Automatic water feeding and supplementing device and method for slurry pump

Abstract

The invention provides a slurry pump water feeding automatic liquid supplementing device and method, and relates to the field of intelligent equipment.The method comprises the steps that local time sequence feature extraction is conducted on a time sequence set of slurry pump outlet end pressure data through the data coding and processing technology based on artificial intelligence, and then time sequence correlation is conducted on the local time sequence features of the outlet end pressure; and meanwhile, performing multi-door structure propagation on each outlet end pressure local time sequence characteristic, so as to adaptively control the rotating speed value of the variable frequency motor at the next time point according to the fusion representation between the outlet end pressure time sequence characteristic significant propagation representation and the outlet end pressure local time sequence distribution characteristics after time sequence association. Therefore, the working parameters of the variable-frequency liquid supplementing pump can be timely adjusted through the change feedback of the outlet pressure of the slurry pump, timely and appropriate liquid supplementing is ensured, and the working efficiency and stability of the slurry pump are improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent devices, and more specifically, to an automatic water replenishment device and method for a mud pump. Background Art

[0002] In modern industrial automation and oil drilling, mud pumps are indispensable key equipment. Their main function is to suck mud from the mud purification tank, transport the mud to the bottom of the well after pressurization, and effectively carry the cuttings generated during the drilling process back to the ground and send them to the mud purification system for treatment. With the increase in well depth and the development of ultra-deep wells, it is particularly important to maintain a stable mud pressure during long-distance transportation, because pressure fluctuations may cause damage to drilling tools or cause underground accidents.

[0003] The traditional approach is to increase the height of the tank to achieve the mud suction port pressure or to match a sand pump, and start the sand pump to pump slurry to the mud pump suction port at the same time as the mud pump is started. However, this method of increasing the height of the tank is not only costly, but also often has unsatisfactory effects, and it is difficult to meet the demand for stable high pressure in deep and ultra-deep well drilling. Although the method of starting a sand pump at the same time as the mud pump can solve the problem to a certain extent, it cannot achieve on-demand replenishment, resulting in inaccurate replenishment, affecting the working efficiency and stability of the mud pump.

[0004] Therefore, an optimized mud pump water replenishment solution is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a mud pump water automatic liquid replenishing device and method.

[0006] According to one aspect of the present application, there is provided an automatic water replenishment device for a mud pump, which comprises: a mud tank; a mud pump for sucking, pressurizing and discharging the mud in the mud tank; a pressure sensor disposed at the outlet end of the mud pump, for collecting the outlet pressure data of the mud pump in real time; a variable frequency replenishment pump disposed between the mud tank and the mud pump, for replenishing the mud pump with water; a central processing unit, for controlling the speed of a variable frequency motor of the variable frequency replenishment pump based on the outlet pressure data of the mud pump; wherein the central processing unit comprises: a pressure data acquisition unit, for acquiring a time series set of outlet pressure data acquired by the pressure sensor; a pressure time series extraction unit, for performing local time series on the time series set of outlet pressure data; Feature extraction to obtain a sequence of local timing features of the outlet pressure; a pressure data timing association unit, used to perform timing association on the sequence of local timing features of the outlet pressure to obtain local timing distribution features of the outlet pressure; a pressure timing propagation unit, used to perform feature significant propagation of the sequence of local timing features of the outlet pressure with a multiple gate structure to obtain a significant propagation representation of the timing features of the outlet pressure; an outlet pressure timing multi-dimensional representation unit, used to fuse the local timing distribution features of the outlet pressure and the significant propagation representation of the timing features of the outlet pressure to obtain a multi-dimensional representation of the timing of the outlet pressure, and based on the multi-dimensional representation of the timing of the outlet pressure, obtain a control result, and the control result is used to represent the recommended decoding value of the speed of the variable frequency motor at the next time point.

[0007] According to another aspect of the present application, a method for automatic water replenishment of a mud pump is provided, which is used to execute the automatic water replenishment device of the mud pump as described above, and comprises: acquiring a time series set of outlet pressure data collected by the pressure sensor; performing local time series feature extraction on the time series set of the outlet pressure data to obtain a sequence of local time series features of the outlet pressure; performing time series correlation on the sequence of local time series features of the outlet pressure to obtain a local time series distribution feature of the outlet pressure; performing feature significant propagation of a multiple gate structure on the sequence of local time series features of the outlet pressure to obtain a significant propagation representation of the outlet pressure time series feature; fusing the local time series distribution feature of the outlet pressure and the significant propagation representation of the outlet pressure time series feature to obtain a multi-dimensional representation of the outlet pressure time series, and obtaining a control result based on the multi-dimensional representation of the outlet pressure time series, wherein the control result is used to represent a recommended decoding value of the speed of the variable frequency motor at the next time point.

[0008] Compared with the prior art, the automatic refilling device and method for the mud pump provided by the present application uses artificial intelligence-based data encoding and processing technology to extract local time series features from the time series set of the pressure data at the outlet of the mud pump, and then performs time series association on each local time series feature of the outlet pressure, and at the same time, performs multiple gate structure propagation on each local time series feature of the outlet pressure, so as to adaptively control the speed value of the variable frequency motor at the next time point according to the fusion representation between the significant propagation representation of the outlet pressure time series feature and the local time series distribution feature of the outlet pressure after time series association. In this way, the working parameters of the variable frequency refilling pump can be adjusted in time through the feedback of the change of the outlet pressure of the mud pump, ensuring timely and appropriate refilling, which is beneficial to improving the working efficiency and stability of the mud pump. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0010] In the attached picture: Figure 1 It is a block diagram of the automatic liquid replenishing device for a mud pump according to an embodiment of the present application.

[0011] Figure 2 The block diagram is a central processing unit in the automatic water replenishment device for a mud pump according to an embodiment of the present application.

[0012] Figure 3 It is a block diagram of a pressure timing propagation unit in a mud pump water automatic replenishment device according to an embodiment of the present application.

[0013] Figure 4 It is a block diagram of a multi-dimensional characterization unit of the outlet pressure timing in the automatic water replenishment device for a mud pump according to an embodiment of the present application.

[0014] Figure 5 This is an operation flow chart of the automatic liquid replenishing device for a mud pump according to other embodiments of the present application.

[0015] Figure 6 The present invention is a flow chart of a method for automatically replenishing water in a mud pump according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0017] In the field of modern industrial automation and oil drilling, mud pumps are key equipment that sucks mud from the mud purification tank and pressurizes it to the bottom of the well, while carrying rock cuttings back to the surface for processing. As the well depth increases, it is crucial to keep the mud pressure stable during long-distance transportation. Traditional methods such as increasing the tank height or using a sand pump for assistance have problems such as high cost, unsatisfactory results, or inaccurate fluid replenishment, which are difficult to meet the needs of deep well drilling.

[0018] Based on this, the present application proposes an automatic liquid replenishing device for a mud pump. Figure 1 FIG. 1 is a block diagram of an automatic liquid replenishing device for a mud pump according to an embodiment of the present application. Figure 1 As shown, in the mud pump water filling automatic fluid replenishment device 100, it includes: a mud tank 110; a mud pump 120 for sucking, pressurizing and discharging the mud in the mud tank; a pressure sensor 130 deployed at the outlet end of the mud pump, for collecting the outlet pressure data of the mud pump in real time; a variable frequency fluid replenishment pump 140 deployed between the mud tank and the mud pump, for filling the mud pump with water; a central processing unit 150, for controlling the speed of the variable frequency motor of the variable frequency fluid replenishment pump based on the outlet pressure data of the mud pump.

[0019] In the embodiment of the present application, the mud pump water automatic refilling device includes a mud tank 110. It should be understood that the mud tank is a storage container for mud. During the oil drilling process, the mud is used to cool and lubricate drilling tools such as drill pipes, and to carry the rock cuttings generated during the drilling process back to the ground. The mud tank provides sufficient mud reserves for the entire drilling process to ensure that the drilling process can continue.

[0020] In an embodiment of the present application, the mud pump water supply automatic liquid replenishing device includes a mud pump 120 for sucking, pressurizing and discharging the mud in the mud tank. It should be understood that the primary function of the mud pump is to generate suction to suck mud from the mud tank. In addition, the mud pump can also pressurize the sucked mud. With the increase of well depth, especially in deep well and ultra-deep well drilling, sufficient pressure is required to transport the mud to the bottom of the well. Through the pressurization treatment of the mud pump, the mud can overcome resistances including friction in the wellbore and the gravity of the mud itself and be smoothly transported to the bottom of the well, thereby ensuring that the drilling work can be carried out normally at all depths underground. The mud pump also has the function of discharging mud. The mud discharged by the mud pump washes the cuttings at the bottom of the well, and carries the cuttings to the ground through its own flow, and enters the mud purification system, thereby maintaining the cleanliness and normal operation of the drilling operation environment.

[0021] The core function of the mud pump is to accurately and efficiently suck in the mud in the mud tank, pressurize it, and then discharge it to meet the needs of the operation. This process involves multiple complex links, which are closely linked and work together to ensure the stable operation of the mud pump.

[0022] The pump body structure design of the mud pump is the basis for realizing its function. Among them, the impeller is a key component, and its exquisite design is reflected in many aspects. The centrifugal impeller is commonly used, and its curved blades give the mud a strong centrifugal force when it rotates at high speed. This centrifugal force pushes the mud from the center of the impeller to the edge, thereby driving the suction and transportation process of the mud. Parameters such as the number, angle and thickness of the blades are not set arbitrarily, but are rigorously calculated and optimized. The number of blades needs to seek a balance between the fluidity of the mud and the efficiency of the pump. Although more blades can enhance the mud pushing ability, they may also increase the flow resistance due to mutual interference between the blades; the precise design of the blade angle directly affects the speed and direction of the mud inlet and outlet, and thus has a decisive impact on the overall performance of the pump. The choice of impeller material is even more related to its durability and reliability. High-chromium alloy cast iron, stainless steel and other high-strength, wear-resistant and corrosion-resistant materials are the first choice. These materials enable the impeller to maintain stable operation for a long time under the continuous scouring of solid particles in the mud and the erosion of chemical substances, ensuring the continuous and efficient operation of the mud pump.

[0023] The design of the pump casing should not be overlooked either. The volute-shaped pump casing not only provides a suitable space for the impeller to rotate, but also cleverly guides the mud to the correct inlet and outlet positions. Its smooth inner wall minimizes the resistance to mud flow and effectively improves the working efficiency of the pump. The carefully designed inlet and outlet pipe connection interfaces on the pump casing can ensure that the mud can smoothly enter and exit the pump body, and achieve seamless connection with the mud tank and the delivery pipeline. In terms of strength, the pump casing must be able to withstand high pressure. Sealing is also a key consideration in the design of the pump casing. Advanced sealing devices such as sealing gaskets and mechanical seals are used at the connections between the pump casing and the impeller, inlet and outlet pipes, etc., which can effectively prevent mud leakage and ensure the normal operation of the pump.

[0024] The suction process is the starting point of the mud pump, and its core is to form negative pressure to drive the mud into the pump body. When the mud pump starts, the impeller starts to rotate at high speed, instantly creating a local vacuum environment in the center of the impeller, that is, the negative pressure area. At this time, the mud in the mud tank flows into the pump body along the suction pipe under the strong thrust of the external atmospheric pressure. The diameter of the suction pipe needs to be carefully matched according to the flow rate and viscosity of the mud. Although a larger diameter suction pipe can reduce the suction resistance and make the mud flow smoother, it is also necessary to weigh the installation space and cost factors of the pump. In order to further enhance the suction capacity, some mud pumps are equipped with auxiliary devices. The vacuum pump first extracts the air in the pump body before the pump is started, further reducing the pressure in the pump, thereby significantly enhancing the suction effect; the jet pump uses the high-speed jet generated by the high-pressure fluid to create additional negative pressure in the suction pipe to help it enter the pump body smoothly. However, during the suction process, cavitation phenomenon needs to be vigilant at all times. Cavitation is caused by low local pressure in the pump, which causes the liquid in the mud to vaporize and form bubbles. These bubbles burst instantly when the mud flows to the high-pressure area, releasing a strong high-pressure shock, causing serious damage to the pump body components. To resist this threat, the installation height of the mud pump must be strictly controlled to ensure that the suction port pressure is always higher than the saturated vapor pressure of the mud at the current temperature. At the same time, it is crucial to optimize the layout of the suction pipe to reduce pipe bending and resistance and keep the pressure in the suction pipe stable. In addition, the choice of impeller speed must be carefully weighed. Although too high a speed can improve the suction capacity, it may also cause a cavitation crisis due to low local pressure.

[0025] The pressurization process is a key stage for the mud pump to increase the energy of the mud, which mainly relies on centrifugal force. The strong centrifugal force generated by the rotation of the impeller not only throws the mud from the center to the edge, but also gives it considerable kinetic energy. As the mud flows in the volute, the shape of the volute gradually changes, causing the mud flow rate to gradually decrease. According to the law of conservation of energy, the kinetic energy of the mud is smoothly converted into pressure energy in this process, achieving effective pressurization of the mud. The impeller speed plays an important role in this process. The higher the speed, the stronger the centrifugal force, the more kinetic energy the mud obtains, and the more significant the pressurization effect. However, the higher the impeller speed is, the better. It is necessary to comprehensively consider many factors such as the mechanical strength of the pump, power consumption, and the characteristics of the mud itself. For some application scenarios with extremely high pressure requirements, the mud pump adopts a multi-stage pressurization structure. For example, a multi-stage pump is composed of multiple impellers and volutes connected in series. The mud flows through each stage of impellers and volutes in turn, and the pressure is further increased after each stage. This multi-stage boosting method can achieve high output pressure at a relatively low impeller speed, while reducing the load on the single-stage impeller, improving the reliability and service life of the pump. In a multi-stage pump, the parameters of the impeller and volute at each stage must be carefully designed and perfectly matched to ensure the efficiency and stability of the entire boosting process.

[0026] The discharge process marks the final output of the mud pump's work. The pressurized mud accumulates a high pressure in the pump body. Driven by this strong pressure, the mud is discharged from the pump body through the outlet pipe and enters the conveying pipe to the bottom of the well or other destinations. The diameter and material selection of the outlet pipe need to be carefully considered based on factors such as mud flow, pressure and conveying distance. A larger diameter outlet pipe can reduce the mud flow rate and reduce flow resistance, but it also means higher construction costs; the material must have strong pressure resistance and corrosion resistance. High-strength steel pipes or composite pipes lined with wear-resistant materials have become common choices, which can protect the mud from damage during long-distance transportation.

[0027] In summary, from the careful design of the mud pump body structure to the precise control of the suction, pressurization and discharge processes, the mud pump can operate efficiently and stably in fields such as oil drilling, thereby ensuring the smooth progress of various operations.

[0028] In the embodiment of the present application, the automatic refilling device for the mud pump includes a pressure sensor 130 disposed at the outlet end of the mud pump, which is used to collect the outlet pressure data of the mud pump in real time. It should be understood that the stable operation of the mud pump is crucial in the process of oil drilling. With the increase of well depth and complex downhole working conditions, the pressure at the outlet end of the mud pump is easily affected by various factors and fluctuates. Real-time collection of pressure data at the outlet end of the mud pump can timely discover the change of pressure so as to adjust the working state of the mud pump and prevent system failures caused by excessive or low pressure, such as damage to the drilling tool or downhole accidents. Specifically, in order to enable the mud pump to work effectively at different drilling depths and working conditions, it is necessary to control the amount of refilling according to the actual pressure conditions. Only by understanding the outlet pressure data in real time can accurate refilling operation be performed through the automatic refilling device to ensure that the output pressure of the mud pump is maintained within a suitable range, thereby maintaining stable mud circulation in deep well and ultra-deep well drilling.

[0029] In an embodiment of the present application, the automatic water replenishment device for the mud pump includes a variable frequency replenishment pump 140 disposed between the mud tank and the mud pump, and the variable frequency replenishment pump is used to replenish the mud pump with water. It should be understood that water replenishment specifically refers to replenishing mud in the mud pump. As the mud is continuously sucked in, pressurized and transported to the bottom of the well by the mud pump, the amount of mud inside the mud pump will gradually decrease. If the mud is not replenished in time, the pressure in the mud pump will drop. By replenishing water, a sufficient amount of mud can be maintained in the mud pump, thereby maintaining a stable output pressure.

[0030] In the embodiment of the present application, the automatic water replenishment device for the mud pump includes a central processor 150, and the central processor is used to control the speed of the variable frequency motor of the variable frequency replenishment pump based on the outlet pressure data of the mud pump.

[0031] It should be understood that the variable frequency refill pump replenishes fluid according to the feedback of the mud pump outlet pressure, so the immediacy and accuracy of the pressure data are crucial. However, traditional variable frequency refill pumps usually rely on fixed parameter settings or simple threshold control, and cannot be dynamically adjusted according to the real-time changes in the mud pump outlet pressure. This static control method is difficult to cope with complex and changing working conditions, resulting in inaccurate refilling and easily causing problems of excessive or insufficient fluid supply.

[0032] Based on this, in the central processing unit, the technical concept of the present application is to obtain the time series set of the outlet pressure data collected by the pressure sensor, and use data encoding and processing technology based on artificial intelligence to extract the local time series features of the outlet pressure data, and then perform time series association on each local time series feature of the outlet pressure, and at the same time, perform multiple gate structure propagation on each local time series feature of the outlet pressure, so as to adaptively control the variable frequency motor speed value at the next time point according to the fusion representation between the significant propagation representation of the outlet pressure time series feature and the local time series distribution feature of the outlet pressure after time series association. In this way, the pressure data can be analyzed in real time, and the complex time dependence between each time point can be captured, so as to ensure that the variable frequency infusion pump is adaptively and accurately adjusted according to actual needs, avoiding the problem of inaccurate infusion caused by fixed parameter settings in traditional methods.

[0033] Figure 2 FIG. 1 is a block diagram of a central processing unit in a mud pump water automatic replenishment device according to an embodiment of the present application. Figure 2 As shown, the central processor 150 includes: a pressure data acquisition unit 151, which is used to acquire the time series set of the outlet pressure data collected by the pressure sensor; a pressure time series extraction unit 152, which is used to extract local time series features from the time series set of the outlet pressure data to obtain a sequence of local time series features of the outlet pressure; a pressure data time series association unit 153, which is used to perform time series association on the sequence of local time series features of the outlet pressure to obtain local time series distribution features of the outlet pressure; a pressure time series propagation unit 154, which is used to perform feature significant propagation of multiple gate structures on the sequence of local time series features of the outlet pressure to obtain a significant propagation representation of the outlet pressure time series features; an outlet pressure time series multi-dimensional representation unit 155, which is used to fuse the local time series distribution features of the outlet pressure and the significant propagation representation of the outlet pressure time series features to obtain a multi-dimensional representation of the outlet pressure time series, and obtain a control result based on the outlet pressure time series multi-dimensional representation, and the control result is used to represent the recommended decoding value of the speed of the variable frequency motor at the next time point.

[0034] In an embodiment of the present application, the pressure data acquisition unit 151 is used to acquire a time series set of outlet pressure data collected by the pressure sensor. It should be understood that the time series set of outlet pressure data contains the outlet pressure data of the mud pump arranged in chronological order. By analyzing this time series set, the changing trend of the pressure of the variable frequency rehydration pump can be obtained. For example, whether there is a linear growth or decline trend in pressure, or whether there is a periodic pressure fluctuation, these trend information helps to understand the working state of the mud pump. It provides feedback information for controlling the speed of the variable frequency motor. Specifically, when the pressure shows a slow downward trend, the central processing unit can gradually increase the speed of the variable frequency motor to achieve a smooth rehydration process. However, when the pressure suddenly drops sharply, the central processing unit may quickly increase the speed of the variable frequency motor to quickly replenish the mud and restore the pressure. Similarly, in the case of pressure fluctuations, the central processing unit can adjust the speed of the variable frequency motor according to the fluctuation frequency and amplitude, so that it can compensate for the pressure fluctuations in time and maintain the relative stability of the pressure at the outlet of the mud pump.

[0035] In order to ensure that the mud pump can maintain a stable pressure when transporting mud over long distances, accurately obtaining the time series set of the pressure data at the outlet of the mud pump is a key prerequisite for achieving efficient control and precise fluid replenishment. The following will elaborate on the specific implementation process of obtaining the time series set of the outlet pressure data collected by the pressure sensor deployed at the outlet of the mud pump: First, the selection and installation of the pressure sensor. In terms of sensor selection, it is necessary to carefully select a pressure sensor with high precision, high reliability and good anti-interference ability. This sensor can accurately measure the pressure changes at the outlet of the mud pump in a complex and harsh industrial environment, and is not affected by the mud flow characteristics, strong vibration and various electromagnetic interferences. In terms of installation position, the sensor should be placed at the key position of the outlet of the mud pump to ensure that the collected data can truly reflect the working status of the mud pump. During the installation process, strictly control the installation direction to ensure that the sensor measurement surface is perpendicular to the mud flow direction to obtain the most accurate pressure value; at the same time, pay attention to sealing to prevent mud from leaking into the sensor and affecting the measurement accuracy, and use high-quality sealing materials and fastening processes to ensure that the sensor is stable and reliable in long-term operation.

[0036] Next is the setting of data acquisition frequency. This link requires comprehensive consideration of many factors such as the working characteristics of the mud pump, the sensitivity of the response to pressure changes, the processing power of the central processor, and the data storage capacity. For mud pumps with rapid changes in working conditions and frequent pressure fluctuations, in order to capture every subtle pressure change in time, it may be necessary to set a higher acquisition frequency, such as several times per second or even higher. In this way, a large number of pressure data points can be obtained in a short period of time, providing a sufficient data basis for subsequent real-time analysis. However, for mud pumps with relatively stable operation and relatively slow pressure changes, too high an acquisition frequency may cause data redundancy, increase the processing burden of the central processor, and also take up too much storage space. At this time, the acquisition frequency can be appropriately reduced, but it is still necessary to ensure that the trend changes in pressure can be accurately reflected to avoid missing important pressure fluctuation information due to too low an acquisition frequency.

[0037] Data transmission and storage are important steps to obtain a time series set. The pressure sensor transmits the collected pressure data to the central processor. The transmission method can be wired or wireless according to the actual working conditions. Wired transmission, such as shielded cables, can provide a stable and reliable data transmission link, effectively resist external electromagnetic interference, and ensure the integrity and accuracy of the data. However, this method may face certain difficulties in wiring, especially in complex industrial field environments, and the wiring path needs to be carefully planned to avoid mechanical damage to the cable. Wireless transmission methods such as Bluetooth, Wi-Fi or dedicated wireless communication modules have the advantage of high installation flexibility, which is convenient for rapid deployment of sensors in different locations. However, wireless transmission requires special attention to the stability and security of communication. By adopting advanced communication protocols, encryption technology and signal enhancement measures, it is ensured that the data is not lost or tampered with during transmission. After receiving the data, the central processor will store it in chronological order to construct a time series set of pressure data. During the storage process, the recently collected data is usually cached in the memory first, and the high-speed read and write characteristics of the memory are used to facilitate rapid processing and analysis. At the same time, in order to prevent data loss due to system failure or unexpected power outage, the data in the memory will be regularly written to the disk for long-term preservation to achieve persistent storage of data.

[0038] In the process of data collection, data preprocessing and calibration are indispensable. Data cleaning is an important part of preprocessing, and its purpose is to remove outliers and noise in the collected data. Outliers may be caused by sudden failure of sensors, instantaneous strong interference signals or other abnormal working conditions. By setting a reasonable threshold range, these outliers can be effectively identified and eliminated. For example, according to the normal working pressure range of the mud pump, combined with a certain safety factor, a pressure upper and lower limit threshold is determined. When the collected data exceeds the threshold range, it can be judged as an outlier and eliminated. In addition, in order to improve the accuracy of the data, a filtering algorithm is also used to smooth the data. Common filtering algorithms include mean filtering and median filtering. Mean filtering replaces the original data by calculating the average value of pressure data in a certain time window, effectively smoothing out high-frequency noise; median filtering takes the median of the data in the time window as the filtering result, which is effective in removing impulse noise. According to the characteristics of pressure data and noise type, selecting a suitable filtering method can significantly improve data quality.

[0039] Data calibration is also crucial. To ensure the accuracy of the pressure sensor measurement data, the sensor needs to be calibrated regularly using a standard pressure source. During the calibration process, the known precise pressure value generated by the standard pressure source is applied to the sensor, and the difference between the sensor's measured output value and the standard value is compared. According to the difference, the sensor's measurement parameters, such as gain, offset, etc., are adjusted to keep the sensor's measurement accuracy at a high level. At the same time, considering the impact of environmental factors on pressure measurement, especially the large impact of temperature changes on sensor measurement results, temperature compensation technology will be used to correct the data. By integrating a temperature sensor inside the sensor, the ambient temperature is monitored in real time, and the pre-established temperature-pressure compensation model is used to compensate the measured pressure data, eliminate the measurement error caused by temperature changes, and further improve the reliability of the data.

[0040] Finally, the time series set management. When constructing the time series set, each collected pressure data is arranged in the order of the time of collection to obtain the time series set of the outlet pressure data, which provides a solid foundation for the subsequent in-depth analysis of the time series characteristics of pressure changes. In terms of data storage structure selection, the efficiency of data query, retrieval and processing needs to be comprehensively considered. The array structure is simple and intuitive, suitable for scenarios with small data volume and extremely high real-time requirements; the linked list structure has advantages in dynamic data insertion and deletion operations, but the query efficiency is relatively low; the database table structure is powerful and can easily perform complex query operations, such as retrieval by time range, data statistical analysis, etc., and is convenient for data management and maintenance, so it is an ideal choice in most cases. When using database table storage, the timestamp is generally used as the primary key to ensure the uniqueness and orderliness of the data, and the pressure data is stored as a field. By establishing a suitable index, the data query speed can be greatly improved. With the continuous operation of the mud pump, the pressure data is constantly updated, and the time series set is also in dynamic change. The central processor needs to process the newly collected data in real time, add it to the time series set in an orderly manner, and update the relevant statistical information and feature data in a timely manner. At the same time, in order to ensure the efficient operation of the device, it is necessary to regularly clean up and archive the historical data in the time series set. Deleting old data can not only free up valuable storage space, but also avoid the decline in query and processing efficiency due to excessive data volume, ensuring that the entire device always maintains good performance during long-term operation.

[0041] In the embodiment of the present application, the pressure time series extraction unit 152 is used to extract local time series features from the time series set of the outlet pressure data to obtain a sequence of local time series features of the outlet pressure. Specifically, in the embodiment of the present application, the pressure time series extraction unit is used to: input the time series set of the outlet pressure data into a pressure time series feature extractor based on a one-dimensional causal convolutional neural network model to obtain a sequence of local time series feature vectors of the outlet pressure as the sequence of local time series features of the outlet pressure. It should be understood that the outlet pressure data is constantly changing with the passage of time, and the information at the current time point depends on the previous historical data, and different local time periods have different time series characteristics. Therefore, in order to more clearly understand and analyze the local time series change feature information of the pressure data, such as transient response, periodic pattern, etc., in the technical solution of the present application, the time series set of the outlet pressure data is input into a pressure time series feature extractor based on a one-dimensional causal convolutional neural network model to slide along the time axis using the convolution kernel, and perform a local weighted sum operation with the input data to capture a specific local time pattern or feature, and obtain a sequence of local time series feature vectors of the outlet pressure. In particular, the one-dimensional causal convolutional neural network is a deep learning model specifically designed to process time series data. Specifically, it maintains causality when processing sequence data, that is, only the data at the current moment and before are used for prediction or analysis, without considering future data points. In other words, the one-dimensional causal convolutional neural network is particularly suitable for processing time series data because it can capture local temporal dependencies without violating the causal principle, which means that when predicting or analyzing data at a certain moment, only the data before that moment will be used, ensuring the physical rationality of the model and providing rich data support for the subsequent recommendation of pressure values.

[0042] In the embodiment of the present application, the pressure data timing association unit 153 is used to perform timing association on the sequence of the local timing characteristics of the outlet pressure to obtain the local timing distribution characteristics of the outlet pressure. Specifically, in the embodiment of the present application, the pressure data timing association unit is used to: cascade each outlet pressure local timing characteristic vector in the sequence of the outlet pressure local timing characteristic vector to obtain the outlet pressure local timing distribution characteristic vector as the outlet pressure local timing distribution characteristic. It should be understood that each outlet pressure local timing characteristic represents the local timing change information of the outlet pressure at a local time point. Therefore, in order to more comprehensively describe the dynamic behavior of the system over a period of time, in the technical solution of the present application, each outlet pressure local timing characteristic vector in the sequence of the outlet pressure local timing characteristic vector is cascaded to capture and integrate the overall timing pattern and trend within the entire time series range to obtain the outlet pressure local timing distribution characteristic vector.

[0043] In an embodiment of the present application, the pressure time series propagation unit 154 is used to perform feature significant propagation of multiple gate structures on the sequence of the local time series features of the outlet pressure to obtain a significant propagation representation of the outlet pressure time series features. In particular, considering that there are time series correlation characteristics between the local time series features of the outlet pressure at each time point in the sequence of the local time series feature vectors of the outlet pressure, and there are inherent implicit correlation characteristics between the various time points. Therefore, in order to further understand the inherent representation of the local time series features of the outlet pressure during the time transfer process, in the technical solution of the present application, the sequence of the local time series features of the outlet pressure is processed by feature significant propagation of multiple gate structures to obtain a significant propagation representation of the outlet pressure time series features. That is, the multiple gate structure allows the model to adaptively select the most relevant features for propagation and suppress irrelevant or redundant information, which helps to improve the model's sensitivity to key features, thereby more accurately reflecting the temporal changes in the outlet pressure state.

[0044] Specifically, Figure 3 FIG. 1 is a block diagram of a pressure timing propagation unit in a mud pump water automatic replenishment device according to an embodiment of the present application. Figure 3 As shown, the pressure time series propagation unit 154 includes: an outlet pressure local time series feature space mapping subunit 1541, which is used to map each outlet pressure local time series feature vector in the sequence of the outlet pressure local time series feature vectors to the Poincare space to obtain a sequence of outlet pressure local time series feature heterogeneous modulation coding feature vectors; an outlet pressure local time series feature heterogeneous propagation representation generating subunit 1542, which is used to perform forward LSTM sequence encoding on the sequence of the outlet pressure local time series feature heterogeneous modulation coding feature vectors to obtain an outlet pressure local time series feature heterogeneous propagation representation vector; an outlet pressure local time series feature heterogeneous propagation representation generating subunit 1542, which is used to perform forward LSTM sequence encoding on the sequence of the outlet pressure local time series feature heterogeneous modulation coding feature vectors to obtain an outlet pressure local time series feature heterogeneous propagation representation vector; The outlet pressure local time series significant propagation subunit 1543 is used to perform significant propagation on the sequence of the outlet pressure local time series characteristic alien modulation coding feature vector based on the outlet pressure local time series characteristic alien propagation representation vector to obtain the outlet pressure alien characteristic significant propagation representation vector; the outlet pressure local time series characteristic aggregation subunit 1544 is used to perform residual mapping processing on the outlet pressure alien characteristic significant propagation representation vector and the outlet pressure local time series characteristic alien propagation representation vector to obtain the outlet pressure timing characteristic significant propagation representation vector as the outlet pressure timing characteristic significant propagation representation.

[0045] Specifically, first, each outlet pressure local time series feature vector in the sequence of the outlet pressure local time series feature vector is mapped to the Poincare space to obtain a sequence of outlet pressure local time series feature heterogeneous modulation coding feature vectors. The above process can be expressed by the formula: ;in, is the sequence of local time series characteristic vectors of the outlet pressure, and are the first, second, and third in the sequence of the local time series characteristic vector of the outlet pressure respectively. and The local time series characteristic vector of the outlet pressure, and are the first weight matrix and the second weight matrix respectively, and are the first, second, and third in the sequence of the characteristic vectors of the local temporal characteristics of the outlet pressure in the heterogeneous modulation coding. and The local temporal characteristics of the outlet pressure are modulated and coded in a special domain. It is a sequence of the characteristic vectors of the heterogeneous modulation coding of the local time series characteristics of the outlet pressure.

[0046] It should be understood that the Poincare space is a hyperbolic space that can maintain the distance relationship between nodes while effectively reducing the curse of dimensionality. By mapping each local temporal feature vector of the outlet pressure in the sequence of the local temporal feature vector of the outlet pressure to the Poincare space, the unique geometric properties of the Poincare space can be used to capture the semantic association between the local temporal features at different temporal nodes of the outlet pressure.

[0047] Next, the sequence of the outlet pressure local time series feature heterogeneous modulation coding feature vectors is forward LSTM sequence encoded to obtain the outlet pressure local time series feature heterogeneous propagation representation vector. The above process can be expressed by the formula: ;in, is the sequence of the characteristic vectors of the heterogeneous modulation coding of the local temporal characteristics of the outlet pressure, Encode the forward LSTM. It is the vector representing the alien propagation of the local time series characteristics of the outlet pressure.

[0048] It should be understood that the forward long short-term memory network plays a key role in the forward LSTM sequence encoding of the sequence of the feature vector of the local temporal characteristics of the outlet pressure. Its forget gate can control the degree of retention of the information at the previous moment and determine which information is forgotten; the input gate determines the importance of the current input information and filters it for inclusion; the output gate is responsible for determining which information in the current hidden state will be output for the next moment calculation. Through forward LSTM encoding, the trend and subtle fluctuations of the outlet pressure change can be accurately captured, and the overfitting problem caused by processing long time series can be effectively avoided.

[0049] Specifically, in the embodiment of the present application, the outlet pressure local time series significant propagation subunit 1543 is used to: calculate the semantic association between each outlet pressure local time series feature heterogeneous modulation coding feature vector in the sequence of the outlet pressure local time series feature heterogeneous modulation coding feature vector and the outlet pressure local time series feature heterogeneous propagation representation vector to obtain a sequence of outlet pressure semantic associations. This process can be expressed as: ;in, is the sequence of the characteristic vector of the outlet pressure local time series characteristic heterogeneous modulation coding. The local temporal characteristics of the outlet pressure are modulated and coded in a special domain. is the vector representing the propagation of the local temporal characteristics of the pressure at the outlet, To calculate the length of a vector, is the inverse hyperbolic cosine function, and They are the first, second, and third in the sequence of semantic relevance of outlet pressure. and The semantic relevance of the pressure at the outlet, is a sequence of the semantic associations of the outlet pressure.

[0050] The semantic space jump value of each outlet pressure local time series feature heterogeneous modulation coding feature vector in the sequence of the outlet pressure local time series feature heterogeneous modulation coding feature vector relative to the outlet pressure local time series feature heterogeneous propagation representation vector is calculated to obtain a sequence of outlet pressure semantic space jump values. This process can be expressed as: ;in, is the sequence of the characteristic vector of the outlet pressure local time series characteristic heterogeneous modulation coding. The local temporal characteristics of the outlet pressure are modulated and coded in a special domain. is the vector representing the propagation of the local temporal characteristics of the pressure at the outlet, To calculate the length of a vector, and are the first, second, and third values ​​in the sequence of the semantic space jump value of the outlet pressure. and The exit pressure semantic space jump value, is a sequence of the outlet pressure semantic space jump values.

[0051] The sequence of the outlet pressure semantic associations and the sequence of the outlet pressure semantic space jump values ​​are subjected to information transmission of a multiple gate structure to obtain a sequence of outlet pressure information transmission modulation weights. The process can be expressed as: ;in, is the sequence of the semantic associations of the outlet pressure, is the sequence of the exit pressure semantic space jump values, and are the semantic association weight matrix and the semantic space weight matrix respectively. is matrix multiplication, is the bias vector, yes function, is the sequence of modulation weights for the outlet pressure information transmission.

[0052] Based on the sequence of the outlet pressure information transmission modulation weights, the position-weighted sum of the sequence of the outlet pressure local temporal feature heterogeneous modulation coding feature vectors is calculated to obtain the outlet pressure heterogeneous feature significant propagation representation vector. This process can be expressed as: ;in, is the number of vectors in the sequence of the outlet pressure local time series characteristic heterogeneous modulation coding feature vector, is the sequence of the characteristic vector of the outlet pressure local time series characteristic heterogeneous modulation coding. The local temporal characteristics of the outlet pressure are modulated and coded in a special domain. yes The corresponding outlet pressure information transmission modulation weight, is the vector representing the significant propagation of the pressure heterogeneity characteristics at the outlet.

[0053] That is, by calculating the semantic association between each outlet pressure local time series feature heterogeneous modulation coding feature vector in the sequence of the outlet pressure local time series feature heterogeneous modulation coding feature vector and the outlet pressure local time series feature heterogeneous propagation representation vector, the semantic intrinsic association between the outlet pressure local time series propagation feature and each outlet pressure local time series feature heterogeneous modulation coding feature vector in the Poincare space can be expressed, which is conducive to the subsequent introduction of the attention mechanism to enhance the model's attention to the specific outlet pressure local time series feature. Next, by calculating the semantic space jump value of each outlet pressure local time series feature heterogeneous modulation coding feature vector in the sequence of the outlet pressure local time series feature heterogeneous modulation coding feature vector relative to the outlet pressure local time series feature heterogeneous propagation representation vector, the change trend and rate of the feature expression of the outlet pressure local time series feature over time can be measured, which can not only reveal the dynamic characteristics of the outlet pressure local time series feature, but also discover abnormal data in the original outlet pressure data. Then, the sequence of the outlet pressure semantic association and the sequence of the outlet pressure semantic space jump value are subjected to information transmission of multiple gate structures. Here, by combining the semantic association and the semantic space jump value, the network can adaptively selectively strengthen or weaken the transmission of specific outlet pressure information according to the relationship between the local temporal features of the outlet pressure, thereby improving the accuracy and robustness of the outlet pressure temporal feature representation. Afterwards, based on the sequence of the outlet pressure information transmission modulation weights, the position-weighted sum of the sequence of the outlet pressure local temporal feature heterogeneous modulation coding feature vectors is calculated to obtain the outlet pressure heterogeneous feature significant propagation representation vector. This weighted summation method comprehensively considers the importance and temporal changes of the outlet pressure local temporal features, and obtains an outlet pressure heterogeneous feature significant propagation representation vector that comprehensively reflects the temporal evolution characteristics of the outlet pressure local temporal features, that is, the obtained propagation representation vector can fully reflect the spatiotemporal characteristics of the original outlet pressure data.

[0054] Specifically, in the embodiment of the present application, the outlet pressure local temporal feature aggregation subunit 1544 is used to: perform residual connection on the outlet pressure heterogeneous feature significant propagation representation vector and the outlet pressure local temporal feature heterogeneous propagation representation vector to obtain the outlet pressure heterogeneous feature significant propagation compensation representation vector. This process can be expressed as: ;in, is the vector representing the propagation of the local temporal characteristics of the pressure at the outlet, is the vector representing the significant propagation of the pressure heterogeneity characteristics at the outlet, is the compensation vector representing the significant propagation of the outlet pressure heterogeneity characteristics.

[0055] The significant propagation compensation representation vector of the outlet pressure heterogeneous feature is mapped into Euclidean space to obtain the significant propagation representation vector of the outlet pressure time series feature. This process can be expressed as: ;in, is the compensation vector representing the significant propagation of the pressure heterogeneity characteristics at the outlet, and are the third weight matrix and the fourth weight matrix respectively, is the Euclidean space mapping operation, is the vector representing the significant propagation of the outlet pressure time series characteristics.

[0056] That is, by introducing deep residual connections, not only can the gradient vanishing problem that may occur in network training be solved, but also the direct transmission of the outlet pressure time series features can be promoted, which is conducive to enhancing the richness and diversity of feature representation. Finally, it is necessary to perform Euclidean space mapping on the outlet pressure heterogeneous feature significant propagation compensation representation vector. In this way, the outlet pressure feature vector after complex processing can be transformed back to the traditional Euclidean space, realizing the effective projection from high-dimensional feature space to low-dimensional space, which is convenient for subsequent data processing.

[0057] In the embodiment of the present application, the outlet pressure time series multi-dimensional representation unit 155 is used to fuse the outlet pressure local time series distribution characteristics and the outlet pressure time series characteristic significant propagation representation to obtain the outlet pressure time series multi-dimensional representation, and based on the outlet pressure time series multi-dimensional representation, obtain a control result, and the control result is used to represent the recommended decoded value of the variable frequency motor speed at the next time point. Specifically, Figure 4 FIG. 1 is a block diagram of a multi-dimensional characterization unit for the pressure sequence at the outlet of a slurry pump automatic water replenishment device according to an embodiment of the present application. Figure 4 As shown, the outlet pressure timing multi-dimensional representation unit 155 includes: an outlet pressure timing multi-dimensional representation generation sub-unit 1551, which is used to use the outlet pressure local timing distribution feature vector as a pressure timing steady-state signal, and use the outlet pressure timing feature significant propagation representation vector as a pressure timing dynamic signal, and fuse the pressure timing steady-state signal and the pressure timing dynamic signal to obtain the outlet pressure timing multi-dimensional representation vector as the outlet pressure timing multi-dimensional representation; a variable frequency motor speed recommendation sub-unit 1552, which is used to obtain the control result based on the outlet pressure timing multi-dimensional representation vector, and the control result is used to represent the recommended decoding value of the variable frequency motor speed at the next time point.

[0058] In the embodiment of the present application, the outlet pressure time series multi-dimensional representation generating subunit 1551 is used to use the outlet pressure local time series distribution feature vector as the pressure time series steady-state signal, and use the outlet pressure time series feature significant propagation representation vector as the pressure time series dynamic signal, and fuse the pressure time series steady-state signal and the pressure time series dynamic signal to obtain the outlet pressure time series multi-dimensional representation vector as the outlet pressure time series multi-dimensional representation. The above process can be expressed by the formula: ;in, is the local time series distribution characteristic vector of the outlet pressure, is the significant propagation vector of the outlet pressure time series characteristics, yes The transposed vector of yes Length, is matrix multiplication, is the last outlet pressure local time series feature vector of the sequence of outlet pressure local time series feature vectors, yes function, It is the multi-dimensional characterization vector of the outlet pressure time series.

[0059] It should be understood that the local time-series distribution characteristic vector of the outlet pressure mainly reflects the average performance or stable state of the outlet of the system over a longer time range. The outlet pressure time-series characteristic significant propagation representation vector is the key and significant pressure feature obtained by dynamically propagating and aggregating the pressure features at each time point. Therefore, a more comprehensive and detailed description of the outlet pressure state is obtained, so that future behavior can be predicted more accurately and a more precise control strategy can be formulated accordingly. In the technical solution of the present application, the local time-series distribution characteristic vector of the outlet pressure is used as a pressure time-series steady-state signal, and the outlet pressure time-series characteristic significant propagation representation vector is used as a pressure time-series dynamic signal. The pressure time-series steady-state signal and the pressure time-series dynamic signal are fused to obtain a multi-dimensional representation vector of the outlet pressure time-series. In particular, here, by performing timing interaction of the converter structure on the features of the last position of the sequence of the pressure time series steady-state signal, the pressure time series dynamic signal and the local time series feature vector of the outlet pressure, the sensitivity to the pressure data at the current time point can be maintained while the pressure trend and time series dynamic change law within a longer time span can be effectively captured, thereby enhancing the overall prediction performance.

[0060] In the embodiment of the present application, the variable frequency motor speed recommendation subunit 1552 is used to obtain the control result based on the outlet pressure time series multidimensional characterization vector, and the control result is used to represent the recommended decoded value of the variable frequency motor speed at the next time point. Specifically, in the embodiment of the present application, the variable frequency motor speed recommendation subunit is used to: pass the outlet pressure time series multidimensional characterization vector through the decoder-based water replenishment controller to obtain the control result, and the control result is used to represent the recommended decoded value of the variable frequency motor speed at the next time point. It should be understood that the outlet pressure time series multidimensional characterization vector contains rich pressure information, but this information is in a highly abstract and comprehensive form. In order to convert the abstract pressure characteristics into specific control instructions so as to be directly applied to the control of the variable frequency motor, it is necessary to input the outlet pressure time series multidimensional characterization vector into the decoder-based water replenishment controller. The decoder is a neural network model, which can accurately output a suitable motor speed recommendation value according to the input pressure characteristics through a large amount of data training. After obtaining this recommended value, the motor can adjust the speed accordingly according to this value, thereby changing the working state of the mud pump. In this way, the mud pump rehydration process can be made more efficient and reasonable, avoiding the situation of over-rehydration or under-rehydration. In particular, in a specific embodiment of the present application, a control result of passing the outlet pressure timing multidimensional characterization vector through a decoder-based water replenishment controller to obtain a recommended decoded value for representing the speed of the variable frequency motor at the next time point can be implemented as follows: multiplying the decoder weight matrix of the decoder with the outlet pressure timing multidimensional characterization vector to obtain a decoded outlet pressure timing multidimensional global characterization vector, and accumulating and summing all eigenvalues ​​of the decoded outlet pressure timing multidimensional global characterization vector to obtain the recommended decoded value of the speed of the variable frequency motor.

[0061] In particular, when the local time series distribution feature vector of the outlet pressure and the significant propagation representation vector of the outlet pressure time series feature respectively represent the full time domain local time domain correlation set features and the cross-local time domain time series dynamic propagation features of the outlet pressure data, after the fusion of the time series steady-state signal and the time series dynamic signal, the outlet pressure time series multi-dimensional representation vector will also have significant fusion distribution space structure differences due to the differences in the distribution of local time domain time series features, affecting the convergence consistency of the decoder and thus affecting the accuracy of the control results obtained by the water replenishment controller.

[0062] Therefore, in one example, the outlet pressure time series multidimensional characterization vector is optimized, and the optimization process includes: calculating the absolute value sum of all eigenvalues ​​of the outlet pressure time series multidimensional characterization vector and the square root of the square sum to obtain the first outlet pressure time series multidimensional space structure value and the second outlet pressure time series multidimensional space structure value, and the process can be expressed as: ;in, The first multidimensional representation vector of the outlet pressure time series The eigenvalues ​​at the positions, Represents the multidimensional spatial structure value of the first outlet pressure time series, Represents the multidimensional spatial structure value of the pressure time series at the second outlet.

[0063] Determine the number of eigenvalues ​​of all eigenvalues ​​of the outlet pressure time series multidimensional representation vector, that is, the length of the outlet pressure time series multidimensional representation vector .

[0064] For each eigenvalue of the outlet pressure time series multidimensional characterization vector, the multidimensional spatial structure value of the first outlet pressure time series minus the product of the eigenvalue and the number of eigenvalues ​​is calculated to obtain the multidimensional long-range dependency value of the first outlet pressure time series. This process can be expressed as: ;in, The first multidimensional representation vector of the outlet pressure time series The eigenvalues ​​at the positions, Represents the multidimensional spatial structure value of the first outlet pressure time series, represents the length of the multi-dimensional representation vector of the outlet pressure time series, Represents the multidimensional long-range dependence value of the pressure time series at the first outlet.

[0065] The multidimensional long-range dependency value of the second outlet pressure time series is obtained by calculating the square root of the number of eigenvalues ​​multiplied by the product of the eigenvalues ​​minus the multidimensional spatial structure value of the second outlet pressure time series. The process can be expressed as: ;in, The first multidimensional representation vector of the outlet pressure time series The eigenvalues ​​at the positions, represents the multidimensional spatial structure value of the pressure time series at the second outlet, represents the length of the multi-dimensional representation vector of the outlet pressure time series, Represents the multi-dimensional long-range dependence value of the pressure time series at the second outlet.

[0066] The exponential value calculated by taking the first outlet pressure time series multidimensional long-range dependence value as the exponent of the natural constant and the inverse of the second outlet pressure time series multidimensional long-range dependence value are weighted summed to obtain the optimized eigenvalue corresponding to each eigenvalue. The process can be expressed as: ;in, represents the multidimensional long-range dependence of the pressure time series at the first outlet, represents the multi-dimensional long-range dependence of the pressure time series at the second outlet, and represents the weighted hyperparameter, The first multidimensional representation vector of the outlet pressure time series The optimized eigenvalues ​​are combined into an optimized outlet pressure time series multi-dimensional representation vector.

[0067] That is, in view of the possible spatial structure loss of the feature set of the outlet pressure time series multidimensional characterization vector in the high-dimensional space, which causes the decoder weight to implicitly infer the spatial structure information based on the features, resulting in inconsistent convergence. A long-distance feature dependency relationship is established based on the spatial structure representation of the outlet pressure time series multidimensional characterization vector based on the overall feature scale of the outlet pressure time series multidimensional characterization vector, so as to establish the feature local connectivity of the outlet pressure time series multidimensional characterization vector, and capture the spatial ambiguous information of the object feature value through the unstructured feature value point prediction of the outlet pressure time series multidimensional characterization vector, thereby improving the spatial inductive deviation perception ability of the feature set of the outlet pressure time series multidimensional characterization vector, improving the convergence consistency of the decoder, and improving the accuracy of the control result obtained by the outlet pressure time series multidimensional characterization vector through the water replenishment controller based on the decoder.

[0068] In summary, the automatic rehydration device 100 for the mud pump water supply based on the embodiment of the present application is explained, which uses artificial intelligence-based data encoding and processing technology to extract local time series features of the time series set of the mud pump outlet pressure data, and then performs time series association on each outlet pressure local time series feature, and at the same time, performs multiple gate structure propagation on each outlet pressure local time series feature, so as to adaptively control the variable frequency motor speed value at the next time point according to the fusion representation between the significant propagation representation of the outlet pressure time series feature and the local time series distribution feature of the outlet pressure after time series association. In this way, the working parameters of the variable frequency rehydration pump can be adjusted in time through the feedback of the change of the mud pump outlet pressure, ensuring timely and appropriate rehydration, which is beneficial to improving the working efficiency and stability of the mud pump.

[0069] In other embodiments of the present application, a mud pump water automatic liquid replenishing device is also provided. Figure 5 FIG. 1 is an operation flow chart of the automatic liquid replenishing device for the mud pump according to other embodiments of the present application. Figure 5As shown, the mud starts from the mud tank and enters the mud pump through two paths. One path is to directly suck the mud from the mud tank into the mud pump; the other path is that the mud first enters the variable frequency pump from the mud tank, and then the variable frequency pump supplies the mud pump. The mud pump discharges the mud, and at the same time there is a pressure transmitter to monitor the pressure of the mud pump and feed back the pressure signal to the PLC master control. The PLC master control can control the start and stop of the variable frequency pump, and receive the mud pump parameter input from the touch screen of the driver's control room, and then adjust the operation of the entire mud pump water filling automatic liquid device according to these parameters and pressure feedback. In this way, the entire device can realize the control of the mud pump and the transportation of mud.

[0070] In an embodiment of the present application, a method for automatically replenishing water in a mud pump is also provided, which is used to execute the automatic replenishing water in a mud pump as described above. Figure 6 Flow chart of the automatic liquid replenishment method for a mud pump according to an embodiment of the present application. Figure 6 As shown, in the automatic water replenishment method for the mud pump, it includes: S110, obtaining a time series set of outlet pressure data collected by the pressure sensor; S120, performing local time series feature extraction on the time series set of the outlet pressure data to obtain a sequence of local time series features of the outlet pressure; S130, performing time series association on the sequence of local time series features of the outlet pressure to obtain a local time series distribution feature of the outlet pressure; S140, performing feature significant propagation of multiple gate structures on the sequence of local time series features of the outlet pressure to obtain a significant propagation representation of the outlet pressure time series feature; S150, fusing the local time series distribution feature of the outlet pressure and the significant propagation representation of the outlet pressure time series feature to obtain a multi-dimensional representation of the outlet pressure time series, and based on the multi-dimensional representation of the outlet pressure time series, obtaining a control result, and the control result is used to represent the recommended decoding value of the speed of the variable frequency motor at the next time point.

[0071] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned method for automatically replenishing water in the mud pump have been described in detail above. Figures 1 to 4 The description of the automatic water replenishing device for the mud pump has been described in detail, and therefore, its repeated description will be omitted.

[0072] In summary, the automatic refilling method for the mud pump based on the embodiment of the present application is explained, which uses artificial intelligence-based data encoding and processing technology to extract local time series features of the time series set of the mud pump outlet pressure data, and then performs time series association on each outlet pressure local time series feature, and at the same time, performs multiple gate structure propagation on each outlet pressure local time series feature, so as to adaptively control the variable frequency motor speed value at the next time point according to the fusion representation between the significant propagation representation of the outlet pressure time series feature and the local time series distribution feature of the outlet pressure after time series association. In this way, the working parameters of the variable frequency refilling pump can be adjusted in time through the feedback of the change of the mud pump outlet pressure, ensuring timely and appropriate refilling, which is beneficial to improving the working efficiency and stability of the mud pump.

Claims

1. A mud pump water automatic liquid replenishing device, characterized in that: include: Mud tank; A mud pump for sucking, pressurizing and discharging the mud in the mud tank; A pressure sensor deployed at the outlet of the mud pump, used to collect the outlet pressure data of the mud pump in real time; A variable frequency fluid replenishment pump disposed between the mud tank and the mud pump, used for replenishing water to the mud pump; A central processing unit, used to control the speed of the variable frequency motor of the variable frequency fluid replenishment pump based on the outlet pressure data of the mud pump; wherein the central processing unit includes: a pressure data acquisition unit, used to acquire the time series set of the outlet pressure data collected by the pressure sensor; a pressure time series extraction unit, used to extract local time series features of the time series set of the outlet pressure data to obtain a sequence of local time series features of the outlet pressure; a pressure data time series association unit, used to perform time series association on the sequence of local time series features of the outlet pressure to obtain a local time series distribution feature of the outlet pressure; a pressure time series propagation unit, used to perform feature significant propagation of the sequence of local time series features of the outlet pressure with a multiple gate structure to obtain a significant propagation representation of the outlet pressure time series feature; an outlet pressure time series multi-dimensional representation unit, used to fuse the local time series distribution feature of the outlet pressure and the significant propagation representation of the outlet pressure time series feature to obtain a multi-dimensional representation of the outlet pressure time series, and based on the multi-dimensional representation of the outlet pressure time series, obtain a control result, and the control result is used to represent the recommended decoding value of the speed of the variable frequency motor at the next time point.

2. The automatic liquid replenishing device for mud pump water supply according to claim 1 is characterized in that: The pressure time series extraction unit is used to: input the time series set of the outlet pressure data into a pressure time series feature extractor based on a one-dimensional causal convolutional neural network model to obtain a sequence of local time series feature vectors of the outlet pressure as a sequence of local time series features of the outlet pressure.

3. The automatic liquid replenishing device for mud pump water supply according to claim 2 is characterized in that: The pressure data time series association unit is used to: cascade each outlet pressure local time series feature vector in the sequence of the outlet pressure local time series feature vector to obtain an outlet pressure local time series distribution feature vector as the outlet pressure local time series distribution feature.

4. The automatic liquid replenishing device for a mud pump according to claim 3, characterized in that: The pressure time series propagation unit includes: an outlet pressure local time series feature space mapping subunit, used for mapping each outlet pressure local time series feature vector in the sequence of the outlet pressure local time series feature vectors to the Poincare space to obtain a sequence of outlet pressure local time series feature heterogeneous modulation coding feature vectors; an outlet pressure local time series feature heterogeneous propagation representation generating subunit, used for forward LSTM sequence encoding the sequence of the outlet pressure local time series feature heterogeneous modulation coding feature vectors to obtain an outlet pressure local time series feature heterogeneous propagation representation vector; the outlet pressure The local temporal significant propagation subunit is used to perform significant propagation on the sequence of the outlet end pressure local temporal feature alien modulation coding feature vectors based on the outlet end pressure local temporal feature alien propagation representation vector to obtain the outlet end pressure alien feature significant propagation representation vector; the outlet end pressure local temporal feature aggregation subunit is used to perform residual mapping processing on the outlet end pressure alien feature significant propagation representation vector and the outlet end pressure local temporal feature alien propagation representation vector to obtain the outlet end pressure temporal feature significant propagation representation vector as the outlet end pressure temporal feature significant propagation representation.

5. The automatic liquid replenishing device for a mud pump according to claim 4, characterized in that: The outlet pressure local temporal significant propagation subunit is used to: calculate the semantic association between each outlet pressure local temporal feature foreign-domain modulation coding feature vector in the sequence of the outlet pressure local temporal feature foreign-domain modulation coding feature vector and the outlet pressure local temporal feature foreign-domain propagation representation vector to obtain a sequence of outlet pressure semantic associations; calculate the semantic space jump value of each outlet pressure local temporal feature foreign-domain modulation coding feature vector in the sequence of the outlet pressure local temporal feature foreign-domain modulation coding feature vector relative to the outlet pressure local temporal feature foreign-domain propagation representation vector to obtain a sequence of outlet pressure semantic space jump values; perform information transfer of multiple gate structures on the sequence of outlet pressure semantic associations and the sequence of outlet pressure semantic space jump values ​​to obtain a sequence of outlet pressure information transfer modulation weights; and calculate the position-weighted sum of the sequence of the outlet pressure local temporal feature foreign-domain modulation coding feature vectors based on the sequence of outlet pressure information transfer modulation weights to obtain the outlet pressure foreign-domain feature significant propagation representation vector.

6. The automatic liquid replenishing device for mud pump water supply according to claim 5, characterized in that: The outlet pressure local temporal feature aggregation subunit is used to: perform residual connection on the outlet pressure exotic feature significant propagation representation vector and the outlet pressure local temporal feature exotic propagation representation vector to obtain the outlet pressure exotic feature significant propagation compensation representation vector; perform Euclidean space mapping on the outlet pressure exotic feature significant propagation compensation representation vector to obtain the outlet pressure temporal feature significant propagation representation vector.

7. The automatic liquid replenishing device for a mud pump according to claim 6, characterized in that: The outlet pressure timing multi-dimensional representation unit includes: an outlet pressure timing multi-dimensional representation generation sub-unit, which is used to use the outlet pressure local timing distribution feature vector as a pressure timing steady-state signal, and use the outlet pressure timing feature significant propagation representation vector as a pressure timing dynamic signal, and fuse the pressure timing steady-state signal and the pressure timing dynamic signal to obtain the outlet pressure timing multi-dimensional representation vector as the outlet pressure timing multi-dimensional representation; a variable frequency motor speed recommendation sub-unit, which is used to obtain the control result based on the outlet pressure timing multi-dimensional representation vector, and the control result is used to represent the recommended decoding value of the variable frequency motor speed at the next time point.

8. The automatic liquid replenishing device for a mud pump according to claim 7, characterized in that: The variable frequency motor speed recommendation subunit is used to: pass the outlet pressure time series multi-dimensional characterization vector through the decoder-based water replenishment controller to obtain the control result, and the control result is used to represent the recommended decoded value of the variable frequency motor speed at the next time point.

9. A method for automatically replenishing water in a mud pump, used to implement the automatic replenishing water in a mud pump as claimed in claim 1, characterized in that: The automatic replenishment method for filling water on a mud pump includes: acquiring a time series set of outlet pressure data collected by the pressure sensor; performing local time series feature extraction on the time series set of outlet pressure data to obtain a sequence of local time series features of outlet pressure; performing time series correlation on the sequence of local time series features of outlet pressure to obtain local time series distribution features of outlet pressure; performing feature significant propagation of multiple gate structures on the sequence of local time series features of outlet pressure to obtain a significant propagation representation of outlet pressure time series features; fusing the local time series distribution features of outlet pressure and the significant propagation representation of outlet pressure time series features to obtain a multi-dimensional representation of outlet pressure time series, and obtaining a control result based on the multi-dimensional representation of outlet pressure time series, wherein the control result is used to represent a recommended decoding value of the variable frequency motor speed at the next time point.

10. The automatic liquid replenishing method for a mud pump according to claim 9, characterized in that: Local time series feature extraction is performed on the time series set of the outlet pressure data to obtain a sequence of local time series features of the outlet pressure, including: inputting the time series set of the outlet pressure data into a pressure time series feature extractor based on a one-dimensional causal convolutional neural network model to obtain a sequence of local time series feature vectors of the outlet pressure as the sequence of local time series features of the outlet pressure.

Images