Low-delay 5G-A communication method and system for hydraulic data acquisition

Abstract

The invention discloses a low-delay 5G-A communication method and system for hydraulic data acquisition, and belongs to the technical field of communication. According to the method, a discretization state space model based on a hydraulic system first principle is constructed and solidified in a data acquisition terminal, and a hydraulic state measured value is acquired in real time and compared with a model predicted value to obtain a residual signal. According to an MCS index issued by a 5G-A base station, a quantization encoding mode is dynamically selected from a pre-stored encoding strategy table to encode a residual signal, when a residual amplitude exceeds a preset threshold value, a high-priority mark is set in an encoding frame, high-precision encoding is forcibly adopted, the encoding frame is sent through a 5G-A communication module, and the base station is triggered to allocate URLLC resources for preferential transmission. The system further comprises an edge server which maintains a model parameter library synchronized with the terminal and is used for receiving the coding frame and reconstructing the original hydraulic state data. According to the invention, the wireless transmission data volume is greatly reduced, reliable transmission of key abnormal information is ensured, and high-precision remote monitoring of the hydraulic state is realized.

Description

Technical Field

[0001] This application belongs to the field of communication technology, specifically relating to a low-latency 5G-A communication method and system for hydraulic data acquisition. Background Technology

[0002] Against the backdrop of the deep integration of the Industrial Internet of Things (IIoT) and 5G mobile communication technology, the condition monitoring of hydraulic equipment is evolving towards higher real-time performance and higher reliability. As the core power unit of key equipment such as engineering machinery and intelligent manufacturing, the operating status of the hydraulic system directly affects the overall performance and safety of the machine. Currently, mainstream monitoring solutions generally use sensors to collect analog signals such as pressure and flow at fixed frequencies. After analog-to-digital conversion, the raw data bitstream is uploaded to a remote server for analysis via a network. Although this communication paradigm achieves basic data feedback, it is difficult to adapt to the stringent end-to-end latency requirements of ultra-reliable low-latency communication scenarios.

[0003] Several improvement schemes have been proposed in existing technologies. For example, some schemes employ a controller-initiated reporting mechanism, sending data to the server and configuring priority sorting based on set threshold conditions; others attempt to dynamically adjust the acquisition frequency and data type by calling a pre-defined acquisition strategy library according to the communication scenario type. Furthermore, some schemes focus on the time-series storage and rule updates of backend data, generating filtering rules through invalid data detection and machine learning analysis to improve the real-time performance and accuracy of data processing. However, these schemes mostly concentrate on optimizing general data acquisition and transmission architectures, and still have limitations when facing the unique nonlinear time-varying characteristics of hydraulic systems.

[0004] When early faults such as minor leaks or valve core wear occur, the abnormal signals are weak and short-lived, easily getting lost in the flood of regular data and resulting in missed detections. Simultaneously, continuously uploading highly redundant data under steady-state conditions severely wastes scarce wireless channel resources. More critically, existing technologies lack physical interpretability of the inherent dynamic laws of hydraulic systems, making high-precision state reconstruction difficult and unable to link physical domain anomalies with communication domain resource allocation across layers.

[0005] Therefore, there is an urgent need for a new low-latency communication architecture that deeply integrates hydraulic physical priors, has adaptive semantic extraction capabilities, and is compatible with 5G-A dynamic scheduling mechanisms, in order to achieve efficient, reliable, and accurate remote monitoring of hydraulic status. Summary of the Invention

[0006] The purpose of this invention is to provide a low-latency 5G-A communication method and system for hydraulic data acquisition, which can effectively solve the problems in the background art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, a low-latency 5G-A communication method for hydraulic data acquisition includes the following steps: In the data acquisition terminal, a discretized state-space model based on the first principles of the hydraulic system is pre-built and solidified. The model includes a system matrix, an input matrix, and an output matrix, which is used to predict the system state at the current moment based on the system state vector at the previous moment and the current control command. The measured values ​​of the hydraulic system's state are acquired in real time and compared with the predicted output values ​​generated by the discretized state-space model based on the control command at the same moment to obtain the residual signal. Obtain the modulation and coding scheme (MCS) index from the 5G-A base station. Based on the channel quality level corresponding to the MCS index, dynamically select the appropriate quantization and coding method from the pre-stored coding strategy table, quantize and encode the residual signal, and generate a coded frame. When the amplitude of the residual signal exceeds the preset threshold, a high priority flag is set in the coded frame, and a coding method with a higher quantization precision than the current channel quality level is forced to be used. The coded frame is sent to the 5G-A base station through the 5G-A communication module, and the high priority flag triggers the base station to allocate ultra-reliable low-latency communication (URLLC) resources for priority transmission.

[0008] Preferably, the process of constructing the discretized state-space model includes: A continuous-time nonlinear dynamic model of a hydraulic system is established based on the fundamental laws of fluid mechanics and dynamics. The nonlinear dynamic model includes at least the pressure dynamic equation, the kinematic equation, and the flow continuity equation. The nonlinear dynamic model is transformed into a linear state-space model by using a small-deviation linearization method near a preset typical operating point. The linear state-space model is discretized using a zero-order hold method with a step size matching the system sampling period to obtain the system matrix, input matrix, and output matrix. The matrix coefficients are then tuned through parameter identification experiments or online learning algorithms.

[0009] Preferably, the step of generating the predicted output value includes: The current control instructions are obtained in real time from the programmable logic controller (PLC), and the normalized control instructions are used as input to the discretized state-space model. Simultaneously, the system state vector of the previous moment is read, and the state prediction value of the current moment is calculated recursively through the system matrix and the input matrix. Then, the state prediction value is linearly projected onto the output matrix to obtain the predicted output value. Among them, the control command is used as the feedforward input model to eliminate the influence of state fluctuations caused by normal operation on the residual signal.

[0010] Preferably, the preset threshold is set based on the statistical characteristics of multiple consecutive residual samples collected when the system is in a steady state, specifically including: Under steady-state conditions, N residual samples are collected, their sample standard deviation σ is calculated, and a preset threshold is set to 3σ, where the value of N is determined according to the dynamic characteristics of the system.

[0011] Preferably, the encoding strategy table contains a mapping relationship between different MCS index ranges and quantization encoding methods; When the MCS index corresponds to a high channel quality level, 8-bit uniform quantization is used, and the quantization step size is fixed at 1 / 256 of the full scale of the residual signal. When the MCS index corresponds to a low channel quality level, 4-bit non-uniform quantization is used. The quantization interval of the non-uniform quantization is obtained offline by training the Lloyd-Max algorithm based on the probability density function of the residual signal and is pre-stored as a quantization table. At the same time, Hamming codes are added to the quantized codewords for forward error correction.

[0012] Preferably, the high-priority flag triggering the base station to allocate URLLC resources includes: After detecting the high priority flag, the base station scheduler allocates at least two consecutive resource blocks (RBs) to the data acquisition terminal and uses the highest available modulation order. At the same time, it configures the physical downlink shared channel (PDSCH) to URLLC mode, where the subcarrier spacing is set to 120 kHz and the transmission time interval (TTI) is compressed to 0.125 milliseconds.

[0013] Preferably, the method further includes a data reconstruction step on the edge server side: The edge server pre-maintains a discretized state-space model parameter library corresponding to each data acquisition terminal; Receive the encoded frames forwarded by the base station, and perform channel decoding and inverse quantization operations on the encoded frames to recover the residual signal; At the same time, control instructions corresponding to the timestamp of the encoded frame are obtained from the core network, and the physical model synchronized with the terminal side is run using the locally stored model parameters to generate the predicted output value at the same time. The predicted output value is added to the recovered residual signal to reconstruct the original measured hydraulic state value, which is then stored as a time-series database record in chronological order according to timestamps.

[0014] Preferably, the model parameter library is kept synchronized with the data acquisition terminal through the 5G core network. The synchronization process uses a bidirectional transport layer security protocol (TLS) based on X.509 certificates to establish an encrypted channel. The triggering conditions include the arrival of a timed period or changes in model parameters caused by terminal firmware upgrades.

[0015] Preferably, the frame structure of the encoded frame includes a frame start flag, a terminal unique identifier, a sequence number, a microsecond-level timestamp, a quantized residual data field, an anomaly flag, and a cyclic redundancy check (CRC) field. The exception flag is used to indicate whether the current residual signal exceeds a preset threshold.

[0016] Secondly, a low-latency 5G-A communication system for hydraulic data acquisition includes: The data acquisition terminal is deployed on the hydraulic equipment side. The data acquisition terminal integrates a sensor interface, a physical model module, a residual calculator, a semantic encoder, and a 5G-A communication module. The sensor interface is used to connect with external sensors to acquire hydraulic state analog signals in real time, perform analog-to-digital conversion, and output digital measured values ​​with timestamps. The physical model module is implemented using an embedded microcontroller. Its program memory contains the system matrix, input matrix, and output matrix of a discretized state-space model based on the first principles of the hydraulic system. This matrix is ​​used to recursively predict the current state value based on the system state vector of the previous moment and the current control command, and then output the predicted output value. The residual calculator connects to the sensor interface and physical model module to calculate the difference between the measured value and the predicted output value at the same time as the residual signal, and monitors whether the amplitude of the residual signal exceeds the preset threshold. When it exceeds the threshold, the abnormal flag is set. The semantic encoder includes a quantization lookup table unit and a channel coding logic array. It is used to dynamically select a quantization coding strategy based on the modulation and coding scheme (MCS) index received from the downlink control channel of the 5G-A base station, quantize and encode the residual signal to generate a coded frame, and force the use of high-precision coding and set a high-priority flag when the abnormal flag bit is valid. The 5G-A communication module is connected to the semantic encoder via an internal bus to send the encoded frame to the 5G-A base station through the 5G-A wireless air interface, and uses a high-priority flag to trigger the base station to allocate ultra-reliable low-latency communication (URLLC) resources for priority transmission. The system also includes an edge server connected to a 5G-A base station. The edge server maintains a discretized state-space model parameter library synchronized with each data acquisition terminal, which is used to receive encoded frames forwarded by the base station and reconstruct the original hydraulic state data.

[0017] In summary, this application includes at least one of the following beneficial technical effects: 1. This invention constructs and solidifies a discretized state-space model based on the first principles of a hydraulic system, generates predicted output values ​​in real time at the data acquisition terminal, and transmits only the residual signal between the measured and predicted values. Compared with the traditional scheme of transmitting the original bit stream, this invention can significantly reduce the amount of wireless data transmitted under steady-state conditions and significantly save scarce wireless channel resources.

[0018] 2. This invention dynamically selects the quantization coding strategy based on the MCS index issued by the 5G-A base station. When the channel quality is high, 8-bit uniform quantization is used to ensure high fidelity. When the channel quality is low, 4-bit non-uniform quantization based on the Lloyd-Max algorithm and Hamming code are added. This realizes cross-domain linkage between communication parameters and physical signal probability distribution, effectively improving the transmission reliability under poor channel conditions.

[0019] 3. When the residual signal exceeds a preset threshold, the present invention sets a high-priority flag and forces the use of high-precision coding, triggering the base station to allocate URLLC dedicated resources for priority transmission. This establishes a feedback closed loop from physical layer residual to MAC layer resource scheduling, ensuring that early fault characteristic data such as minor leaks and valve core wear in the hydraulic system are transmitted instantaneously and reliably, meeting the requirements of industrial-grade ultra-low latency monitoring. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a low-latency 5G-A communication method for hydraulic data acquisition according to the present invention. Figure 2 This is a schematic diagram of the process of prediction and residual generation based on physical model in this invention; Figure 3 This is a flowchart illustrating the adaptive semantic coding and dynamic wireless scheduling process in this invention. Figure 4 This is a schematic diagram of the process of edge server state reconstruction and upper-layer application collaboration in this invention. Detailed Implementation

[0021] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of specific embodiments based on the present invention is provided in conjunction with the accompanying drawings and preferred embodiments.

[0022] Reference Figures 1 to 4 As shown, a low-latency 5G-A communication method for hydraulic data acquisition includes the following steps: Step S1 involves deploying the system architecture, specifically including: Step S101: Configure a data acquisition terminal on the hydraulic equipment side. The terminal establishes a control command communication link with the programmable logic controller (PLC) of the hydraulic equipment through an industrial Ethernet or dedicated bus to ensure that control command signals can be received in real time. At the same time, the terminal establishes an uplink data transmission channel with a 5G-A base station through a 5G-A wireless air interface to send the processed data to the base station.

[0023] In step S102, the 5G-A base station is connected to the edge server via the fronthaul interface, and at the same time, the 5G-A base station is connected to the 5G core network via the NG interface. The edge server is deployed in an edge computing node close to the base station and is connected to the base station via a fiber optic link. The round-trip latency is designed to be less than 1 millisecond to meet the requirements of the ultra-reliable low-latency communication (URLLC) scenario. The edge server is also connected to the 5G core network via the N6 interface for data interaction with the core network.

[0024] Step S103: The data acquisition terminal integrates four types of functional modules: sensor interface, physical model module, residual calculator, and semantic encoder. The sensor interface is responsible for interfacing with external sensors. It supports analog voltage input range of 0 to 10 volts or current loop of 4 to 20 mA. It has a built-in 16-bit analog-to-digital converter (ADC) and includes a double buffer for temporarily storing the converted digital measured values. The sampling clock is provided by a local temperature-compensated crystal oscillator with frequency stability better than ±50ppm, ensuring the consistency of acquisition accuracy and time reference.

[0025] The data acquisition terminal also integrates a 5G-A communication module, which includes a radio frequency front-end and a baseband processor. It is connected to the sensor interface, physical model module, residual calculator and semantic encoder through an internal bus. It is used to send the encoded data frame to the base station through the 5G-A wireless air interface. At the same time, it transmits the modulation and coding scheme (MCS) index obtained from the downlink control channel of the base station to the semantic encoder in real time through the internal bus, so that the encoder can dynamically adjust the coding strategy.

[0026] Step S104: The physical model module is implemented using an embedded microcontroller. Its program memory stores the state-space model parameter matrix of the discretized state-space equations. These coefficient matrices include the system matrix. Input matrix The output matrix C is not a simple black box model, but a white box model built based on the first principles of hydraulic systems. Its construction process deeply reflects the integration of physical prior knowledge of the nonlinear time-varying characteristics of hydraulic systems.

[0027] First, for core actuators of hydraulic equipment, such as hydraulic cylinders, a continuous-time nonlinear dynamic model is established based on the fundamental laws of fluid mechanics and dynamics. This model includes at least the following equations: The pressure dynamic equation describes the relationship between the pressures p_1 and p_2 in the two chambers of a hydraulic cylinder and the change in oil volume, specifically: Where β is the effective bulk elastic modulus of the oil, and its value varies with oil temperature and gas content, V is the cavity volume, Q is the inflow or outflow flow rate, A is the effective area of ​​the piston, and v is the piston speed.

[0028] The kinematic equations describe the motion of the piston and the load, specifically: Where m is the equivalent load mass. For velocity-dependent nonlinear friction models, such as the Stribeck effect, which exhibits nonlinearity, This is the external load force.

[0029] The flow continuity equation describes the relationship between the flow rate Q through a servo or proportional valve and the control command u(t) and the valve orifice pressure difference Δp, specifically: in, Here, ρ is the flow coefficient, w is the valve orifice area gradient, and ρ is the oil density. This equation reflects the nonlinearity of the system and the correlation with the control input.

[0030] Secondly, for the aforementioned set of nonlinear differential equations, a small-deviation linearization method is used near the preset typical operating point of the hydraulic system to transform it into a linear state-space model. The linearization process preserves the dominant dynamic characteristics of the system under this operating condition. By repeating this process at different typical operating points, a family of linear models characterizing the dynamic behavior of the system under different operating conditions is obtained.

[0031] Finally, the continuous-time linear state-space model is precisely discretized using the zero-order hold method with a step size of 1 millisecond, which is strictly matched with the system sampling period, to obtain the coefficient matrix. , The coefficients of these matrices, C, are tuned through parameter identification experiments or real-time online learning algorithms before the hydraulic equipment leaves the factory, ensuring that they accurately match the physical characteristics of the specific equipment. It is a 2×2 system matrix. C is a 2×1 input matrix and C is a 1×2 output matrix, used to predict the current state based on the previous state vector and the current control command.

[0032] In this embodiment, after linearizing the above nonlinear model, the pressure p in the rodless chamber of the hydraulic cylinder and the piston speed v are selected as state variables to form a second-order state vector. Wherein, pressure p corresponds to the pressure in the dynamic equation of pressure and the two-chamber pressure in the kinematic equation. and By pre-set relationships , By approximating α and β, which are constants obtained through parameter identification, the nonlinear model is simplified into a linear state-space model with p and v as state variables.

[0033] Step S105 The residual calculator consists of a hardware subtractor that supports 32-bit fixed-point arithmetic in a single cycle. The residual calculator is responsible for calculating the difference between the sensor's measured value and the physical model's predicted value, generating a residual signal. This residual signal is used to characterize the degree of deviation between the actual dynamics of the system and the model.

[0034] The residual calculator has built-in peak detection logic to continuously monitor whether the absolute value of the residual exceeds a preset threshold τ. This preset threshold τ can be determined based on the standard deviation of multiple consecutive residual samples collected when the system is in a steady state. For example, the three-times-standard-deviation principle can be used to ensure a low false alarm rate within the normal fluctuation range. If the absolute value of the residual exceeds this threshold, an anomaly is determined and the anomaly flag is immediately set.

[0035] Step S106: The semantic encoder includes a quantization lookup table unit and a channel coding logic array. It can dynamically switch coding strategies based on the modulation and coding scheme (MCS) index received from the downlink control channel of the 5G-A base station. The semantic encoder has a pre-stored coding strategy table that maps different MCS index ranges to corresponding quantization coding methods. Specifically: When the MCS index corresponds to a high channel quality level, such as a Channel Quality Indicator (CQI) greater than or equal to 10, the semantic encoder enables 8-bit uniform quantization with a fixed quantization step size of 1 / 256 of the full scale to transmit residual data with high fidelity.

[0036] When the MCS index corresponds to a low channel quality level, such as CQI less than 10, the semantic encoder enables 4-bit non-uniform quantization. Its quantization interval is obtained offline by training the Lloyd-Max algorithm based on the probability density function of the residual signal and is pre-stored as a quantization table, making the quantization of small amplitude regions more refined. At the same time, Hamming codes are added for forward error correction to improve anti-interference capability.

[0037] In abnormal situations, i.e. when the abnormal flag bit of the residual calculator is valid, the semantic encoder ignores the MCS indication, forces the use of 8-bit uniform quantization to retain more details, sets a high priority flag in the encoded frame header, requests the base station to allocate URLLC dedicated resources, and ensures that abnormal event data is transmitted with priority and reliability.

[0038] Through the deployment of the above system architecture, this invention constructs a complete link from sensor data acquisition to edge server reconstruction. The data acquisition terminal deeply integrates the hydraulic physical model and adaptive semantic coding, which can significantly reduce the amount of wireless data transmission while ensuring the reliable transmission of key abnormal information, providing hardware support and infrastructure guarantee for real-time status monitoring and fault diagnosis in subsequent steps.

[0039] Step S2 acquires real-time hydraulic status data. The specific execution process is as follows: In step S201, the pressure sensor or flow sensor sets the sampling frequency to 1000 Hz according to the dynamic response characteristics of the hydraulic system. This frequency can cover the key frequency band of the main dynamic behavior of the hydraulic system, ensuring that the sampling signal reflects the transient changes of the system without distortion. The sensor continuously outputs an analog signal at this frequency.

[0040] In step S202, the analog signal is transmitted to the sensor interface of the data acquisition terminal via a shielded twisted pair cable. The shielded twisted pair cable has good anti-electromagnetic interference capability, which can effectively suppress common-mode and differential-mode interference generated by equipment such as motors and frequency converters in industrial sites, and ensure signal transmission quality.

[0041] In step S203, the sensor interface performs anti-aliasing filtering on the input analog signal. The filter adopts a second-order Butterworth low-pass filter structure, and its cutoff frequency is set to 0.45 times the sampling frequency, i.e. 450 Hz. This design ensures that the amplitude-frequency characteristics within the passband are flat while providing sufficient attenuation for signal components with frequencies higher than the Nyquist frequency, thereby eliminating the aliasing effect of high-frequency noise on sampling accuracy.

[0042] In step S204, the filtered signal is sent to the built-in 16-bit successive approximation analog-to-digital converter (ADC) for analog-to-digital conversion. The reference voltage of the ADC is set to 5 volts, which corresponds to the full-scale output of the pressure sensor. That is, when the measured pressure reaches 35 MPa, the sensor outputs 5 volts. Therefore, the pressure resolution represented by each digital unit can be obtained by dividing the full-scale pressure by 2 to the power of 16 and then multiplying by the ratio of the reference voltage to the full-scale voltage. Specifically, 35 MPa divided by 65536 multiplied by 5 volts and then divided by 5 volts yields a result of approximately 0.534 kPa. This resolution is sufficient to capture minute pressure fluctuations.

[0043] Step S205, the measured value of the converted digital quantity is recorded as follows: Where t represents the current sampling time, the system adds a timestamp with microsecond precision to each sampling point. The timestamp is generated by a high-resolution timer driven by a temperature-compensated crystal oscillator built into the terminal and calibrated through a precise time protocol synchronized with the 5G-A base station to ensure that the time base of each terminal is consistent with that of the edge server, thereby achieving strict alignment of data and time.

[0044] In step S206, the measured digital value with timestamp is temporarily stored in the dual buffer of the sensor interface. The dual buffer adopts a ping-pong operation mechanism, with the two buffers taking turns as the receiving area and the processing area. One buffer is used to receive newly arrived data, and the other buffer is used for subsequent physical model prediction and residual calculator reading. The buffer depth is set to 8 sampling points. This depth takes into account factors such as module processing delay, bus contention and burst data flow, and can effectively absorb short-term processing fluctuations and avoid data loss due to untimely processing.

[0045] In step S207, the timestamp information and the measured value are packaged together according to a predefined frame format to form a raw data frame with a time sequence label. The frame structure includes a frame start flag, a terminal identifier, a timestamp field, a measured value field, and a frame check field. This data frame only flows between modules within the terminal and serves as the input for subsequent physical model prediction and residual calculator. It is not wirelessly transmitted to the outside world to ensure that the raw data always remains on the local terminal, thus meeting data privacy and security requirements.

[0046] Step S3 performs predictions based on the physical model, and its detailed operation is broken down into the following sub-steps: In step S301, the physical model module reads the system state vector x(t-1) from the internal state register of the previous moment. This vector has a dimension of 2 and contains two physical quantities: the pressure p in the hydraulic cylinder cavity and the piston speed v. All data is stored in 32-bit fixed-point number format. The range of the fixed-point number is predetermined according to the sensor range and the dynamic range of the model to ensure the accuracy and efficiency of the calculation.

[0047] At system startup, the initial state vector x(0) is set using the first measured value of the sensor. For example, the measured value of pressure is used as the initial value of p, and the initial value of velocity is set to zero.

[0048] In step S302, the physical model module obtains the current control instruction u(t) from the programmable logic controller (PLC) of the hydraulic equipment via the industrial bus. This instruction is the duty cycle signal of the electromagnetic directional valve, with an original value range of 0 to 100%. The system normalizes and maps it to a dimensionless control quantity of 0 to 1 so as to match the input matrix in the state space model.

[0049] Step S303: The physical model module calls the discretized system matrix pre-stored in the program memory. and Perform matrix-vector multiplication to calculate the predicted state value at the current time. The calculation formula is: in The system matrix is ​​2×2. The input matrix is ​​2×1. The two are obtained by discretizing the continuous-time hydraulic dynamics model using the zero-order hold method. The sampling period is strictly matched to 1 millisecond to ensure that the numerical integration error is less than 0.1% of the full scale.

[0050] In step S304, the physical model module uses the output matrix C to perform a linear projection on the state prediction value to obtain the prediction output value at the current time. The calculation formula is: Where C is a 1×2 row vector, and its specific elements are preset according to the sensor installation position and the measurement target. For example, if the sensor measures pressure, then C=[1 0], if it measures speed, then C=[0 1], and if it measures a combination of both, then the corresponding coefficients are set according to the physical relationship.

[0051] In step S305, the physical model module writes the calculated current state vector x(t) into the internal state register as the initial state for the next time step, thus forming a recursive update.

[0052] In step S306, the physical model module executes a complete prediction loop every 1 millisecond, which includes all the calculations from steps S301 to S305 above.

[0053] Step S4 calculates the residual signal, and the specific implementation mechanism is as follows: Step S401: The residual calculator reads the measured digital value at the current moment from the double buffer of the sensor interface. This value includes a timestamp t with microsecond precision; simultaneously, the residual calculator obtains the predicted output value corresponding to the same time t from the physical model module. .

[0054] When generating each predicted output value, the physical model module also attaches a corresponding timestamp, which is consistent with the timestamp of the input data used for prediction. The residual calculator achieves strict temporal alignment between the two by comparing the timestamps, ensuring that the measured value and the predicted value used for subtraction are from the same moment.

[0055] Step S402: The residual calculator performs fixed-point subtraction to calculate the residual signal. in, and Both are 32-bit fixed-point number formats, representing the same physical quantity, such as pressure or flow rate, and have the same dimensions and scaling factor, so fixed-point subtraction can be performed directly.

[0056] The calculation results are stored in 32-bit signed integer format, retaining positive and negative sign information to reflect the direction and magnitude of the deviation of the measured value from the predicted value.

[0057] In step S403, when the system is in steady-state operation, since the physical model accurately describes the dynamic characteristics of the hydraulic cylinder, the absolute value of the residual Δy(t) is usually less than 0.5% of the full scale, and it exhibits small random fluctuations around the zero point. When abnormal events such as minor hydraulic oil leakage, increased internal leakage due to valve core wear, or sudden changes in external load occur, the actual system behavior deviates from the model prediction, and the amplitude of Δy(t) increases significantly, possibly reaching more than 5% of the full scale. This characteristic provides a basis for anomaly detection.

[0058] The amplitude and fluctuation pattern of the residual signal Δy(t) are closely related to specific physical process changes. In-depth analysis of the residual signal can further enable pre-diagnosis of fault types. The residual signal Δy(t) contains all unmodeled information between the actual dynamics of the system and the ideal model, for example: When a minor hydraulic oil leak occurs, the leak will cause the equivalent bulk modulus β of the hydraulic cylinder pressure dynamic equation to decrease, which will slow down the system response. At this time, the residual Δy(t) will no longer be a zero-mean white noise, but will exhibit low-frequency drift or phase lag characteristics related to the control command u(t).

[0059] When valve core wear leads to increased internal leakage, the wear will affect the flow coefficient in the flow equation. Changes occur, and additional leakage flow is introduced, causing the damping ratio of the system to change. At this time, the residual Δy(t) will exhibit transient characteristics such as increased overshoot or slower oscillation decay after a step change in u(t) during the dynamic response process.

[0060] When sensor zero-point drift occurs, the physical model itself is accurate, but the measured values... By superimposing a fixed offset, the residual Δy(t) will behave as a constant DC bias, the magnitude of which does not change with u(t) or the system state.

[0061] Therefore, the subsequent step S404 will determine the threshold. This triggers anomaly detection, but this judgment is only a preliminary condition. In practical applications, the data acquisition terminal or edge server of this invention can also perform more refined analysis on the residual sequence {Δy(t)} to identify the fault type. Specifically, the following methods can be used: Spectral analysis is performed on the residual sequence, such as calculating its power spectral density using Fast Fourier Transform to extract energy features in specific frequency bands; time-series statistical features such as kurtosis, root mean square value, and skewness are calculated; and the extracted features are compared with a pre-established residual-fault feature library. This feature library can be obtained by acquiring residual samples under different fault types through fault simulation experiments and trained using classifiers such as support vector machines or neural networks.

[0062] By comparing features, possible fault types can be initially identified while detecting anomalies, such as slowed pressure response suggesting leakage or increased dynamic oscillation suggesting valve core wear, providing valuable information for subsequent accurate diagnosis and maintenance decisions.

[0063] Step S404: The residual calculator has a built-in peak detection logic, which can be implemented by a hardware comparator or a numerical comparison instruction inside the microcontroller to continuously monitor whether the absolute value of the current residual Δy(t) exceeds the preset threshold τ.

[0064] The threshold τ is set as follows: After the system has just started up or after the system is confirmed to be in a steady state by monitoring the rate of change of control commands to be less than a certain threshold, the residual calculator collects residual data from N consecutive sampling points. The value of N is determined according to the dynamic characteristics of the system, for example, 1000 sampling points.

[0065] Then, calculate the sample standard deviation σ of these residual data using the following formula: in For the i-th residual sample, The mean of N residual samples is given, and the threshold τ is finally set to 3σ. Based on the assumption of normal distribution, this threshold can ensure that the false alarm probability is extremely low within the normal fluctuation range.

[0066] During real-time monitoring, if the absolute value of the residual exceeds the threshold, an anomaly is determined, and the residual calculator immediately sets the anomaly flag. This anomaly flag setting can be a bit in a dedicated register or a notification to subsequent modules via an interrupt signal.

[0067] In step S405, the residual calculator packages the current residual Δy(t), the corresponding microsecond-level timestamp t, and the anomaly flag into a data unit according to a predefined frame format. The structure of this data unit includes a frame start flag, data length, timestamp field, residual data field, anomaly flag field, and frame verification field, such as using cyclic redundancy check (CRC) for verification.

[0068] After packaging, the data units are sent to the input queue of the semantic encoder via direct memory access (DMA) or written to a first-in-first-out (FIFO) queue, awaiting subsequent encoding and transmission processing. The input queue is managed using a circular buffer to avoid data overwriting or loss.

[0069] Step S5 implements adaptive semantic encoding, and its complete process includes the following steps: In step S501, the semantic encoder continuously listens to the physical downlink control channel (PDCCH) from the 5G-A base station, parses the downlink control information carried by the channel according to the 3GPP protocol specifications, and extracts the modulation and coding scheme (MCS) index allocated to the terminal by the network side.

[0070] The MCS index is dynamically calculated and generated by the base station based on the Channel Quality Indicator (CQI) obtained by the uplink sounding reference signal (SRS). The terminal only needs to follow the standard protocol stack to complete the parsing to obtain the current channel quality level information without any additional intervention.

[0071] In step S502, after receiving the MCS index, the semantic encoder immediately queries the internally stored encoding strategy table. This strategy table uses the MCS index range as the key and maps it to the corresponding quantization encoding method. For example, the high channel quality range corresponds to 8-bit uniform quantization, and the low channel quality range corresponds to 4-bit non-uniform quantization with Hamming code appended.

[0072] The specific mapping relationship of the strategy table is pre-set according to the system design goals. It can be fixed at the factory or updated through subsequent firmware upgrades.

[0073] Step S503: Determine the current quantization and coding strategy based on the query results. If the MCS index corresponds to a high channel quality level, such as CQI greater than or equal to 10, proceed to step S504; if it corresponds to a low channel quality level, such as CQI less than 10, proceed to step S505; if an abnormal flag is valid, jump directly to step S508.

[0074] In step S504, under high channel quality conditions, the semantic encoder enables 8-bit uniform quantization.

[0075] First, it is necessary to determine the full-scale range of the residual signal. This range can be preset according to the system design parameters. For example, based on the range of the hydraulic sensor and the statistical distribution of the model prediction error, the full-scale range can be set to ±5 MPa or the corresponding engineering units to ensure that all possible residual amplitudes can be covered.

[0076] The quantization step size Δq is fixed at 1 / 256 of the full scale, that is, if the full scale is R, then Each residual sample is divided by Δq and rounded to obtain a quantization index between 0 and 255, thus achieving high-fidelity transmission.

[0077] Step S505: Under low channel quality conditions, the semantic encoder enables 4-bit non-uniform quantization to improve anti-interference capability. The specific implementation is as follows: First, during the system development phase, a large number of residual signal samples under typical operating conditions were collected, covering steady-state, transient, and various abnormal states. The probability density function of the residual signal was estimated using statistical methods. Then, the Lloyd-Max algorithm was used for iterative optimization. This algorithm divides the value range of the residual signal into 16 non-uniform quantization intervals by minimizing the quantization distortion criterion, and determines an optimal reconstruction level for each interval.

[0078] The iterative process includes: initializing the quantization boundary, calculating the centroid of each interval as the new reconstructed level, updating the boundary according to the new level, and repeating this process until convergence.

[0079] The 16 interval boundary values ​​and the corresponding 16 reconstructed levels are finally obtained to form a non-uniform quantization table, which is pre-stored in the terminal's non-volatile memory in the form of an array. During encoding, the current residual Δy(t) is compared with the quantization table boundary to determine the interval index 0 to 15, which is a 4-bit codeword.

[0080] Meanwhile, in order to resist channel errors, a (7,4) Hamming code is added to this 4-bit codeword to generate a 7-bit encoded output, which the receiver can use for error detection and correction.

[0081] Step S506: The quantized data, along with the corresponding timestamp and anomaly flag, is combined into an encoded frame according to a predefined frame format. The frame structure includes the following fields: The start-of-frame flag is used for synchronization, the terminal unique identifier is used to distinguish different devices, the sequence number is used to detect packet loss and reordering, the timestamp field carries microsecond-level time, the data field stores the quantization index, the anomaly flag indicates whether the current data corresponds to an abnormal event, and finally, the cyclic redundancy check (CRC) field is attached for integrity verification.

[0082] The specific byte length and arrangement of the frame format can be designed according to actual needs.

[0083] In step S507, the assembled encoded frame is sent to the baseband processor of the 5G-A communication module via the internal bus. The baseband processor completes the subsequent physical layer processing, including channel coding, modulation, resource mapping, etc., and finally transmits it through the radio frequency front end.

[0084] In step S508, when the exception flag bit set by the residual calculator is valid, the semantic encoder forcibly ignores the indication of the MCS index and directly uses 8-bit uniform quantization to retain more exception details. The quantization step size is the same as in step S504.

[0085] At the same time, a dedicated high-priority flag is set in the frame header of the encoded frame. This high-priority flag can be a predefined bit. After the base station scheduler receives this flag, it will trigger the emergency scheduling process to allocate URLLC dedicated resources to the terminal, ensuring that abnormal event data is transmitted with priority and reliability.

[0086] Through the above adaptive semantic coding process, this invention dynamically selects the optimal quantization and coding method based on real-time channel quality and abnormal states. While ensuring the reliable transmission of key information, it significantly reduces the amount of transmitted data, providing high-quality data frames for subsequent dynamic scheduling and low-latency communication. Next, the base station will perform dynamic wireless scheduling based on information such as the high priority flag of the coded frame to achieve end-to-end low-latency reliable communication.

[0087] Step S6 involves dynamically scheduling wireless transmission, specifically performed by the 5G-A base station as follows: In step S601, the MAC layer scheduler of the 5G-A base station continuously receives uplink sounding reference signals (SRS) sent from various terminal devices. These SRS signals are sent in strict accordance with the time and frequency resource format specified in the 3GPP protocol. The scheduler obtains the measurement results of the SRS through the physical layer reporting interface, including information such as received power and time of arrival.

[0088] Step S602: The scheduler calculates the instantaneous Channel State Information (CSI) for each terminal based on the received SRS signal. The specific calculation method is as follows: First, the received power of the SRS is used as the signal power. Simultaneously, the noise power is estimated by measuring the received power on unallocated resource blocks. Then, according to the formula... First, the signal-to-noise ratio (SNR) is calculated. Second, the rate of phase change over time is obtained by continuously measuring the phase of the SRS, and then the Doppler frequency shift formula is applied. The Doppler frequency shift is estimated, which reflects the impact of the terminal's movement speed on the channel.

[0089] The CSI obtained in this way includes signal-to-noise ratio and Doppler shift, which can accurately describe the quality and dynamic characteristics of the current uplink channel.

[0090] In step S603, the scheduler allocates transmission resources of the Physical Downlink Shared Channel (PDSCH) to each terminal based on the CSI and service requirements of each terminal. The allocation includes the number of resource blocks (RBs), the modulation order (QPSK, 16QAM, or 64QAM), and the channel coding rate.

[0091] Under normal circumstances, the scheduler determines resource allocation based on algorithms such as proportional fairness or maximum load-to-interference ratio, aiming to improve the overall spectrum efficiency and user experience of the system while ensuring fairness for users.

[0092] Step S604: When the data acquisition terminal of the present invention reports a high priority flag, the scheduler immediately performs emergency scheduling for it.

[0093] Specifically, the scheduler allocates at least two consecutive RBs to the terminal in order to improve transmission reliability by utilizing frequency domain diversity gain, and to maximize transmission efficiency by using the highest modulation order available under the current channel conditions, thereby completing data transmission in the shortest possible time. In addition, the scheduler will prioritize the terminal's scheduling request to reduce its waiting time in the scheduling queue.

[0094] In step S605, the scheduler configures the PDSCH carrying the terminal data to Ultra-Reliable Low-Latency Communication (URLLC) mode. In this mode, the subcarrier spacing is set to 120 kHz, and the transmission time interval (TTI) is compressed to 0.125 milliseconds, which is equivalent to the length of one OFDM symbol. This effectively reduces air interface transmission latency and improves reliability.

[0095] In step S606, the base station maps the encoded residual data to the RB allocated in step S604 or S603.

[0096] The physical layer processing flow is then executed: First, an inverse fast Fourier transform (IFFT) is performed to convert the frequency domain signal into a time domain signal. Then, a cyclic prefix is ​​added to resist multipath interference. Next, a digital-to-analog conversion is performed, and finally, the signal is transmitted through the antenna via the radio frequency front end.

[0097] Step S607: The delay from the entire scheduling decision to the air interface transmission is strictly controlled within 0.5 milliseconds. This delay includes the scheduler decision time, the resource allocation signaling issuance time, the baseband processing time, and the radio frequency transmission time.

[0098] To achieve the latency target, hardware acceleration methods are adopted, such as using FPGAs or dedicated ASICs to execute scheduling algorithms and baseband processing, avoiding the additional overhead of software processing; at the same time, optimizations are made at the protocol level, such as simplifying the signaling interaction process, adopting pre-scheduling mechanisms and fast HARQ feedback, to ensure that high-priority data can pass through the processing pipeline quickly.

[0099] Through these measures, the end-to-end communication link fully meets the stringent latency constraints of URLLC.

[0100] Through the above dynamic scheduling process, this invention achieves rapid response and reliable transmission of high-priority data. At the same time, combined with URLLC configuration, it further ensures low latency requirements and provides real-time and reliable data support for subsequent edge server state reconstruction.

[0101] Step S7 reconstructs the complete state sequence, and its specific implementation steps on the edge server are as follows: Step S701: The edge server maintains a physical model parameter library, organized in a hash table structure, where the key is a unique identifier for the terminal, and the value is the state space model parameter matrix corresponding to that terminal, including the system matrix. Input matrix And the output matrix C.

[0102] These matrices are used to run a physical model that is completely consistent with the terminal side on the edge server side. The data in the physical model parameter library originally comes from the terminal device. When the terminal updates the model parameters through factory identification or online learning, it will send the new matrix to the 5G core network through the 5G uplink. The core network then distributes these parameters to the edge server.

[0103] In step S702, the parameter library is updated synchronously from the 5G core network through a security authentication mechanism. The synchronization process uses a bidirectional transport layer security protocol (TLS) based on X.509 certificates to establish an encrypted channel to ensure the confidentiality and integrity of parameter transmission.

[0104] The synchronization trigger condition is a timed period of 1 hour, or an update is triggered immediately when the terminal firmware upgrade causes changes in the model parameters. Each time an update is performed, the edge server receives a data packet containing the terminal identifier and the new matrix, and updates the corresponding entry in the local hash table.

[0105] In step S703, after the state reconstruction module receives the residual data packet forwarded by the base station, it first performs integrity verification on the data packet. The verification method can be a preset cyclic redundancy check (CRC) or checksum algorithm. If the verification fails, the packet is discarded and the error is recorded. If the verification passes, the next step is performed.

[0106] Step S704: After successful verification, the state reconstruction module parses the terminal's unique identifier in the data packet header and uses this identifier as the key to retrieve the corresponding discretized matrix from the hash table of the physical model parameter library. , and .

[0107] If the retrieval fails, it means that the model parameters of the terminal have not been synchronized. In this case, the current data packet is discarded and the parameter request process is triggered.

[0108] Step S705: The state reconstruction module runs a physical model that is fully synchronized with the terminal side locally.

[0109] To ensure the model's initial state is consistent with the terminal's, when the edge server first receives residual data from a terminal, it extracts the measured values ​​from the first complete time step of the data packet and sets the initial state vector in the same way as the terminal. For example, the measured pressure value is used as the initial pressure state, and the initial velocity value is set to zero. In subsequent runs, the model uses the same recursive formula as the terminal for prediction.

[0110] Meanwhile, the edge server obtains the control instruction sequence u(t) corresponding to the current residual data moment from the programmable logic controller (PLC) through the core network in real time. These control instructions are generated by the terminal-side PLC and transparently forwarded to the edge server through the core network, ensuring that they are completely consistent with the instructions used by the terminal-side model.

[0111] Then, the state reconstruction module performs the same recursive prediction operation as the terminal side, generating prediction output values ​​consistent with those on the terminal side. The prediction calculation formula is as follows: Where x(t) is the state vector at the current moment, which includes pressure p and velocity v; u(t) is the normalized control command; , C is a matrix obtained from the parameter library.

[0112] Step S706: Perform channel decoding operation on the residual data packet.

[0113] If Hamming code 74 is used for data packet encoding, the decoder first uses the Hamming code parity check matrix to perform error detection. If a single bit error is detected, it is automatically corrected and the original quantization index is recovered. If an uncorrectable error is detected, the packet is marked as an error and discarded.

[0114] If Hamming codes are not used during data packet encoding, the quantization index is extracted directly from the data packet.

[0115] Step S707: Perform dequantization on the decoded quantization index to restore the residual value. Dequantization is distinguished according to the quantization type used during encoding.

[0116] If 8-bit uniform quantization is used during encoding, the quantization step size Δq is used, which is completely consistent with that on the terminal side. The floating-point residual value is restored by multiplying the index by Δq and adding the offset. The quantization step size Δq is pre-stored in the edge server and kept synchronized with the terminal side.

[0117] If four-bit non-uniform quantization is used during encoding, the index is mapped to the corresponding reconstructed level value as the floating-point residual value based on the quantization table that is exactly the same as that on the terminal side. This quantization table is also pre-stored in the edge server and synchronized with the terminal side.

[0118] Step S708: Perform an addition operation to combine the predicted value and the residual value to reconstruct the complete original hydraulic data. The calculation formula is as follows: in This represents the complete state value after reconstruction. This refers to the predicted output value generated by the physical model at the same time in step S705. The residual signal obtained after dequantization in step S707 is ideally consistent with the original residual Δy(t) generated by the transmitting end, and the deviation introduced by the quantization error is within an acceptable range.

[0119] Through addition operations, the edge server accurately reconstructs the original hydraulic data collected by the terminal-side sensors.

[0120] Step S709: Sort all the reconstructed state values ​​in order of timestamps to form a complete time series.

[0121] Since network transmission may cause data packets to arrive out of order, the state reconstruction module first temporarily stores the residual data packets in a sorting buffer after receiving them. Once the timestamps in the buffer are consecutive or have timed out, they are written into the high-performance time series database in ascending order of timestamps.

[0122] This high-performance time-series database adopts a columnar storage structure, supports fast time range queries and data compression, and facilitates subsequent analysis and long-term storage.

[0123] In step S710, the reconstruction data is pushed to the upper-layer application in real time through the gRPC interface. The data format used for pushing is Protocol Buffers serialization, and each message contains a terminal identifier, a timestamp, and the corresponding reconstruction status value.

[0124] Upper-layer applications include condition monitoring systems, fault diagnosis platforms, or data visualization services. These systems can obtain real-time status information of hydraulic equipment by subscribing to the corresponding data streams, thus meeting the real-time requirements of business systems.

[0125] Through step S7 above, the edge server completes the entire process from parameter library maintenance to state reconstruction, storage and distribution. This process ensures that the original hydraulic state information can be accurately restored even with a small amount of residual data, significantly reducing the wireless transmission load, while ensuring high data fidelity and low latency, providing reliable data support for remote monitoring and intelligent operation and maintenance.

[0126] Furthermore, based on the communication method, this invention also provides a low-latency 5G-A communication system for hydraulic data acquisition, comprising: The data acquisition terminal is deployed on the hydraulic equipment side. The data acquisition terminal integrates a sensor interface, a physical model module, a residual calculator, a semantic encoder, and a 5G-A communication module.

[0127] The sensor interface is responsible for interfacing with external sensors to acquire analog signals of hydraulic status in real time. This interface has a built-in 16-bit analog-to-digital converter (ADC) that converts analog signals into digital measured values. It also uses a high-resolution timer driven by a local temperature-compensated crystal oscillator to add a microsecond-level timestamp to each sampling point. At the same time, the sensor interface synchronizes with the 5G-A base station through a precise time protocol to ensure that the time base of data acquisition is consistent with the edge server. The converted digital measured values ​​with timestamps are temporarily stored in a dual buffer inside the sensor interface. This buffer adopts a ping-pong operation mechanism, with one buffer used to receive new data and the other for subsequent modules to read. The buffer depth is set to 8 sampling points, which can effectively absorb short-term processing fluctuations and avoid data loss.

[0128] The physical model module is implemented using an embedded microcontroller. Its program memory stores the system matrix, input matrix, and output matrix of a discretized state-space model based on first-principles calculations of the hydraulic system. This module obtains normalized control instructions in real-time from the programmable logic controller (PLC) and reads the system state vector from the previous time step from its internal state register, using recursive formulas... Calculate the predicted state value at the current moment, and then output the matrix. The predicted output value is obtained by projection, and the calculated current state vector is written back to the state register to form a recursive update, realizing a prediction loop every 1 millisecond.

[0129] The residual calculator connects to the sensor interface and the physical model module, simultaneously reading the timestamped measured values ​​and predicted output values. The residual signal is obtained through fixed-point subtraction. The residual calculator has built-in peak detection logic to continuously monitor whether the absolute value of the residual exceeds a preset threshold. This threshold is determined based on the statistical characteristics of multiple consecutive residual samples collected when the system is in steady state. For example, if N samples are collected under steady-state conditions and the standard deviation σ is calculated, the threshold is set to 3σ. If the absolute value of the residual exceeds the threshold, an anomaly is determined, the anomaly flag is immediately set, and the residual signal, timestamp, and anomaly flag are packaged into a data unit and sent to the input queue of the semantic encoder.

[0130] The semantic encoder comprises a quantization lookup table unit and a channel coding logic array. It obtains the Modulation and Coding Scheme (MCS) index from the downlink control channel parsing of the base station via an internal bus from the 5G-A communication module. The semantic encoder pre-stores a coding strategy table, mapping different MCS index ranges to corresponding quantization coding methods. When the MCS index corresponds to a high channel quality level, 8-bit uniform quantization is used, with a fixed quantization step size of 1 / 256 of the residual signal's full scale. When the MCS index corresponds to a low channel quality level, 4-bit non-uniform quantization is used. Its quantization range is obtained offline through Lloyd-Max algorithm training and pre-stored in a quantization table. Simultaneously, Hamming codes are appended to the quantized codewords for forward error correction. When the anomaly flag is valid, the semantic encoder forcibly ignores the MCS index and directly uses 8-bit uniform quantization to retain more details, setting a high-priority flag in the encoded frame header. The quantized data, timestamp, and anomaly flag are combined to form an encoded frame. The frame structure includes a frame start flag, a unique terminal identifier, a sequence number, a microsecond-level timestamp, a quantized data field, an anomaly flag, and a Cyclic Redundancy Check (CRC) field.

[0131] The 5G-A communication module connects to the semantic encoder via an internal bus. After receiving the encoded frame, the baseband processor performs physical layer processing, including channel coding, modulation, and resource mapping, and finally transmits it to the 5G-A base station via the antenna through the radio frequency front end. After the high-priority flag in the encoded frame is recognized by the base station MAC layer scheduler, an emergency scheduling process is triggered to allocate ultra-reliable low-latency communication (URLLC) resources to the data acquisition terminal for priority transmission. This includes allocating at least two consecutive resource blocks, using the highest available modulation order, and compressing the transmission time interval to 0.125 milliseconds.

[0132] The system also includes edge servers connected to 5G-A base stations. These edge servers maintain a discretized state-space model parameter library synchronized with each data acquisition terminal. The data in the parameter library is updated synchronously with the terminals via the 5G core network. The synchronization process uses a bidirectional Transport Layer Security (TLS) protocol based on X.509 certificates to establish an encrypted channel. Triggering conditions include the arrival of a timed period or changes in model parameters due to terminal firmware upgrades. The edge server receives encoded frames forwarded by the base station, performs channel decoding and inverse quantization to recover the residual signal, and simultaneously obtains control commands corresponding to the timestamps of the encoded frames from the core network. It then uses locally stored model parameters to run a physical model identical to that on the terminal side, generating predicted output values ​​at the same time. Finally, the predicted output values ​​are added to the recovered residual signal to reconstruct the original measured hydraulic state values, which are stored in time-series database records in timestamp order. These records are then pushed in real-time to upper-layer applications via the gRPC interface for status monitoring, fault diagnosis, or data visualization.

[0133] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects.

[0134] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment includes only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A low-latency 5G-A communication method for hydraulic data acquisition, characterized in that, Includes the following steps: In the data acquisition terminal, a discretized state-space model based on the first principles of the hydraulic system is pre-built and solidified. The model includes a system matrix, an input matrix, and an output matrix, which is used to predict the system state at the current moment based on the system state vector at the previous moment and the current control command. The measured values ​​of the hydraulic system's state are acquired in real time and compared with the predicted output values ​​generated by the discretized state-space model based on the control command at the same moment to obtain the residual signal. Obtain the modulation and coding scheme (MCS) index from the 5G-A base station. Based on the channel quality level corresponding to the MCS index, dynamically select the appropriate quantization and coding method from the pre-stored coding strategy table, quantize and encode the residual signal, and generate a coded frame. When the amplitude of the residual signal exceeds the preset threshold, a high priority flag is set in the coded frame, and a coding method with a higher quantization precision than the current channel quality level is forced to be used. The coded frame is sent to the 5G-A base station through the 5G-A communication module, and the high priority flag triggers the base station to allocate ultra-reliable low-latency communication (URLLC) resources for priority transmission.

2. The low-latency 5G-A communication method for hydraulic data acquisition according to claim 1, characterized in that, The process of constructing the discretized state-space model includes: A continuous-time nonlinear dynamic model of a hydraulic system is established based on the fundamental laws of fluid mechanics and dynamics. The nonlinear dynamic model includes at least the pressure dynamic equation, the kinematic equation, and the flow continuity equation. The nonlinear dynamic model is transformed into a linear state-space model by using a small-deviation linearization method near a preset typical operating point. The linear state-space model is discretized using a zero-order hold method with a step size matching the system sampling period to obtain the system matrix, input matrix, and output matrix. The matrix coefficients are then tuned through parameter identification experiments or online learning algorithms.

3. The low-latency 5G-A communication method for hydraulic data acquisition according to claim 1, characterized in that, The steps for generating the predicted output value include: The current control instructions are obtained in real time from the programmable logic controller (PLC), and the normalized control instructions are used as input to the discretized state-space model. Simultaneously, the system state vector of the previous moment is read, and the state prediction value of the current moment is calculated recursively through the system matrix and the input matrix. Then, the state prediction value is linearly projected onto the output matrix to obtain the predicted output value. Among them, the control command is used as the feedforward input model to eliminate the influence of state fluctuations caused by normal operation on the residual signal.

4. The low-latency 5G-A communication method for hydraulic data acquisition according to claim 1, characterized in that, The preset threshold is set based on the statistical characteristics of multiple consecutive residual samples collected when the system is in a steady state, specifically including: Under steady-state conditions, N residual samples are collected, their sample standard deviation σ is calculated, and a preset threshold is set to 3σ, where the value of N is determined according to the dynamic characteristics of the system.

5. The low-latency 5G-A communication method for hydraulic data acquisition according to claim 1, characterized in that, The encoding strategy table contains the mapping relationship between different MCS index ranges and quantization encoding methods; When the MCS index corresponds to a high channel quality level, 8-bit uniform quantization is used, and the quantization step size is fixed at 1 / 256 of the full scale of the residual signal. When the MCS index corresponds to a low channel quality level, 4-bit non-uniform quantization is used. The quantization interval of the non-uniform quantization is obtained offline by training the Lloyd-Max algorithm based on the probability density function of the residual signal and is pre-stored as a quantization table. At the same time, Hamming codes are added to the quantized codewords for forward error correction.

6. The low-latency 5G-A communication method for hydraulic data acquisition according to claim 1, characterized in that, The high-priority flag triggers the base station to allocate URLLC resources, including: After detecting the high priority flag, the base station scheduler allocates at least two consecutive resource blocks (RBs) to the data acquisition terminal and uses the highest available modulation order. At the same time, it configures the physical downlink shared channel (PDSCH) to URLLC mode, where the subcarrier spacing is set to 120 kHz and the transmission time interval (TTI) is compressed to 0.125 milliseconds.

7. The low-latency 5G-A communication method for hydraulic data acquisition according to claim 1, characterized in that, The method also includes a data reconstruction step on the edge server side: The edge server pre-maintains a discretized state-space model parameter library corresponding to each data acquisition terminal; Receive the encoded frames forwarded by the base station, and perform channel decoding and inverse quantization operations on the encoded frames to recover the residual signal; At the same time, control instructions corresponding to the timestamp of the encoded frame are obtained from the core network, and the physical model synchronized with the terminal side is run using the locally stored model parameters to generate the predicted output value at the same time. The predicted output value is added to the recovered residual signal to reconstruct the original measured hydraulic state value, which is then stored as a time-series database record in chronological order according to timestamps.

8. The low-latency 5G-A communication method for hydraulic data acquisition according to claim 7, characterized in that, The model parameter library is kept synchronized with the data acquisition terminal through the 5G core network. The synchronization process uses a bidirectional transport layer security protocol (TLS) based on X.509 certificates to establish an encrypted channel. The triggering conditions include the arrival of a timed period or changes in model parameters caused by terminal firmware upgrades.

9. The low-latency 5G-A communication method for hydraulic data acquisition according to claim 1, characterized in that, The frame structure of the encoded frame includes a frame start flag, a terminal unique identifier, a sequence number, a microsecond-level timestamp, a quantized residual data field, an anomaly flag, and a cyclic redundancy check (CRC) field. The exception flag is used to indicate whether the current residual signal exceeds a preset threshold.

10. A low-latency 5G-A communication system for hydraulic data acquisition, used to implement the low-latency 5G-A communication method for hydraulic data acquisition as described in any one of claims 1 to 9, characterized in that, include: The data acquisition terminal is deployed on the hydraulic equipment side. The data acquisition terminal integrates a sensor interface, a physical model module, a residual calculator, a semantic encoder, and a 5G-A communication module. The sensor interface is used to connect with external sensors to acquire hydraulic state analog signals in real time, perform analog-to-digital conversion, and output digital measured values ​​with timestamps. The physical model module is implemented using an embedded microcontroller. Its program memory contains the system matrix, input matrix, and output matrix of a discretized state-space model based on the first principles of the hydraulic system. This matrix is ​​used to recursively predict the current state value based on the system state vector of the previous moment and the current control command, and then output the predicted output value. The residual calculator connects to the sensor interface and physical model module to calculate the difference between the measured value and the predicted output value at the same time as the residual signal, and monitors whether the amplitude of the residual signal exceeds the preset threshold. When it exceeds the threshold, the abnormal flag is set. The semantic encoder includes a quantization lookup table unit and a channel coding logic array. It is used to dynamically select a quantization coding strategy based on the modulation and coding scheme (MCS) index received from the downlink control channel of the 5G-A base station, quantize and encode the residual signal to generate a coded frame, and force the use of high-precision coding and set a high-priority flag when the abnormal flag bit is valid. The 5G-A communication module is connected to the semantic encoder via an internal bus to send the encoded frame to the 5G-A base station through the 5G-A wireless air interface, and uses a high-priority flag to trigger the base station to allocate ultra-reliable low-latency communication (URLLC) resources for priority transmission. The system also includes an edge server connected to a 5G-A base station. The edge server maintains a discretized state-space model parameter library synchronized with each data acquisition terminal, which is used to receive encoded frames forwarded by the base station and reconstruct the original hydraulic state data.

Images