present application, step S342 includes: inputting the temporal inference encoding vector of the stirring state into a stirring time analyzer based on a decoder to obtain an estimated value of the remaining stirring time. It should be understood that the temporal inference encoding vector of the stirring state is obtained after a series of complex processes, and it contains rich temporal and state information during the stirring process, but this information exists in a coded form. The stirring time analyzer based on the decoder can decode this coded information and convert it into a more intuitive and easier-to-understand estimated value of the remaining stirring time, thus realizing the conversion from complex coded features to the information required for practical applications. In particular, the stirring time analyzer based on the decoder is usually trained, and it has learned the relationship between the temporal inference encoding vector of the stirring state and the actual remaining stirring time. By inputting the encoding vector into the analyzer, it can utilize its existing learning results and knowledge to analyze and infer the new encoding vector, so as to obtain an accurate estimated value of the remaining stirring time. In this way, by estimating the remaining stirring time in real time, the system can dynamically adjust the stirring operation according to the actual situation, such as stopping stirring in a timely manner, avoiding insufficient or excessive stirring, ensuring the preparation quality of the lubricating oil, and at the same time enabling engineers to have a forward-looking prediction of the entire stirring progress, thus facilitating process control and optimization.

[0076] In step S35, an estimated value of the remaining stirring time is displayed. It should be understood that displaying the estimated value of the remaining stirring time helps the operator to more precisely control the stirring time. In traditional preparation methods, due to the use of a fixed stirring duration setting, problems such as insufficient stirring or over-stirring are likely to occur. By displaying the remaining stirring time in real time, the operator can flexibly adjust the stirring operation according to the actual situation, ensure that the stirring stops when the best mixing effect is achieved, and thus improve the preparation quality of the lubricating oil.

[0077] In summary, step S3 is clearly described. First, a high-frame-rate industrial camera is deployed at the final mixing stage of the rust inhibitor and the metal cleaner to capture the microscopic rheological morphological evolution characteristics of the mixture in real time. Combining with a deep learning algorithm, spatio-temporal feature extraction and timestamp coding fusion are performed on the key frames of the stirring state to establish a non-linear mapping model between the mixing uniformity and the stirring time. Then, by introducing context encoding with dynamic constraints, the model can separate environmental noise and effective state features, and capture the temporal evolution law of the mixing dynamics to dynamically predict the remaining stirring time and guide the process termination point in real time. In this way, the problems of functional component agglomeration or thermal decomposition caused by a fixed stirring duration in the traditional process are effectively overcome. It not only maintains the molecular structure integrity of the amine antioxidant, but also ensures the uniform distribution of the rust inhibitor nanoparticles in the oil phase, thereby significantly improving the anti-micro-pitting ability and high-temperature oxidation stability of the lubricating oil under extreme working conditions.

[0078] In summary, the preparation method of the high-temperature and high-pressure resistant lubricating oil based on the embodiments of the present application is clarified. First, a predetermined proportion of α-olefin synthetic oil is added to a blending kettle equipped with a stirring device. Then, an antioxidant preservative, an extreme pressure phosphorus-containing ashless anti-wear agent, and an amine high-temperature antioxidant are sequentially added to the blending kettle, and stirred evenly under the first constant temperature condition to obtain a preliminary mixture. Subsequently, a rust inhibitor, a metal cleaner, and a pour point depressant are sequentially added to the mixture, and stirring is continued under the second constant temperature condition. During the stirring process, the stirring time is intelligently controlled by deeply analyzing the stirring state of the mixture, and finally, a high-temperature and high-pressure resistant lubricating oil is prepared. In this way, the quality and performance of the high-temperature and high-pressure resistant lubricating oil product can be effectively ensured.

[0079] In other embodiments of the present application, a high-temperature and high-pressure resistant lubricating oil is also provided, wherein the high-temperature and high-pressure resistant lubricating oil is prepared by the preparation method of the high-temperature and high-pressure resistant lubricating oil as described above.

[0080] The basic principles of the present application have been described in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are only examples and not limitations. It cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present application. Additionally, the specific details of the above application are only for the purposes of illustration and facilitating understanding, rather than limitations. The above details do not limit the present application to necessarily adopt the above specific details for implementation.

[0081] Words such as "including", "comprising", "having", etc. in the present application are open-ended terms, meaning "including but not limited to", and can be used interchangeably with each other. The words "or" and "and" used herein refer to the term "and / or", and can be used interchangeably with it, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to", and can be used interchangeably with it.

[0082] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

[0083] The above description has been given for purposes of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A preparation method of a high-temperature and high-pressure resistant lubricating oil, comprising: S1: Add a-olefin synthetic oil in a predetermined proportion into a blending kettle equipped with a stirring device; S2: Sequentially add an antioxidant preservative, an extreme pressure phosphorus-containing ashless anti-wear agent, and an amine high-temperature antioxidant in a predetermined proportion into the blending kettle, and perform stirring treatment under a first constant temperature condition to obtain a mixture; S3: Sequentially add a rust inhibitor, a metal cleaner, and a pour point depressant into the mixture, and perform stirring treatment under a second constant temperature condition to obtain a high-temperature and high-pressure resistant lubricating oil; It is characterized in that the step S3 includes: Obtain a monitoring video of the stirring state of the mixture collected by a camera; Perform key frame sampling on the monitoring video of the stirring state of the mixture to obtain a time series of key frames of the stirring state of the mixture; Perform timestamp annotation on the stirring state features of each key frame of the stirring state of the mixture in the time series of key frames of the stirring state of the mixture to obtain a time series of stirring state features with time information annotation; Perform encoding and decoding on the time series of stirring state features with time information annotation to obtain an estimated value of the remaining stirring time; Display the estimated value of the remaining stirring time.

2. The preparation method of the high-temperature and high-pressure resistant lubricating oil according to claim 1, wherein, Performing timestamp annotation on the stirring state features of each key frame of the stirring state of the mixture in the time series of key frames of the stirring state of the mixture to obtain a time series of stirring state features with time information annotation includes: Respectively record the timestamps of each key frame of the stirring state of the mixture in the time series of key frames of the stirring state of the mixture to obtain a time series of key frame timestamps; Extract the stirring state features of each key frame of the stirring state of the mixture in the time series of key frames of the stirring state of the mixture to obtain a time series of stirring state feature vectors of the mixture; Perform one-hot embedding encoding on each key frame timestamp in the time series of key frame timestamps to obtain a time series of one-hot embedding encoded vectors of key frame timestamps; Fuse the time series of stirring state feature vectors of the mixture and the time series of one-hot embedding encoded vectors of key frame timestamps to obtain a time series of stirring state feature vectors with time information annotation as the time series of stirring state features with time information annotation.

3. The preparation method of the high-temperature and high-pressure resistant lubricating oil according to claim 2, characterized in that Extracting the stirring state features of each key frame of the stirring state of the mixture in the time series of key frames of the stirring state of the mixture to obtain a time series of stirring state feature vectors of the mixture includes: Using a feature extractor based on a dilated feature pyramid network to extract the stirring state features of each key frame of the stirring state of the mixture in the time series of key frames of the stirring state of the mixture to obtain the time series of stirring state feature vectors of the mixture.

4. The preparation method of the high-temperature and high-pressure resistant lubricating oil according to claim 3, characterized in that, Fuse the time series of the mixture stirring state feature vectors and the time series of the one-hot embedded encoding vectors of the key frame timestamps to obtain a time information annotated time series of the mixture stirring state feature vectors, including: concatenating each corresponding mixture stirring state feature vector and key frame timestamp one-hot embedded encoding vector in the time series of the mixture stirring state feature vectors and the time series of the key frame timestamp one-hot embedded encoding vectors to obtain the time information annotated time series of the mixture stirring state feature vectors.

5. The preparation method of the high-temperature and high-pressure resistant lubricating oil according to claim 1, wherein, Perform stirring state time series context sequence encoding and decoding on the time series of the time information annotated mixture stirring state features to obtain an estimated remaining stirring time, including: Perform stirring state time series context sequence encoding based on dynamic constraints on the time series of the time information annotated mixture stirring state feature vectors to obtain stirring state time series inference encoding vectors; Perform feature decoding on the stirring state time series inference encoding vectors to obtain the estimated remaining stirring time.

6. The preparation method of the high-temperature and high-pressure resistant lubricating oil according to claim 5, characterized in that, Perform stirring state time series context sequence encoding based on dynamic constraints on the time series of the time information annotated mixture stirring state feature vectors to obtain stirring state time series inference encoding vectors, including: Calculate the time information annotated mixture stirring state end information flow constraint parameters and time information annotated mixture stirring state axial information flow constraint parameters of each time information annotated mixture stirring state feature vector in the time series of the time information annotated mixture stirring state feature vectors; Based on the time information annotated mixture stirring state end information flow constraint parameters and time information annotated mixture stirring state axial information flow constraint parameters of each time information annotated mixture stirring state feature vector in the time series of the time information annotated mixture stirring state feature vectors, calculate the time information annotated mixture stirring state context propagation constraint parameters of each time information annotated mixture stirring state feature vector; Based on the time information annotated mixture stirring state context propagation constraint parameters of each time information annotated mixture stirring state feature vector, perform time series context dynamic constraint propagation inference on the sequence distribution of the time information annotated mixture stirring state feature vectors to obtain the stirring state time series inference encoding vectors.

7. The preparation method of the high-temperature and high-pressure resistant lubricating oil according to claim 6, characterized in that, Calculate the time information annotated mixture stirring state end information flow constraint parameters and time information annotated mixture stirring state axial information flow constraint parameters of each time information annotated mixture stirring state feature vector in the time series of the time information annotated mixture stirring state feature vectors, including: Extract the last time information annotated mixture stirring state feature vector from the time series of the time information annotated mixture stirring state feature vectors as the time information annotated mixture stirring state propagation space end constraint anchoring encoding vector; Perform clustering analysis on the time series of the time information annotated mixture stirring state feature vectors to obtain the time information annotated mixture stirring state propagation space axial constraint anchoring encoding vectors; Calculate the time information annotation mixture stirring state end information flow constraint parameters of each time information annotation mixture stirring state feature vector in the time series of the time information annotation mixture stirring state feature vector relative to the time information annotation mixture stirring state propagation space end constraint anchoring coding vector; Calculate the time information annotation mixture stirring state axial information flow constraint parameters of each time information annotation mixture stirring state feature vector in the time series of the time information annotation mixture stirring state feature vector relative to the time information annotation mixture stirring state propagation space axial constraint anchoring coding vector.

8. The preparation method of the high-temperature and high-pressure resistant lubricating oil according to claim 7, characterized in that, Based on the time information annotation mixture stirring state end information flow constraint parameters and the time information annotation mixture stirring state axial information flow constraint parameters of each time information annotation mixture stirring state feature vector in the time series of the time information annotation mixture stirring state feature vector, calculate the time information annotation mixture stirring state context propagation constraint parameters of each time information annotation mixture stirring state feature vector, including: Use the standard space constraint matrix to perform transverse field-induced compactification on the time information annotation mixture stirring state end information flow constraint parameters and the time information annotation mixture stirring state axial information flow constraint parameters corresponding to the time information annotation mixture stirring state feature vector to obtain the time information annotation mixture stirring state end information flow compact constraint parameters and the time information annotation mixture stirring state axial information flow compact constraint parameters; Based on the time information annotation mixture stirring state end information flow compact constraint parameters and the time information annotation mixture stirring state axial information flow compact constraint parameters, perform constraint condition consistency on the time information annotation mixture stirring state end information flow constraint parameters and the time information annotation mixture stirring state axial information flow constraint parameters to obtain the optimized time information annotation mixture stirring state end information flow constraint parameters and the optimized time information annotation mixture stirring state axial information flow constraint parameters; Perform non-linear weighted processing on the optimized time information annotation mixture stirring state end information flow constraint parameters and the optimized time information annotation mixture stirring state axial information flow constraint parameters to obtain the time information annotation mixture stirring state context propagation constraint parameters corresponding to the time information annotation mixture stirring state feature vector.

9. The preparation method of the high-temperature and high-pressure resistant lubricating oil according to claim 8, characterized in that, Perform feature decoding on the stirring state time series inference coding vector to obtain an estimated value of the remaining stirring time, including: inputting the stirring state time series inference coding vector into a stirring time analyzer based on a decoder to obtain an estimated value of the remaining stirring time.

10. A high-temperature and high-pressure resistant lubricating oil, characterized in that, The high-temperature and high-pressure resistant lubricating oil is prepared by the preparation method of the high-temperature and high-pressure resistant lubricating oil according to any one of claims 1-9.

Images