Intelligent data acquisition system of special pump

Abstract

The invention discloses an intelligent data acquisition system of a special pump, and relates to the technical field of equipment management. Comprising a data acquisition module, a data preprocessing module, a data storage module, a data analysis and prediction module, an intelligent control module, a remote monitoring module, a maintenance and optimization module, a multi-source data fusion module, a digital twinborn simulation module and a collaboration module. According to the intelligent data acquisition system of the special pump, by arranging multi-dimensional data acquisition and an intelligent optimization model based on the self-adaptive neural network and reinforcement learning, the defects that in an existing system, data integration is not comprehensive, a maintenance decision is lagged and intelligent optimization is lacked are overcome, and real-time monitoring and dynamic adjustment of the equipment state are achieved. The system can accurately predict the equipment fault, optimize the maintenance strategy, significantly reduce the operation cost, and improve the operation efficiency and reliability of the equipment.

Description

Technical Field

[0001] The invention relates to the technical field of equipment management, and in particular to an intelligent data acquisition system for a special pump. Background Art

[0002] With the continuous advancement of intelligent industrial equipment, special pumps, as key equipment, are widely used in the fields of petroleum, chemical industry, energy, etc. However, existing data acquisition systems usually rely on a single type of sensor data and lack comprehensive monitoring and analysis of multi-dimensional information, making it difficult to accurately evaluate the operating status of the equipment. In addition, traditional maintenance decisions are mostly based on fixed cycles or empirical rules, lack dynamic adaptive capabilities, and are difficult to cope with complex and changing working conditions.

[0003] Therefore, the existing systems have many deficiencies in fault prediction and maintenance optimization, mainly including the following aspects: Existing technologies often rely on single-dimensional data, such as vibration or temperature, which leads to a one-sided assessment of the health status of the equipment and fails to fully reflect the complex operating characteristics of the pump.

[0004] Traditional maintenance methods based on experience or fixed cycles cannot adapt to real-time changes in operating conditions, and can easily lead to excessive or untimely maintenance of equipment, increasing operating costs and downtime risks.

[0005] The existing system lacks intelligent learning mechanisms in decision-making optimization and is unable to dynamically adjust strategies based on real-time and historical data, making it difficult to improve long-term operational efficiency and reliability.

[0006] To solve these problems, we introduced adaptive neural network models and reinforcement learning models, which enable the system to optimize equipment operation strategies based on real-time data and business scenarios. Through dynamic optimization and continuous learning, the system can not only provide accurate equipment health assessment and maintenance recommendations, but also significantly improve the operating efficiency of the pump and reduce the occurrence of failures. Summary of the invention

[0007] In view of the deficiencies in the prior art, the present invention provides an intelligent data acquisition system for a special pump to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions: an intelligent data acquisition system for special pumps, comprising a data acquisition module, a data preprocessing module, a data storage module, a data analysis and prediction module, an intelligent control module, a remote monitoring module, a maintenance and optimization module, a multi-source data fusion module and a digital twin simulation module; The data acquisition module collects data on the operating status of special pumps in real time based on different types of sensors; The data preprocessing module preprocesses the raw data collected from the sensor and removes noise and interference in the data based on the filtering algorithm to improve the signal-to-noise ratio of the data; The data storage module stores and manages the pre-processed data, adopts a distributed database architecture, supports the storage and rapid retrieval of large-scale data, and has a multi-level data backup and redundancy mechanism to prevent data loss and damage; The data analysis and prediction module uses an adaptive neural network model to perform real-time analysis on the operating data of special pumps, dynamically adjusts model parameters based on the pump's operating history data, performs status monitoring and fault prediction, and identifies potential fault modes by analyzing the changing trends of key parameters such as flow, vibration, temperature, and pressure, and issues early warning signals; Intelligent control module, which automatically adjusts the operating status of the pump based on real-time analysis results, automatically adjusts the speed, pressure and flow of the pump according to the analysis data, and optimizes the operating parameters of the pump in real time according to the external environment and the workload of the pump to ensure that the pump works in the best condition; Remote monitoring module, which can check the operation status and data of special pumps in real time through the Internet, provide a visual monitoring interface, access the system in real time at different locations based on the remote access function, diagnose and analyze the operation status of the pump, and support remote operation, allowing remote adjustment of the pump's operating parameters or emergency shutdown operations when necessary; The maintenance and optimization module automatically generates customized maintenance plans and operation optimization suggestions based on real-time and historical data. It uses a reinforcement learning model to dynamically calculate the optimal maintenance cycle of the pump and adjust the maintenance plan according to the actual operating status of the equipment. It also provides targeted operation optimization suggestions based on different working conditions and environmental conditions. Multi-source data fusion module integrates sensor data from inside the pump body and external environment data to conduct multi-dimensional data monitoring and analysis; The digital twin simulation module uses digital twin technology to replicate the operating status of special pumps in real time in virtual space. By establishing a digital model of the pump, real-time simulation is performed in a virtual environment to test the impact of different operating conditions on the pump.

[0009] To further optimize the technical solution, in the data acquisition module, sensors are installed and integrated in various parts of the special pump for monitoring the operating status, so as to monitor the operating status of the pump in real time; Sensor types include temperature sensors, pressure sensors, flow sensors, vibration sensors and speed sensors. The sensors transmit data to the data acquisition module via wired or wireless means and support multiple communication protocols and transmission methods.

[0010] Further optimizing the technical solution, in the data analysis and prediction module, the adaptive neural network model ANN is based on the multi-dimensional data collected during the operation of the special pump, and the model weight is dynamically adjusted to optimize the fault prediction and equipment health status assessment; When the data distribution of the pump changes under different operating conditions, the weight of the neural network is adaptively adjusted in real time; Set the structure of the neural network to: Input layer: contains Sensor data input, indicating different operating status parameters; Hidden layer: contains multiple layers of hidden neurons and uses adaptive activation functions; Output layer: predict the health status of the pump and failure probability ; Inputting features into a neural network: vibration data , temperature data , pressure data , traffic data , forming the input vector: ; Output predictions include: Health status prediction: Indicates that the pump is at time The health status of the device is in the range of [0,1], where 0 means the device is seriously damaged and 1 means the device is in good condition; Failure probability prediction: is the failure probability, which indicates the probability that the equipment will fail in the short term.

[0011] Further optimizing the technical solution, the adaptive neural network model ANN dynamically adjusts the model weights based on the dynamic weight adjustment mechanism, assuming that the model adaptively adjusts the weights according to the changes in the input data , the dynamic weight adjustment mechanism is as follows: in, Representing neurons arrive The weight between The value of is the learning rate; For the current moment The loss function of It is a dynamic item of weight change, which is adjusted in real time according to the operating status of the pump: when the health status or working condition of the pump fluctuates greatly, the model needs a higher weight adjustment speed to adapt to the fluctuation. By calculating the gradient change of each sensor data, the comprehensive gradient change of the sensor data is integrated based on the weighted average algorithm. This volatility is obtained through the comprehensive gradient change, and the dynamic item is adjusted accordingly. is the adaptive factor, the adaptive factor The adjustment of is controlled by: ; in, It is the comprehensive gradient change of sensor data, which is used to reflect the volatility of the system's working conditions. If the sensor data fluctuates greatly, The corresponding increase is used to speed up the weight adjustment of the model; is the adjustment coefficient, controlling the fluctuation of health status the extent of the impact; It is the change in health status, indicating the difference between the current health status and the previous moment.

[0012] To further optimize the technical solution, in the adaptive neural network model ANN, in order to optimize the prediction of health status and failure probability, the loss function is designed into two parts: Health status prediction loss: Use mean square error MSE to measure the gap between the predicted health status and the actual status: in, Indicates the number of samples, used to average the errors of all samples; Indicates The index of the samples, starting from Sample starts at samples; is the actual health status; is the health state predicted by the model; Failure probability prediction loss: The cross entropy loss function is used to measure the gap between the predicted failure probability and the actual failure: ; in, Indicates The true failure probability of a sample is 0 or 1, indicating a normal or faulty state; The model predicts the The failure probability of a sample is a continuous value in the interval [0, 1], indicating the possibility of failure. The total loss function is: ; in, and is a weight parameter used to adjust the importance of health status and fault prediction; An adaptive activation function is set to improve the generalization ability of the model under different operating conditions. The adaptive activation function is as follows: ; in, is a dynamically changing adaptive parameter that is adjusted according to the volatility of the input data: ; in, is the initial steepness parameter of the adaptive activation function, which indicates the basic steepness when the input data does not change significantly; It is a regulating factor, which means that when the input data change rate is large, the steepness of the activation function increases, which is used to improve the response speed of the network; is the input vector in the neural network.

[0013] To further optimize the technical solution, in the maintenance and optimization module, the reinforcement learning model is constructed by a Markov decision process MDP, including: State Space : The operating status set of a special pump includes a combination of different sensor data, such as ; Action Space : A collection of maintenance decisions or optimization suggestions taken under different conditions, such as adjusting the speed of the pump, replacing parts, starting or stopping the pump; Reward Function : When performing an action The immediate feedback or reward after the decision reflects the positive or negative impact of the decision on the pump operation; Strategy : Given state Next, select Action The probability distribution of Discount Factor : The discount rate for future rewards, which represents the relationship between long-term benefits and current benefits, ranging from ; And use an optimization algorithm to optimize the decision-making process of the model. The algorithm estimates the long-term value of each state-action pair by updating the Q-value function. The update formula of the Q-value function is as follows: in, is the learning rate, which controls the step size of the update; An immediate reward for the current moment; is the discount factor, which indicates the weight of future rewards; Indicates that in the next state Next, choose the best action to bring the greatest value; In the pump maintenance optimization scenario, the reward function The design is dynamically adjusted according to the health status, maintenance cost and operating efficiency of the pump. Assume that the reward function is as follows: in, Is to perform an action the maintenance costs required; It is the health status of the pump. The higher the value, the better the pump status. is the operating energy consumption of the pump, the higher the value, the greater the energy consumption; and are weight parameters that control the impact of health status and energy consumption respectively.

[0014] To further optimize the technical solution, in the reinforcement learning model, an adaptive exploration strategy is set to optimize the trade-off between exploration and utilization of the algorithm. Greedy strategy control, that is, based on probability Randomly select actions to The probability of selecting the current optimal action is based on the dynamic exploration rate , the formula is as follows: in, is the initial exploration rate; is the exploration rate attenuation factor, which controls the speed at which the exploration rate decreases over time; Over time, the model gradually shifts from exploration mode to exploitation mode, where it takes full advantage of the learned optimal policy.

[0015] To further optimize the technical solution, the multi-source data fusion module adopts a dynamic weight adjustment algorithm to dynamically adjust the weights of different data sources according to the real-time operating status of the pump and changes in the external environment; By fusing multi-source data, the operating status of the pump is evaluated to improve the accuracy of fault prediction and operation optimization.

[0016] To further optimize the technical solution, the digital twin simulation module uses finite element analysis to simulate the mechanical properties and fluid movement of the pump; By combining real-time data with historical data, the digital twin dynamically reflects the actual operating status of the pump and pre-evaluates the operating effects under extreme conditions or high loads.

[0017] To further optimize the technical solution, the system also includes a collaborative module that supports multiple communication protocols and interface standards to meet the integration requirements of different devices and systems; Through data routing and protocol conversion, seamless connection and data sharing are achieved between different modules, and docking with external systems is supported.

[0018] Compared with the prior art, the present invention provides an intelligent data acquisition system for special pumps, which has the following beneficial effects: The intelligent data acquisition system of this special pump solves the defects of incomplete data integration, delayed maintenance decision-making and lack of intelligent optimization in the existing system by setting up multi-dimensional data acquisition and intelligent optimization model based on adaptive neural network and reinforcement learning, and realizes real-time monitoring and dynamic adjustment of equipment status. The system can accurately predict equipment failures, optimize maintenance strategies, significantly reduce operating costs, and improve equipment operation efficiency and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a structural schematic diagram of an intelligent data acquisition system for a special pump proposed by the present invention; Figure 2 A schematic diagram of the process flow of a collaborative module in an intelligent data acquisition system for a special pump proposed by the present invention; Figure 3 A schematic diagram of the process of an adaptive neural network model in an intelligent data acquisition system for a special pump proposed by the present invention; Figure 4 This is a flow chart of a reinforcement learning model in an intelligent data acquisition system for a special pump proposed in the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Embodiment 1: See also Figure 1 , an intelligent data acquisition system for special pumps, including a data acquisition module, a data preprocessing module, a data storage module, a data analysis and prediction module, an intelligent control module, a remote monitoring module, a maintenance and optimization module, a multi-source data fusion module and a digital twin simulation module.

[0022] The data acquisition module collects data on the operating status of special pumps in real time based on different types of sensors.

[0023] In this embodiment, in the data acquisition module, sensors are installed and integrated in various parts of the special pump for monitoring the operating status, and the operating status of the pump is monitored in real time; Sensor types include temperature sensors, pressure sensors, flow sensors, vibration sensors and speed sensors. The sensors transmit data to the data acquisition module via wired or wireless means and support multiple communication protocols and transmission methods.

[0024] The sensor layout has been precisely designed to ensure efficient and accurate data collection in complex pump structures. In order to adapt to different working environments, the module also has a self-diagnosis function that can detect the working status of the sensor in real time, detect and alarm sensor failures in a timely manner, and ensure the continuity and accuracy of data collection.

[0025] The data preprocessing module preprocesses the raw data collected from the sensor and removes noise and interference in the data based on the filtering algorithm to improve the signal-to-noise ratio of the data.

[0026] In this embodiment, outliers are identified and removed through an anomaly detection algorithm to ensure the validity of the data. The preprocessing steps include data format conversion, time series alignment, and missing value filling to ensure the consistency and integrity of data from different sensors. The module also supports real-time data stream processing, which can perform preprocessing while collecting data, reduce data transmission delays, and provide a high-quality data foundation for subsequent data analysis.

[0027] The data storage module stores and manages the pre-processed data. It adopts a distributed database architecture to support the storage and rapid retrieval of large-scale data, ensuring that data storage and management are efficient even in a multi-pump system. It supports multiple data formats, including time series data, log files, and structured databases to meet different analysis needs. It also sets up a multi-level data backup and redundancy mechanism to prevent data loss and damage.

[0028] The data analysis and prediction module uses an adaptive neural network model to perform real-time analysis on the operating data of special pumps, dynamically adjust model parameters based on the pump's historical operating data, perform status monitoring and fault prediction, identify potential failure modes by analyzing the changing trends of key parameters such as flow, vibration, temperature, and pressure, and issue early warning signals.

[0029] In this embodiment, in the data analysis and prediction module, the adaptive neural network model ANN optimizes fault prediction and equipment health status assessment by dynamically adjusting model weights based on multi-dimensional data collected during the operation of the special pump; When the data distribution of the pump changes under different operating conditions, the weight of the neural network is adaptively adjusted in real time; Set the structure of the neural network to: Input layer: contains Sensor data input, indicating different operating status parameters; Hidden layer: contains multiple layers of hidden neurons and uses adaptive activation functions; Output layer: predict the health status of the pump and failure probability ; Inputting features into a neural network: vibration data , temperature data , pressure data , traffic data , forming the input vector: Output predictions include: Health status prediction: Indicates that the pump is at time The health status of the device is in the range of [0,1], where 0 means the device is seriously damaged and 1 means the device is in good condition; Failure probability prediction: is the failure probability, which indicates the probability that the equipment will fail in the short term.

[0030] The adaptive neural network model ANN dynamically adjusts the model weights based on the dynamic weight adjustment mechanism, assuming that the model adaptively adjusts the weights according to the changes in the input data. , the dynamic weight adjustment mechanism is as follows: in, Representing neurons arrive The weight between The value of is the learning rate; For the current moment The loss function of It is a dynamic item of weight change, which is adjusted in real time according to the operating status of the pump: when the health status or working condition of the pump fluctuates greatly, the model needs a higher weight adjustment speed to adapt to the fluctuation. By calculating the gradient change of each sensor data, the comprehensive gradient change of the sensor data is integrated based on the weighted average algorithm. This volatility is obtained through the comprehensive gradient change, and the dynamic item is adjusted accordingly. is the adaptive factor, the adaptive factor The adjustment of is controlled by: ; in, It is the comprehensive gradient change of sensor data, which is used to reflect the volatility of the system's working conditions. If the sensor data fluctuates greatly, The corresponding increase is used to speed up the weight adjustment of the model; is the adjustment coefficient, controlling the fluctuation of health status the extent of the impact; It is the change in health status, indicating the difference between the current health status and the previous moment.

[0031] Furthermore, in the adaptive neural network model ANN, in order to optimize the prediction of health status and failure probability, the loss function is designed to be divided into two parts: Health status prediction loss: Use mean square error MSE to measure the gap between the predicted health status and the actual status: in, Indicates the number of samples, used to average the errors of all samples; Indicates The index of the sample, starting from Sample starts at samples; is the actual health status; is the health state predicted by the model; Failure probability prediction loss: The cross entropy loss function is used to measure the gap between the predicted failure probability and the actual failure: in, Indicates The true failure probability of a sample is 0 or 1, indicating a normal or faulty state; The model predicts the The failure probability of a sample is a continuous value in the interval [0, 1], indicating the possibility of failure. The total loss function is: in, and is a weight parameter used to adjust the importance of health status and fault prediction; An adaptive activation function is set to improve the generalization ability of the model under different operating conditions. The adaptive activation function is as follows: in, is a dynamically changing adaptive parameter that is adjusted according to the volatility of the input data: in, is the initial steepness parameter of the adaptive activation function, which indicates the basic steepness when the input data does not change significantly; It is a regulating factor, which means that when the input data change rate is large, the steepness of the activation function increases, which is used to improve the response speed of the network; is the input vector in the neural network.

[0032] like Figure 3 As shown, the model, when used, includes: Data input: The real-time operation data of the pump (flow, vibration, temperature, pressure, etc.) is input into the adaptive neural network model through the sensor integration module to form an input vector .

[0033] Dynamic adjustment: The neural network adaptively adjusts weights according to the volatility of the input data and the parameters of the activation function , ensuring that the model can respond quickly to changes in device status.

[0034] Health status prediction: The model outputs the health status prediction of the pump ,By comparing with historical data, the system can determine whether the current state of the ,equipment is operating normally, has a minor fault, or is about to have a serious fault.

[0035] Failure probability prediction: The model also outputs the failure probability , which predicts the likelihood of equipment failure within a certain period of time in the future based on historical data and real-time input data.

[0036] Early warning mechanism: Once If the set threshold is exceeded, the system will automatically trigger a warning signal to notify maintenance personnel to conduct inspections or repairs to avoid sudden equipment shutdowns without warning.

[0037] Self-learning and optimization: The model learns from the loss function in each iteration. The model automatically adjusts the weights and activation function parameters based on the feedback from the system, gradually optimizing the fault prediction accuracy. At the same time, the model continuously learns from historical data to improve its ability to predict the status of pumps under different working conditions.

[0038] This enables it to adjust adaptively based on the real-time operation data of the pump, thereby improving the accuracy and response speed of fault prediction. Through this model, the system can not only accurately evaluate the current status of the equipment, but also predict the risk of failure in advance, optimize equipment maintenance strategies, reduce downtime, and improve the reliability and life of special pumps.

[0039] The intelligent control module automatically adjusts the operating status of the pump based on the real-time analysis results. It automatically adjusts the speed, pressure and flow of the pump according to the analysis data. It optimizes the operating parameters of the pump in real time according to the external environment and the pump's workload to ensure that the pump works in the best condition.

[0040] In this embodiment, the module can adjust the speed, pressure and flow of the pump in real time to adapt to different workloads and environmental conditions. For example, when the system detects that the vibration of the pump exceeds the safety threshold, the intelligent control module will automatically reduce the speed, reduce mechanical stress and prevent equipment damage. The module also has self-learning capabilities and can optimize the control strategy based on historical operating data and operational feedback to improve the operating efficiency and stability of the pump. In addition, the intelligent control module supports multiple control modes, such as manual control, automatic control and hybrid control, to meet different operating requirements and improve the flexibility and adaptability of the system.

[0041] The remote monitoring module can view the operating status and data of special pumps in real time through the Internet, provide a visual monitoring interface, and access the system in real time at different locations based on the remote access function to diagnose and analyze the operating status of the pump. It also supports remote operation, allowing remote adjustment of the pump's operating parameters or emergency shutdown operations when necessary.

[0042] In this embodiment, the module also integrates a real-time alarm system, which can immediately notify operators via SMS, email or application when an abnormal state is detected, ensuring that the problem can be handled as soon as possible.

[0043] The maintenance and optimization module automatically generates customized maintenance plans and operation optimization suggestions based on real-time and historical data. It uses a reinforcement learning model to dynamically calculate the optimal maintenance cycle of the pump and adjust the maintenance plan according to the actual operating status of the equipment. It provides targeted operation optimization suggestions based on different working conditions and environmental conditions.

[0044] In this embodiment, in the maintenance and optimization module, the reinforcement learning model is constructed by a Markov decision process MDP, including: State Space : The operating status set of a special pump includes a combination of different sensor data, such as ; Action Space : A collection of maintenance decisions or optimization suggestions taken under different conditions, such as adjusting the speed of the pump, replacing parts, starting or stopping the pump; Reward Function : When performing an action The immediate feedback or reward after the decision reflects the positive or negative impact of the decision on the pump operation; Strategy : Given state Next, select Action The probability distribution of Discount Factor : The discount rate for future rewards, which represents the relationship between long-term benefits and current benefits, ranging from ; And use an optimization algorithm to optimize the decision-making process of the model. The algorithm estimates the long-term value of each state-action pair by updating the Q-value function. The update formula of the Q-value function is as follows: in, is the learning rate, which controls the step size of the update; An immediate reward for the current moment; is the discount factor, which indicates the weight of future rewards; Indicates that in the next state Next, choose the best action to bring the greatest value; In the pump maintenance optimization scenario, the reward function The design is dynamically adjusted according to the health status, maintenance cost and operating efficiency of the pump. Assume that the reward function is as follows: in, Is to perform an action the maintenance costs required; It is the health status of the pump. The higher the value, the better the pump status. is the operating energy consumption of the pump, the higher the value, the greater the energy consumption; and are weight parameters that control the impact of health status and energy consumption respectively.

[0045] In the reinforcement learning model, an adaptive exploration strategy is set to optimize the trade-off between exploration and utilization of the algorithm. Greedy strategy control, that is, based on probability Randomly select actions to The probability of selecting the current optimal action is based on the dynamic exploration rate , the formula is as follows: in, is the initial exploration rate; is the exploration rate attenuation factor, which controls the speed at which the exploration rate decreases over time; Over time, the model gradually shifts from exploration mode to exploitation mode, where it takes full advantage of the learned optimal policy.

[0046] like Figure 4 As shown, the reinforcement learning model is applied through the following steps: Environmental perception and state construction: The system monitors the pump's operating data in real time through the sensor integration module, such as vibration, temperature, pressure, etc. The state vector is constructed based on these data. , and passed to the reinforcement learning model.

[0047] Action selection: At each time step , the model is based on the current state and strategies , select an action , such as adjusting pump operating parameters, recommending component replacement or scheduling maintenance operations.

[0048] Feedback and reward calculation: After the model executes an action, the system will calculate the immediate reward based on the health status and energy consumption of the pump. , and transfer the pump state to the next moment .

[0049] Q-value update and strategy optimization: According to the algorithm, the model optimizes future decision-making strategies by updating the Q-value function to ensure that the action with the highest benefit is selected under the same state.

[0050] Adaptive exploration and strategy improvement: The model is based on the dynamic exploration rate , a process that balances exploring new strategies and leveraging known optimal strategies at different stages of operation. When equipment conditions change significantly, the system can increase the exploration rate to respond to new data patterns.

[0051] Maintenance suggestion generation: Based on the learned optimal strategy, the model generates adaptive maintenance suggestions to users. These suggestions include when to perform maintenance, which parts need to be replaced, whether to adjust the operating parameters of the equipment, etc., to optimize the overall performance of the equipment.

[0052] Long-term maintenance optimization: Through continuous learning of historical equipment data, the model gradually optimizes maintenance cycles and strategies, extending equipment life and reducing the occurrence of sudden failures. The reinforcement learning model continuously adjusts the weights in each maintenance decision, making future maintenance recommendations more accurate and efficient.

[0053] The multi-source data fusion module integrates sensor data from inside the pump body and external environmental data to conduct multi-dimensional data monitoring and analysis.

[0054] In this embodiment, the multi-source data fusion module adopts a dynamic weight adjustment algorithm to dynamically adjust the weights of different data sources according to the real-time operating status of the pump and changes in the external environment; for example, in a high temperature environment, the data weight of the temperature sensor will automatically increase to more accurately reflect the impact of temperature on the operation of the pump.

[0055] By fusing multi-source data, the operating status of the pump is evaluated to improve the accuracy of fault prediction and operation optimization.

[0056] The digital twin simulation module uses digital twin technology to replicate the operating status of special pumps in real time in virtual space. By establishing a digital model of the pump, real-time simulation is performed in a virtual environment to test the impact of different operating conditions on the pump.

[0057] In this embodiment, the digital twin simulation module uses finite element analysis to simulate the mechanical properties and fluid movement of the pump; By combining real-time data with historical data, the digital twin dynamically reflects the actual operating status of the pump and pre-evaluates the operating effects under extreme conditions or high loads.

[0058] like Figure 2 As shown, in this embodiment, the system also includes a collaboration module, which supports multiple communication protocols and interface standards to meet the integration requirements of different devices and systems; Through data routing and protocol conversion, seamless connection and data sharing between different modules are achieved, and docking with external systems is supported, such as enterprise resource planning ERP system, maintenance management system MMS, etc., to improve overall operational efficiency and data consistency.

[0059] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0060] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent data acquisition system for special pumps, characterized in that: It includes data acquisition module, data preprocessing module, data storage module, data analysis and prediction module, intelligent control module, remote monitoring module, maintenance and optimization module, multi-source data fusion module and digital twin simulation module; The data acquisition module collects data on the operating status of special pumps in real time based on different types of sensors; The data preprocessing module preprocesses the raw data collected from the sensor and removes noise and interference in the data based on the filtering algorithm to improve the signal-to-noise ratio of the data; The data storage module stores and manages the pre-processed data, adopts a distributed database architecture, supports the storage and rapid retrieval of large-scale data, and has a multi-level data backup and redundancy mechanism to prevent data loss and damage; The data analysis and prediction module uses an adaptive neural network model to perform real-time analysis on the operating data of special pumps, dynamically adjusts model parameters based on the pump's operating history data, performs status monitoring and fault prediction, and identifies potential fault modes by analyzing the changing trends of key parameters such as flow, vibration, temperature, and pressure, and issues early warning signals; Intelligent control module, which automatically adjusts the operating status of the pump based on real-time analysis results, automatically adjusts the speed, pressure and flow of the pump according to the analysis data, and optimizes the operating parameters of the pump in real time according to the external environment and the workload of the pump to ensure that the pump works in the best condition; Remote monitoring module, which can check the operation status and data of special pumps in real time through the Internet, provide a visual monitoring interface, access the system in real time at different locations based on the remote access function, diagnose and analyze the operation status of the pump, and support remote operation, allowing remote adjustment of the pump's operating parameters or emergency shutdown operations when necessary; The maintenance and optimization module automatically generates customized maintenance plans and operation optimization suggestions based on real-time and historical data. It uses a reinforcement learning model to dynamically calculate the optimal maintenance cycle of the pump and adjust the maintenance plan according to the actual operating status of the equipment. It also provides targeted operation optimization suggestions based on different working conditions and environmental conditions. Multi-source data fusion module integrates sensor data from inside the pump body and external environment data to conduct multi-dimensional data monitoring and analysis; The digital twin simulation module uses digital twin technology to replicate the operating status of special pumps in real time in virtual space. By establishing a digital model of the pump, real-time simulation is performed in a virtual environment to test the impact of different operating conditions on the pump.

2. The intelligent data acquisition system for special pumps according to claim 1 is characterized in that: In the data acquisition module, sensors are installed and integrated at various locations of the special pump for monitoring the operating status, and the operating status of the pump is monitored in real time; Sensor types include temperature sensors, pressure sensors, flow sensors, vibration sensors and speed sensors. The sensors transmit data to the data acquisition module via wired or wireless means and support multiple communication protocols and transmission methods.

3. The intelligent data acquisition system for special pumps according to claim 1 is characterized in that: In the data analysis and prediction module, the adaptive neural network model ANN optimizes fault prediction and equipment health status assessment by dynamically adjusting model weights based on multi-dimensional data collected during the operation of the special pump; When the data distribution of the pump changes under different operating conditions, the weight of the neural network is adaptively adjusted in real time; Set the structure of the neural network to: Input layer: contains Sensor data input, indicating different operating status parameters; Hidden layer: contains multiple layers of hidden neurons and uses adaptive activation functions; Output layer: predict the health status of the pump and failure probability ; Inputting features into the neural network: vibration data , temperature data , pressure data , traffic data , forming the input vector: ； Output predictions include: Health status prediction: Indicates that the pump is at time The health status of the device is in the range of [0,1], where 0 means the device is seriously damaged and 1 means the device is in good condition; Failure probability prediction: is the failure probability, which indicates the probability that the equipment will fail in the short term.

4. The intelligent data acquisition system for special pumps according to claim 3 is characterized in that: The adaptive neural network model ANN dynamically adjusts the model weights based on the dynamic weight adjustment mechanism, assuming that the model adaptively adjusts the weights according to the changes in the input data. , the dynamic weight adjustment mechanism is as follows: ； in, Representing neurons arrive The weight between The value of is the learning rate; For the current moment The loss function of It is a dynamic item of weight change, which is adjusted in real time according to the operating status of the pump: when the health status or working condition of the pump fluctuates greatly, the model needs a higher weight adjustment speed to adapt to the fluctuation. By calculating the gradient change of each sensor data, the comprehensive gradient change of the sensor data is integrated based on the weighted average algorithm. This volatility is obtained through the comprehensive gradient change, and the dynamic item is adjusted accordingly. is the adaptive factor, the adaptive factor The adjustment of is controlled by: ； in, It is the comprehensive gradient change of sensor data, which is used to reflect the volatility of the system's working conditions. If the sensor data fluctuates greatly, The corresponding increase is used to speed up the weight adjustment of the model; is the adjustment coefficient, controlling the fluctuation of health status the extent of the impact; It is the change in health status, indicating the difference between the current health status and the previous moment.

5. The intelligent data acquisition system for special pumps according to claim 3 is characterized in that: In the adaptive neural network model ANN, in order to optimize the prediction of health status and failure probability, the loss function is designed to be divided into two parts: Health status prediction loss: Use mean square error MSE to measure the gap between the predicted health status and the actual status: ； in, Indicates the number of samples, used to average the errors of all samples; Indicates The index of the samples, starting from Sample starts at samples; is the actual health status; is the health state predicted by the model; Failure probability prediction loss: The cross entropy loss function is used to measure the gap between the predicted failure probability and the actual failure: ； in, Indicates The true failure probability of a sample is 0 or 1, indicating a normal or faulty state; The model predicts the The failure probability of a sample is a continuous value in the interval [0, 1], indicating the possibility of failure. The total loss function is: ； in, and is a weight parameter used to adjust the importance of health status and fault prediction; An adaptive activation function is set to improve the generalization ability of the model under different operating conditions. The adaptive activation function is as follows: ； in, is a dynamically changing adaptive parameter that is adjusted according to the volatility of the input data: ； in, is the initial steepness parameter of the adaptive activation function, which indicates the basic steepness when the input data does not change significantly; It is a regulating factor, which means that when the input data change rate is large, the steepness of the activation function increases, which is used to improve the response speed of the network; is the input vector in the neural network.

6. The intelligent data acquisition system for special pumps according to claim 1, characterized in that: In the maintenance and optimization module, the reinforcement learning model is constructed through the Markov decision process MDP, including: State Space : The operating status set of a special pump includes a combination of different sensor data, such as ; Action Space : A collection of maintenance decisions or optimization suggestions taken under different conditions, such as adjusting the speed of the pump, replacing parts, starting or stopping the pump; Reward Function : When performing an action The immediate feedback or reward after the decision reflects the positive or negative impact of the decision on the pump operation; Strategy : Given state Next, select Action The probability distribution of Discount Factor : The discount rate for future rewards, which represents the relationship between long-term benefits and current benefits, ranging from ; And use an optimization algorithm to optimize the decision-making process of the model. The algorithm estimates the long-term value of each state-action pair by updating the Q-value function. The update formula of the Q-value function is as follows: ； in, is the learning rate, which controls the update step size; An immediate reward for the current moment; is the discount factor, which indicates the weight of future rewards; Indicates that in the next state Next, choose the best action to bring the greatest value; In the pump maintenance optimization scenario, the reward function The design is dynamically adjusted according to the health status, maintenance cost and operating efficiency of the pump. Assume that the reward function is as follows: ； in, Is to perform an action the maintenance costs required; It is the health status of the pump. The higher the value, the better the pump status. is the operating energy consumption of the pump, the higher the value, the greater the energy consumption; and are weight parameters that control the impact of health status and energy consumption respectively.

7. The intelligent data acquisition system for special pumps according to claim 6, characterized in that: In the reinforcement learning model, an adaptive exploration strategy is set to optimize the trade-off between exploration and utilization of the algorithm. Greedy strategy control, that is, based on probability Randomly select actions to The probability of selecting the current optimal action is based on the dynamic exploration rate , the formula is as follows: ； in, is the initial exploration rate; is the exploration rate attenuation factor, which controls the speed at which the exploration rate decreases over time; Over time, the model gradually shifts from exploration mode to exploitation mode, where it takes full advantage of the learned optimal policy.

8. The intelligent data acquisition system for special pumps according to claim 1, characterized in that: The multi-source data fusion module adopts a dynamic weight adjustment algorithm to dynamically adjust the weights of different data sources according to the real-time operating status of the pump and changes in the external environment; By fusing multi-source data, the operating status of the pump is evaluated to improve the accuracy of fault prediction and operation optimization.

9. The intelligent data acquisition system for special pumps according to claim 1, characterized in that: The digital twin simulation module uses finite element analysis to simulate the mechanical properties and fluid movement of the pump; By combining real-time data with historical data, the digital twin dynamically reflects the actual operating status of the pump and pre-evaluates the operating effects under extreme conditions or high loads.

10. The intelligent data acquisition system for special pumps according to claim 1, characterized in that: The system also includes collaborative modules that support multiple communication protocols and interface standards to meet the integration requirements of different devices and systems; Through data routing and protocol conversion, seamless connection and data sharing are achieved between different modules, and docking with external systems is supported.

Images