Intelligent negative pressure control clamping arm based on internet of things sensor and remote monitoring system

Abstract

This invention discloses an intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors, belonging to the field of industrial automation control technology. Addressing the problems of delayed leakage response, simplified control strategies, and lack of remote safety monitoring in existing negative pressure clamping systems, this invention uses an extended Kalman filter to fuse pressure data from both the adsorption chamber and the vacuum pump outlet pipe. It estimates the equivalent area of ​​the leakage orifice in real time, and combines temperature and vibration sensors to classify operating modes and match adaptive leakage thresholds. When a leakage warning is triggered, the vacuum pump speed is adjusted based on the leakage area and the rate of change of negative pressure, calculating feedforward compensation. An emergency stop and maximum power operation are executed upon alarm. Simultaneously, a digital twin synchronizes the physical state in real time, and bidirectional digital signatures ensure secure communication between the edge controller and the cloud platform. This invention achieves early proactive leakage compensation, adaptive operating mode control, and safe remote monitoring, significantly reducing the risk of workpiece detachment.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, and in particular to an intelligent negative pressure control clamping arm and remote monitoring system based on Internet of Things (IoT) sensors. Background Technology

[0002] Negative pressure clamping technology is a clamping method that uses the principle of vacuum negative pressure to adsorb and fix workpieces to the end of clamping arms. It is widely used in workpiece handling, positioning, and processing processes in automated production lines. Its core principle is to generate a negative pressure difference relative to atmospheric pressure within the adsorption chamber using a vacuum pump or negative pressure generator, thereby creating a normal clamping force between the adsorption surface and the surface of the workpiece, thus fixing the workpiece. Compared with mechanical grippers, negative pressure clamping has advantages such as simple structure, less damage to the workpiece surface, and strong adaptability to irregularly shaped workpieces, making it irreplaceable in fields such as electronic component assembly, glass panel handling, and thin-walled part processing.

[0003] However, existing negative pressure clamping systems have several technical problems that have not yet been effectively solved in practical applications.

[0004] The first problem lies in the delayed response to leaks. During the continuous operation of automated production lines, micro-leakage often occurs between the suction surface of the negative pressure clamp and the surface of the workpiece due to factors such as microscopic unevenness of the workpiece surface, intrusion of foreign objects, or mechanical vibration. In the initial stage, this micro-leakage manifests as a slow decrease in negative pressure. When the negative pressure continues to drop below a certain critical threshold, the clamping force becomes insufficient to maintain the stable fixation of the workpiece, causing it to slip or even fall off during handling or processing. Existing negative pressure control systems typically employ a simple threshold comparison method—that is, real-time monitoring of the negative pressure value and comparison with a preset safety threshold. Once the negative pressure value falls below the safety threshold, an alarm or shutdown is triggered. This method is a passive response mode, only reacting when the leak has developed to a dangerous level, failing to identify and intervene in the early stages of leakage. In reality, there is often an observable transition period between the onset of micro-leakage and the negative pressure value falling below the safety threshold. Existing threshold comparison methods completely fail to utilize the leakage trend information contained in the negative pressure signal during this transition period.

[0005] The second problem lies in the oversimplification of control strategies. Even some more advanced negative pressure control devices with pressure feedback regulation functions typically limit their control methods to simple on / off control or proportional-integral-derivative (PI-DE) control. These control methods rely on preset fixed parameters and struggle to adapt to changes in working conditions such as workpiece material, surface roughness, ambient temperature, and humidity on the leakage rate. When working conditions change, the fixed-parameter control strategy cannot automatically adjust its response characteristics, leading to a significant decrease in control effectiveness. For example, when switching from adsorbing a smooth metal workpiece to a rough composite material workpiece, the characteristics of micro-leakage change. The fixed-parameter PI-DE controller cannot recognize this change and adjust the control parameters accordingly, resulting in either an excessively slow response or unnecessary overcompensation.

[0006] The third problem lies in the lack of effective remote status awareness and safety control measures. In smart manufacturing scenarios, production managers need to monitor the real-time operating status of the negative pressure grippers at each workstation. However, existing negative pressure control devices typically only provide local indications or simple fieldbus communication interfaces, lacking the capability for remote monitoring over a wide area network. Even if some systems possess basic remote data reporting functions, they often lack security protection mechanisms for communication data, facing the security risks of data tampering or control commands being forged. In the industrial internet environment, the integrity and authenticity of control commands are directly related to production safety. Once uncertified control commands are injected into the system, they may cause the grippers to suddenly release during workpiece clamping, resulting in workpiece damage or even personal injury.

[0007] In summary, existing negative pressure clamping control systems have significant shortcomings in three dimensions: early leakage detection and prediction, adaptive control parameter adjustment, and remote safety monitoring. There is an urgent need for a technical solution that can organically integrate these three dimensions and achieve synergistic optimization. Summary of the Invention

[0008] To achieve the above objectives, the present invention provides an intelligent negative pressure control clamp arm and remote monitoring system based on an Internet of Things (IoT) sensor, comprising: At least one negative pressure control clamping arm unit, the negative pressure control clamping arm unit including a clamping arm body, an adsorption chamber, a vacuum pump, a servo drive module, a first negative pressure sensor, a second negative pressure sensor, a temperature sensor, and a vibration sensor; the first negative pressure sensor is disposed on the inner wall of the adsorption chamber near the adsorption port, and is used to collect a first negative pressure value time series; the second negative pressure sensor is disposed in the vacuum pump outlet pipeline, and is used to collect a second negative pressure value time series; the temperature sensor is disposed on the outer wall of the adsorption chamber, and is used to collect a temperature value time series; the vibration sensor is disposed at the base of the clamping arm body, and is used to collect a vibration acceleration amplitude time series. An edge computing controller is communicatively connected to the first negative pressure sensor, the second negative pressure sensor, the temperature sensor, and the vibration sensor. The edge computing controller is configured as follows: The received first negative pressure value time series and second negative pressure value time series are subjected to outlier removal and low-pass filtering preprocessing to obtain the preprocessed first negative pressure value time series and the preprocessed second negative pressure value time series; The preprocessed first negative pressure value time series and the preprocessed second negative pressure value time series are input into an extended Kalman filter. The state variables of the extended Kalman filter include the equivalent negative pressure estimate, the negative pressure change rate, and the equivalent area of ​​the leakage orifice. The observations of the extended Kalman filter include the preprocessed first negative pressure value and the preprocessed second negative pressure value. The extended Kalman filter outputs the equivalent area estimate of the leakage orifice and the negative pressure change rate. The temperature value time series and the vibration acceleration amplitude time series are input into a pre-trained working condition mode classifier. The working condition mode classifier outputs the current working condition mode identifier based on the mean temperature and root mean square value of vibration acceleration amplitude within the current time window. The estimated equivalent area of ​​the leakage orifice is compared with a pre-stored leakage threshold table corresponding to the current operating condition mode identifier. When the estimated equivalent area of ​​the leakage orifice is greater than the first-level warning threshold and less than or equal to the second-level alarm threshold, a warning status flag is generated. When the estimated equivalent area of ​​the leakage orifice is greater than the second-level alarm threshold, an alarm status flag is generated. The control command generation module is used to, when the warning status flag is generated, calculate the estimated equivalent area of ​​the leakage orifice and the rate of change of negative pressure according to the formula... Calculate the feedforward compensation increment, add the feedforward compensation increment to the current negative pressure setpoint, and generate a vacuum pump speed adjustment command. This is an estimate of the equivalent area of ​​the leakage orifice. The rate of change of the negative pressure. This is the leakage area compensation coefficient. The rate compensation coefficient is used; the control command generation module is also used to generate an emergency stop command and a vacuum pump maximum power operation command when the alarm status flag is generated, the emergency stop command is used to control the servo drive module to stop moving; A cloud platform server is provided, and a secure communication link with the edge computing controller is established between the cloud platform server and the edge computing controller. The cloud platform server is used to receive status data uploaded by the edge computing controller. The remote monitoring terminal is used to receive status data forwarded by the cloud platform server and display the running interface, as well as to receive control commands input by the operator and send the control commands to the edge computing controller via the cloud platform server.

[0009] Preferably, the sampling frequency of the first negative pressure sensor is set to 50Hz to 200Hz, the sampling frequency of the second negative pressure sensor is set to 10Hz to 50Hz, the sampling frequency of the temperature sensor is set to 1Hz, and the sampling frequency of the vibration sensor is set to 100Hz to 500Hz. The negative pressure control clamping arm unit further includes a data acquisition module, which is electrically connected to the first negative pressure sensor, the second negative pressure sensor, the temperature sensor, and the vibration sensor. The data acquisition module is used to read the data collected by each sensor at a fixed period, attach a timestamp synchronized based on the NTP protocol to each sampling data point, encapsulate the timestamp-aligned data within the same time window into Ethernet frames, and send them to the edge computing controller via the EtherNet / IP protocol. The edge computing controller receives the Ethernet frames, unpacks them, and recovers the time series of each sensor based on the timestamps.

[0010] Preferably, the specific process of the edge computing controller performing outlier removal and low-pass filtering preprocessing on the received first negative pressure value time series and second negative pressure value time series includes: using a sliding window method, with the window width set to the number of sampling points obtained by multiplying the sampling frequency by 0.5 seconds, and the sliding step size set to half the window width; for each data point within the window, calculating the median and median absolute deviation of the data points within the window, identifying data points that deviate from the median by more than 3 times the median absolute deviation as outliers and removing them; filling the removed data points using linear interpolation of the previous and next valid values ​​of the removed data points; and smoothing the outlier-removed and filled sequences using an infinite impulse response low-pass filter with a cutoff frequency of 5Hz to obtain the preprocessed first negative pressure value time series and the preprocessed second negative pressure value time series.

[0011] Preferably, the state vector of the extended Kalman filter is defined as follows: ,in This is an estimate of the equivalent negative pressure. The rate of change of negative pressure. The equivalent area of ​​the leakage orifice; the state transition equation uses a nonlinear function. , where nonlinear functions Based on the dynamic differential equation of the pressure in the adsorption chamber Discretization yields, The natural leakage coefficient, Leakage flow coefficient, Let be the process noise vector, which follows a zero-mean Gaussian distribution, and its covariance matrix is... The observation vector is defined as follows: ,in This refers to the first negative pressure value after pretreatment. The second negative pressure value after preprocessing is given; the observation equation is... The observation matrix , which is the static gain ratio between the first negative voltage value and the second negative voltage value. The coefficient representing the impact of leakage on the second negative pressure value. The observed noise vector follows a zero-mean Gaussian distribution, and its covariance matrix is... The extended Kalman filter recursively calculates the estimated equivalent area of ​​the leakage orifice through time update and measurement update steps. and the rate of change of negative pressure .

[0012] Preferably, the operating condition mode classifier employs a pre-trained support vector machine (SVM) model. The input feature vector of the SVM model consists of the average temperature and the root mean square value of vibration acceleration amplitude within the current time window. The SVM model uses a radial basis function (RBF) kernel function. The training dataset includes calibrated samples of average temperature and root mean square value of vibration acceleration amplitude under known operating conditions. During training, the penalty parameter and kernel function parameter are determined through cross-validation. The operating condition mode categories output by the operating condition mode classifier include "normal temperature-low vibration" mode, "normal temperature-high vibration" mode, and so on. The system includes "high temperature-low vibration" mode and "high temperature-high vibration" mode. The pre-stored leakage threshold table is configured with corresponding first-level warning thresholds and second-level alarm thresholds for each operating condition mode category. The first-level warning thresholds for "high temperature-low vibration" mode, "normal temperature-high vibration" mode, and "high temperature-high vibration" mode are all lower than the first-level warning threshold for "normal temperature-low vibration" mode. The threshold for judging high or low temperature is a preset temperature value, and the threshold for judging vibration intensity is a preset root mean square value of vibration acceleration amplitude.

[0013] Preferably, the control command generation module is further configured to obtain the corresponding leakage area compensation coefficient from a pre-established operating condition-coefficient mapping table based on the current operating condition mode identifier. and the rate compensation coefficient In the operating condition-coefficient mapping table, the leakage area compensation coefficient for the "normal temperature-low vibration" mode is set to [value missing]. The leakage area compensation coefficient for the "normal temperature-high vibration" mode is set to [value]. The leakage area compensation coefficient for the "high temperature-low vibration" mode is set to [value]. The leakage area compensation coefficient for the "high temperature-high vibration" mode is taken as: In the operating condition-coefficient mapping table, the rate compensation coefficient corresponding to the "normal temperature-low vibration" mode is set to a value of [value missing]. The rate compensation coefficient for the "normal temperature - high vibration" mode is set to 1. The rate compensation coefficient for the "high temperature-low vibration" mode is set to 1. The rate compensation coefficient for the "high temperature-high vibration" mode is set to 1. .

[0014] Preferably, the edge computing controller further comprises a digital twin model unit, which is used to construct a digital twin of the negative pressure control clamping arm unit; the digital twin receives the equivalent negative pressure estimate output by the extended Kalman filter. The estimated equivalent area of ​​the leakage orifice. The rate of change of negative pressure The current operating mode identifier, the average temperature, and the root mean square value of the vibration acceleration amplitude are used to update the virtual negative pressure value and virtual leakage status parameters within the virtual adsorption cavity of the digital twin, synchronizing the operating status of the digital twin with the real-time physical status of the negative pressure control clamping arm unit. A cloud copy of the digital twin is stored in the cloud platform server, which updates the cloud copy synchronously after receiving status data uploaded by the edge computing controller. The remote monitoring terminal displays a three-dimensional visualization interface of the digital twin by accessing the cloud copy. This three-dimensional visualization interface includes a real-time negative pressure curve, a trend chart of the equivalent area change of the leakage orifice diameter, and early warning alarm status indicators.

[0015] Preferably, the two-way authentication mechanism of the secure communication link specifically includes: the edge computing controller has a built-in first security chip, which stores the edge computing controller's private key and the cloud platform server's public key; the cloud platform server has a built-in second security chip, which stores the cloud platform server's private key and the edge computing controller's public key; before sending status data to the cloud platform server, the edge computing controller uses its private key to generate a digital signature for the status data packet using the SHA-256 hash algorithm and the elliptic curve digital signature algorithm, and attaches the digital signature to the status data packet before sending it; after receiving the data, the cloud platform server uses the edge computing controller's public key to verify the digital signature, and if the verification is successful, it confirms that the source of the status data packet is legitimate; before issuing control commands to the edge computing controller, the cloud platform server uses its private key to generate a digital signature for the control command using the SHA-256 hash algorithm and the elliptic curve digital signature algorithm, and attaches the digital signature to the control command before sending it; after receiving the control command, the edge computing controller uses the cloud platform server's public key to verify the digital signature, and if the verification is successful, it executes the control command.

[0016] Preferably, when the warning status flag is generated, the edge computing controller packages the estimated equivalent area of ​​the leakage orifice, the rate of change of negative pressure, the calculated feedforward compensation increment, and the adjusted vacuum pump speed command into a warning event record, and uploads it to the cloud platform server through the secure communication link; after receiving the warning event record, the remote monitoring terminal pops up a warning prompt box on the display interface and enables an audible alarm; the remote monitoring terminal provides a manual intervention interface, through which the operator inputs a confirmation command or a modified negative pressure setting value, and the confirmation command or the modified negative pressure setting value is sent to the edge computing controller after being signed by the cloud platform server using the cloud platform server's private key, and the edge computing controller executes it after the signature is verified.

[0017] The beneficial effects of this invention are: 1. This invention collects dual-channel pressure data by using a first negative pressure sensor located on the inner wall of the adsorption chamber near the adsorption port and a second negative pressure sensor located in the vacuum pump outlet pipeline. It utilizes an extended Kalman filter to estimate the equivalent area of ​​the leakage orifice as a latent state variable in real time. This allows for accurate identification of micro-leakage and its severity before significant decay of the negative pressure value. Based on the estimated equivalent area of ​​the leakage orifice and the rate of negative pressure change, a feedforward compensation increment is generated, driving the vacuum pump to adjust its speed in advance to proactively compensate for the negative pressure decay caused by the leakage. This transforms the leakage response time from the traditional threshold-triggered post-remediation mode to a proactive compensation mode based on online estimation of the leakage state, significantly shortening the leakage response time and effectively preventing workpieces from falling off due to insufficient negative pressure during handling or processing.

[0018] 2. This invention introduces temperature and vibration sensors into the edge computing controller, uses a pre-trained support vector machine model to classify the current operating conditions in real time, and configures differentiated leakage warning thresholds and compensation coefficients for different operating conditions. This allows the sensitivity of leakage detection and the strength of compensation control to be automatically adjusted according to changes in operating conditions such as temperature and vibration. This solves the problem that fixed thresholds and fixed control parameters cannot balance response speed and false alarm rate under different operating conditions, and realizes intelligent negative pressure control that is adaptive to operating conditions.

[0019] 3. This invention establishes a two-way authentication secure communication link between the edge computing controller and the cloud platform server based on a hardware security chip and an elliptic curve digital signature algorithm. All status data and control commands must be digitally signed before transmission, and the receiving end can only process them after the signature is verified. This effectively prevents the security risks of control commands being forged or tampered with, ensures the integrity and authenticity of remote monitoring commands, and enables the system to meet the secure remote monitoring needs in the industrial internet environment. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a logic block diagram of the system of the present invention. Detailed Implementation

[0022] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0023] Please see Figure 1 This invention provides an intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors. The core inventive concept lies in organically integrating five technical means into a collaborative closed-loop system: extended Kalman filtering fusion of multi-source heterogeneous sensor data, adaptive threshold judgment based on operating condition pattern recognition, feedforward compensation control based on online leakage state estimation, real-time synchronous mapping of a digital twin model, and secure remote communication with two-way authentication. The implementation details of the above technical means are described in detail below with reference to specific embodiments.

[0024] Example This embodiment uses the glass panel handling process in an automated automotive parts production line as an application scenario. In this process, the negative pressure-controlled gripper unit at the end effector of an industrial robot needs to adsorb the glass panel of the car's central control screen and transport it from the material tray to the processing station. The glass panel measures 300 mm x 200 mm x 1.2 mm, has a polished glass surface, and weighs approximately 180 grams. During the handling process, the maximum acceleration of the robot's end effector is 5 m / s². Based on physical mechanics calculations, the gripping force needs to provide a normal adsorption force greater than 2.7 Newtons to simultaneously overcome the panel's weight and the inertia force during handling. Considering the safety factor, the negative pressure setting value under normal operating conditions is determined to be approximately 12 Newtons of theoretical adsorption force when the negative pressure reaches -55 kPa, meeting the required safety margin.

[0025] I. On-site equipment layer configuration The gripper body of the negative pressure control gripper unit is made of aluminum alloy, with a 40 mm diameter circular adsorption chamber installed at the end. The internal volume of the adsorption chamber is approximately 25 cubic centimeters. The vacuum pump is a miniature rotary vane vacuum pump driven by a brushless DC motor, with a maximum pumping speed of 8 liters per minute and an ultimate vacuum of -85 kPa. The servo drive module is integrated with the joint servo controller of the industrial robot, receiving motion commands to drive the gripper arm to move in three-dimensional space.

[0026] The first negative pressure sensor is installed on the inner wall of the adsorption chamber near the adsorption port, specifically in a sensor mounting hole located 5 mm inward from the adsorption port. The first negative pressure sensor is a piezoresistive microelectromechanical system (MEMS) absolute pressure sensor with a range of -100 kPa to 0 kPa, an accuracy of 0.25% of the range, and a response time of less than 1 millisecond. The sampling frequency of the first negative pressure sensor is set to 200 Hz, meaning a first negative pressure value data point is collected every 5 milliseconds to ensure the high-frequency fluctuation characteristics of the pressure within the adsorption chamber are captured. The collected data is a time series of the first negative pressure values.

[0027] The second negative pressure sensor is installed in the vacuum pump outlet pipe, specifically in a sensor mounting hole located 10 mm downstream of the vacuum pump exhaust port on the inner wall of the pipe. The inner diameter of the vacuum pump outlet pipe is 4 mm. The second negative pressure sensor is a piezoresistive MEMS absolute pressure sensor of the same model as the first negative pressure sensor, with the same range and accuracy. The sampling frequency of the second negative pressure sensor is set to 40 Hz, meaning that one second negative pressure value data point is acquired every 25 milliseconds. The sampling frequency of the second negative pressure sensor is lower than that of the first negative pressure sensor because the pressure fluctuation in the vacuum pump outlet pipe is relatively slow. The lower sampling frequency is sufficient to effectively capture the pressure change trend in the pipe, while reducing the processing load of the data acquisition module and the data throughput pressure of the edge computing controller. The acquired data is a time series of the second negative pressure values.

[0028] The temperature sensor is installed on the outer wall of the adsorption chamber, specifically at the middle of the cylindrical outer wall. It is a Pt100 platinum resistance temperature sensor with a range of -20°C to 100°C and an accuracy of ±0.15°C. The sampling frequency of the temperature sensor is set to 1 Hz, meaning it collects one temperature data point per second. The temperature sensor monitors temperature changes on the outer wall of the adsorption chamber. Changes in ambient temperature affect the physical properties of the gas inside the adsorption chamber and the elastic modulus of the sealing material, thus affecting leakage characteristics. Therefore, temperature information needs to be incorporated into the judgment of the operating mode. The collected data is a time series of temperature values.

[0029] The vibration sensor is mounted on the base of the gripper arm, specifically on the back of the connection surface between the gripper arm and the end flange of the industrial robot. It employs a triaxial microelectromechanical system (MEMS) accelerometer with a range of ±16 times the force of gravity, a frequency response range of 0 Hz to 1 kHz, and a sampling frequency of 500 Hz, meaning it acquires one triaxial vibration acceleration data point every 2 milliseconds. The vibration sensor collects instantaneous vibration acceleration values ​​along three orthogonal axes and calculates the resultant vibration acceleration amplitude through built-in vector synthesis. The acquired data is a time series of vibration acceleration amplitude values.

[0030] The data acquisition module is an embedded acquisition board integrated into the electrical cabinet of the negative pressure control clamping arm unit. It is based on a field-programmable gate array (FPGA) and an ARM Cortex-M7 dual-core architecture. The data acquisition module has four independent analog-to-digital conversion channels, which are electrically connected to the first negative pressure sensor, the second negative pressure sensor, the temperature sensor, and the vibration sensor via shielded twisted-pair cables. The data acquisition module has a built-in GPS timing module and an NTP protocol stack. It obtains the reference time from the time server in the production line's local area network using the NTP protocol and adds a timestamp to each sampled data point with millisecond-level accuracy. In each data acquisition cycle, the data acquisition module uses a first sampling frequency as the reference cycle. At the end of each reference cycle, it encapsulates the latest available sampled value from the four channels at that moment, along with the timestamps of each sensor, into a data frame. Each data frame contains the following fields: a 2-byte frame header identifier, an 8-byte timestamp, a 4-byte first negative pressure value, a 4-byte second negative pressure value, a 2-byte temperature value, a 4-byte vibration acceleration amplitude, and a 2-byte checksum, totaling 24 bytes. The data acquisition module sends the encapsulated data frames to the edge computing controller via EtherNet / IP industrial Ethernet protocol in a multicast manner. The sending period is consistent with the period of the first sampling frequency, that is, one frame is sent every 5 milliseconds.

[0031] II. Edge Control Layer Configuration The edge computing controller is deployed in an edge industrial computer on the production line. This industrial computer is equipped with an Intel Core i7-12700H processor, 32GB of memory, and a 1TB solid-state drive, running the Ubuntu 20.04 LTS real-time operating system. The edge computing controller contains a signal preprocessing unit, a leakage state assessment unit, a digital twin model unit, and a control command generation unit. These four functional units run as independent software processes, exchanging data between processes via shared memory to ensure data transmission latency of less than 100 microseconds.

[0032] Signal preprocessing unit: The signal preprocessing unit receives Ethernet frames from the data acquisition module, unpacks them, and reconstructs the time series of the first negative pressure value, the second negative pressure value, the temperature value, and the vibration acceleration amplitude. The signal preprocessing unit performs outlier removal and low-pass filtering preprocessing on the first and second negative pressure value time series. The preprocessing process is as follows: First, outlier removal is performed. A sliding window approach is used, with the width of the sliding window calculated based on the sampling frequency of the first negative pressure sensor (200 Hz) and a preset time window length of 0.5 seconds, resulting in 100 sampling points. The step size of the sliding window is set to half the window width, i.e., 50 sampling points, with a 50-point overlap between adjacent windows to ensure the continuity of outlier removal at the window edges. For the 100 first negative pressure data points within the current sliding window, the median is calculated and denoted as M. Then, the absolute deviation of each data point from the median is calculated, and the median of these 100 absolute deviations is obtained, denoted as _____. For any data point within the window If the conditions are met Then the data point Outliers are marked as such. The median and median absolute deviation are used here instead of the mean and standard deviation for outlier detection because the median and median absolute deviation are more robust to outliers—individual extreme outliers will not significantly skew the median's position, thus avoiding the mislabeling of normal data points due to outlier contamination. Data points marked as outliers are removed from the sequence, and the value at the previous valid time step of that outlier data point is used. and the value of the next effective time step According to the linear interpolation formula Calculate the imputation value and fill it into the removed position. If an outlier appears at the very beginning or end of the sequence, resulting in a lack of preceding or succeeding valid values, then the nearest valid value is used to directly fill the gap. The same outlier removal process is used for the second negative pressure value time series, the only difference being that the sampling frequency of the second negative pressure value is 40 Hz, the corresponding sliding window width is 20 sampling points, and the sliding window step size is 10 sampling points.

[0033] The sequence after outlier removal and imputation was then smoothed using a low-pass filter. A second-order infinite impulse response Butterworth low-pass filter was used, with a cutoff frequency of 5 Hz. This cutoff frequency was chosen because: under normal operating conditions, the change in negative pressure within the adsorption chamber is determined by the speed regulation of the vacuum pump and the microstructure of the workpiece adsorption surface; its effective signal spectrum is mainly distributed below 5 Hz, while sensor noise and electromagnetic interference are mainly concentrated in higher frequency bands. At a sampling frequency of 200 Hz, the 5 Hz cutoff frequency corresponds to a normalized cutoff frequency of 0.05. The coefficients of the filter's difference equation were obtained from the analog prototype filter design using the bilinear transform method; specifically, the feedforward coefficients... , , Feedback coefficient , The filter recursively calculates the input sequence point by point to obtain the filtered output sequence, resulting in the preprocessed first negative voltage time series. The second negative voltage time series is then processed using a low-pass filter with the same structure and a cutoff frequency of 5 Hz to obtain the preprocessed second negative voltage time series.

[0034] Leakage Status Assessment Unit: The leakage condition assessment unit receives the pre-processed first negative pressure value time series, the pre-processed second negative pressure value time series, the temperature value time series, and the vibration acceleration amplitude time series, and performs the following steps.

[0035] Step 1: Data Preparation. The leakage condition assessment unit maintains a first-in, first-out (FIFO) data buffer with a buffer length of 2 seconds of data. For the preprocessed first negative pressure value time series, the buffer stores 400 data points; for the preprocessed second negative pressure value time series, the buffer stores 80 data points; for the temperature value time series, the buffer stores 2 data points; and for the vibration acceleration amplitude time series, the buffer stores 1000 data points. At the end of each calculation cycle, the leakage condition assessment unit retrieves the latest complete dataset from the buffer for calculation. The calculation cycle is set to 50 milliseconds, meaning a complete leakage condition assessment process is executed every 50 milliseconds.

[0036] Step 2: Extended Kalman Filter Data Fusion. The current data values ​​from the preprocessed first negative pressure value time series and the preprocessed second negative pressure value time series are input into an extended Kalman filter for data fusion. The state vector of the extended Kalman filter is defined as a three-dimensional vector:

[0037] in, This is the estimated equivalent negative pressure inside the adsorption chamber at the current moment, in kilopascals. The rate of change of negative pressure is the first derivative of negative pressure with respect to time, and its unit is kilopascals per second. This refers to the equivalent area of ​​the leaking orifice, expressed in square millimeters. It's important to note that this is the equivalent area of ​​the leaking orifice. This is a virtual physical quantity. It does not correspond to a real leak hole with a fixed geometry on the adsorption interface. Instead, it unifies and equates various complex microscopic leakage mechanisms—including laminar flow leakage within micro-grooves on the workpiece surface, gap leakage between the adsorption sealing ring and the workpiece surface, and permeation leakage from the adsorption cavity material itself—to the cross-sectional area of ​​a circular throttling orifice with the same gas flow characteristics. By introducing this virtual state variable, the severity of leakage, which cannot be directly measured, is transformed into a continuous variable that can be recursively calculated within a state estimation framework, thus providing a quantitative basis for subsequent feedforward compensation control. The introduction of this virtual physical quantity is an important component of the inventive concept of this invention.

[0038] The state transition equation describes the evolution of the state vector from time k-1 to time k, and takes the form of a nonlinear function:

[0039] Nonlinear function This is obtained by discretizing the dynamic differential equation of the adsorption chamber pressure. This differential equation describes the relationship between the gas pressure inside the adsorption chamber and time / leakage:

[0040] The first term on the right side of the equation This indicates the pressure drop caused by natural leakage within the adsorption chamber and piping system itself under leak-free conditions. The natural leakage coefficient is expressed in seconds. In this embodiment, its value was determined through offline calibration experiments. Natural leakage coefficient The calibration process is as follows: With the vacuum pump stopped and the adsorption port completely sealed with a smooth glass plate, record the time curve of the negative pressure value in the adsorption chamber decreasing from -60 kPa to -30 kPa. Use the least squares method to fit the solution of the above differential equation at L=0 to obtain... The estimated value. The second term on the right side of the equation. This represents the additional pressure drop caused by an equivalent leakage orifice with a leakage area of ​​L. Leakage flow coefficient, unit: In this embodiment, the value was determined through a standard leakage hole calibration experiment with a known orifice diameter. . The calibration process is as follows: A stainless steel calibration plate with known circular micropore diameters (0.05 mm, 0.10 mm, and 0.20 mm) is installed on the adsorption port. Pressure decay curves are recorded under different initial negative pressure conditions, and the results are obtained through joint calculation. Value. The differential equation is discretized using the forward Euler method, with a discretization time step of 50 milliseconds, consistent with the update period of the extended Kalman filter, resulting in the discretized state transition relationship:

[0041]

[0042]

[0043] in Seconds. Process noise vector It follows a Gaussian distribution with zero mean, and its covariance matrix is... It is a third-order diagonal matrix:

[0044] in The unit is kilopascals squared. The unit is kilopascals per second squared. The unit is squared millimeters. The values ​​of each element in the covariance matrix reflect the degree of confidence in the dynamic model of different state variables: A smaller value indicates that it is assumed that the equivalent area of ​​the leakage orifice changes slowly over a short period of time (a slowly changing leakage state is a reasonable physical assumption). The value is relatively large to accommodate rapid fluctuations in the rate of pressure change.

[0045] The observation vector is defined as a two-dimensional vector:

[0046] in The first negative pressure value after preprocessing at time k is expressed in kilopascals. Let k be the second negative pressure value after preprocessing, in kilopascals. The observation equation describes the mapping relationship between the observation vector and the state vector:

[0047] Observation matrix It is a 2x3 matrix:

[0048] The first row of the observation matrix [1,0,0] represents the negative pressure value directly measured by the first negative pressure sensor within the adsorption chamber. Therefore, the first negative pressure value directly corresponds to the equivalent negative pressure estimate in the state variables. In the second line, The static gain ratio between the first negative pressure value and the second negative pressure value is the steady-state ratio between the pressure in the vacuum pump outlet pipe and the pressure in the adsorption chamber. In this embodiment, it is determined through steady-state calibration. This indicates that there is approximately 8% pressure loss in the pipeline. The influence coefficient of leakage on the second negative pressure value is determined through experimental calibration in this embodiment. The physical meaning is that for every 1 square millimeter increase in the equivalent area of ​​the leakage orifice, the pressure in the vacuum pump outlet pipe will increase by an additional 0.15 kPa under the same adsorption chamber pressure. This is because the increased gas flow rate due to leakage leads to a corresponding increase in the vacuum pump exhaust port pressure. (Observation noise vector) It follows a Gaussian distribution with zero mean, and its covariance matrix is... It is a second-order diagonal matrix:

[0049] in and The units are all kilopascals squared. Greater than This reflects that the noise level of the pressure signal in the vacuum pump outlet pipeline measured by the second negative pressure sensor is higher than the noise level of the pressure signal inside the adsorption chamber. This is because the airflow pulsation in the vacuum pump outlet pipeline introduces additional measurement fluctuations.

[0050] The recursive calculation of the extended Kalman filter is performed alternately by a time update step and a measurement update step. The time update step calculates the state prediction value. and prediction error covariance matrix :

[0051]

[0052] in nonlinear function exist Jacobian matrix at:

[0053] Measurement update step: Calculate Kalman gain matrix Status update value and the updated error covariance matrix :

[0054]

[0055]

[0056] in This is a third-order identity matrix. After the above recursive calculations, the extended Kalman filter outputs the state estimate for the current time step in each operation cycle, from which the estimated equivalent area of ​​the leakage orifice is extracted. and negative pressure change rate .

[0057] Step 3: Operating Condition Classification. Input the data within the current time window from the temperature value time series and vibration acceleration amplitude time series into the operating condition classifier. The time window length is set to 2 seconds, consistent with the data buffer length. For the temperature value time series, there are 2 temperature data points within the current time window. Calculate the arithmetic mean of these 2 temperature data points to obtain the average temperature within the current time window, denoted as . The unit is degrees Celsius. For a vibration acceleration amplitude time series, there are 1000 vibration acceleration amplitude data points within the current time window. Calculate the root mean square value of these 1000 data points using the following formula:

[0058] Where N=1000 is the number of data points within the window. This represents the value of the i-th vibration acceleration amplitude data point, in meters per second squared.

[0059] Average temperature and the root mean square value of vibration acceleration amplitude Composition of two-dimensional input feature vectors The data is fed into a pre-trained support vector machine (SVM) model for classification. The SVM model uses a radial basis function kernel.

[0060] in The width parameter of the radial basis kernel function. For the i-th support vector, This represents the Euclidean distance. The training dataset construction process for the Support Vector Machine (SVM) model is as follows: In a real production environment, four different working condition combinations are simulated by manually controlling the ambient temperature and robot operating cycle. Temperature and vibration acceleration data are continuously collected for 30 minutes under each working condition combination, resulting in four sets of calibration datasets. The mean temperature and root mean square value of vibration acceleration amplitude are calculated for each calibration dataset, forming 400 training samples, with 100 samples corresponding to each working condition mode. A one-to-one multi-class classification strategy is used during training, and the penalty parameter C=10 is determined to be optimal through five-fold cross-validation. After training, the SVM model is serialized and stored in the solid-state drive of the edge computing controller, and loaded into memory during the initialization of the leakage state assessment unit.

[0061] The operating condition classifier outputs four operating condition mode categories: "Normal Temperature - Low Vibration," "Normal Temperature - High Vibration," "High Temperature - Low Vibration," and "High Temperature - High Vibration." The threshold for distinguishing between high and low temperatures is a preset temperature value, set to 35 degrees Celsius in this embodiment. A temperature value greater than 35 degrees Celsius is classified as high temperature, while a value less than or equal to 35 degrees Celsius is classified as normal temperature. The threshold for distinguishing between high and low vibration is a preset root mean square value of vibration acceleration amplitude, set to 2 meters per second squared in this embodiment. A vibration amplitude greater than 2 meters per second squared is classified as high vibration, while a value less than or equal to 2 meters per second squared is classified as low vibration. These two thresholds are used to cross-classify the four operating condition modes. The output of the support vector machine model is the current operating condition mode identifier. The values ​​are 1, 2, 3, and 4, which correspond to the four working modes mentioned above.

[0062] Step 4: Leakage Status Assessment. The estimated equivalent area of ​​the leaking orifice output from Step 2 is used... Corresponding operating condition mode identifiers in the pre-stored leakage threshold table The thresholds are compared. The structure of the leakage threshold table is shown in Table 1: Table 1 Leakage Threshold Table

[0063] As shown in Table 1, the first-level warning thresholds for the "high temperature-low vibration" mode, the "normal temperature-high vibration" mode, and the "high temperature-high vibration" mode are all lower than the first-level warning threshold of 0.08 square millimeters for the "normal temperature-low vibration" mode. This is because: under high temperature conditions, the sealing material of the adsorption chamber undergoes thermal expansion, increasing the microscopic gaps at the sealing contact surface, thus reducing the upper limit of the equivalent area corresponding to the permissible minor leakage state under normal operating conditions; under high vibration conditions, the periodic micro-displacement of the adsorption interface exacerbates the leakage development rate, also requiring earlier warning intervention. Therefore, using a lower warning threshold under high temperature or high vibration conditions allows feedforward compensation control to be triggered when the leakage is still in its very small stage, preventing rapid deterioration of the leakage.

[0064] The specific logic for determining the leakage status is as follows: If If the leakage is less than or equal to the first-level warning threshold, the leakage status is determined to be normal, and the system maintains its current operating state; if If the threshold is greater than the first-level warning threshold and less than or equal to the second-level alarm threshold, a warning status flag is generated; if... If the value exceeds the second-level alarm threshold, an alarm status flag is generated.

[0065] Control command generation unit: The control command generation unit receives the status flags and related status variables output by the leakage status assessment unit, and generates corresponding control commands based on the type of status flags.

[0066] When the leakage status assessment unit outputs a warning status flag, the control command generation unit reads the estimated value of the equivalent area of ​​the leakage orifice from the shared memory at the current moment. and negative pressure change rate And obtain the current operating condition mode identifier. Corresponding leakage area compensation coefficient and rate compensation coefficient The coefficient mapping relationship is shown in Table 2: Table 2. Coefficient Mapping Relationship Table

[0067] A larger compensation coefficient is used under high temperature or high vibration conditions because the leakage development trend is more intense under these conditions, requiring stronger feedforward compensation to effectively suppress the pressure drop caused by leakage.

[0068] Calculate the feedforward compensation increment using the formula:

[0069] The first item This is the leakage area ratio compensation term, which compensates for the pressure loss caused by the steady-state leakage flow rate determined by the equivalent area of ​​the leakage orifice. The larger the equivalent area of ​​the leakage orifice, the greater the required compensation. (Second term) This is the rate compensation term, which compensates for the dynamic losses during the rapid pressure drop phase of a leak. The larger the absolute value of the negative pressure change rate, the faster the leak is deteriorating, requiring additional compensation to suppress the pressure decline. The two terms are superimposed to form the feedforward compensation increment, measured in Pascals. This increment is converted to kilopascals by dividing by 1000 and then algebraically added to the current negative pressure setpoint - 55 kilopascals to obtain the new negative pressure setpoint. For example, when the operating mode is "normal temperature - low vibration"... square millimeter, In kilopascals per second (a negative value indicates a decrease in pressure), the feedforward compensation increment is:

[0070] By adding 21 Pascals to the 55 kPa setpoint, the new negative pressure setpoint is -55.021 kPa. The control command generation unit generates a vacuum pump speed adjustment command based on the new negative pressure setpoint, controlling the brushless DC motor of the vacuum pump to operate at a higher speed, increasing the pumping rate and thus compensating for the pressure drop caused by leakage.

[0071] When the leakage status assessment unit outputs an alarm status flag, it indicates that the leakage has become so severe that it cannot be effectively compensated for by speed adjustment. At this time, the control command generation unit simultaneously generates two commands: an emergency stop command, sent to the servo drive module to immediately stop all movement of the clamping arm body and maintain its current position; and a vacuum pump maximum power operation command, sent to the vacuum pump driver to make the vacuum pump operate at maximum speed to provide maximum pumping capacity, maintain residual suction force as much as possible, and prevent the workpiece from falling off due to inertia during the emergency stop. After generation, the two commands are simultaneously distributed to the corresponding execution interfaces through shared memory, with an execution delay of no more than 10 milliseconds.

[0072] Digital twin model unit: The digital twin model unit is an independent process running within the edge computing controller, responsible for constructing a digital twin that corresponds one-to-one with the physical negative pressure control clamping arm unit. The digital twin employs a simulation model based on physical mechanisms, comprising three components: an adsorption chamber gas dynamics sub-model, a leakage evolution model, and a vacuum pump performance sub-model. The adsorption chamber gas dynamics sub-model is based on the aforementioned pressure dynamic differential equation. A model is constructed to solve for the virtual negative pressure value within a virtual adsorption cavity in real time in digital space. Leakage evolution is modeled based on the estimated equivalent area of ​​the leakage orifice from the extended Kalman filter output. The virtual leakage state parameters are updated. The vacuum pump performance sub-model is constructed based on the speed-flow characteristic curve of the vacuum pump, and the pumping flow rate of the virtual pump is calculated according to the vacuum pump speed adjustment command output by the control command generation unit.

[0073] In each computation cycle, the digital twin model unit retrieves the equivalent negative pressure estimate of the extended Kalman filter output from shared memory. Estimated equivalent area of ​​leakage orifice negative pressure change rate Current operating mode identifier Average temperature and the root mean square value of vibration acceleration amplitude The digital twin model unit uses this data as driving input to update the state parameters of each sub-model of the digital twin. Specifically, the update mechanism is as follows: the virtual negative pressure value of the digital twin is directly derived from the equivalent negative pressure estimate. The digital twin is updated by assignment, rather than being solved autonomously by the model, to ensure that the states of the digital twin and the physical entity are forcibly synchronized; the virtual leakage state parameters of the digital twin are estimated by the equivalent area of ​​the leakage orifice. Update; The temperature of the outer wall of the virtual adsorption chamber is now calculated from the average temperature. Update: The vibration level of the virtual clamping arm base is determined by the root mean square value of the vibration acceleration amplitude. Update. Through this data-driven forced synchronization mechanism, the digital twin maintains a precise state correspondence with the physical negative pressure control clamp unit in each synchronization cycle.

[0074] III. Remote Monitoring Layer Configuration The cloud platform server is deployed in the production company's data center, employing a cluster architecture of three servers to run containerized microservices. The cloud platform server is connected to the edge computing controller via the company's local area network, with a network bandwidth of gigabit Ethernet.

[0075] The two-way authentication mechanism for the secure communication link is implemented collaboratively by a first security chip deployed in the edge computing controller and a second security chip deployed in the cloud platform server. Both the first and second security chips are hardware security modules compliant with the FIPS 140-2 Level 3 certification standard, possessing key generation, secure storage, and encryption / decryption capabilities. The edge computing controller's private key is pre-stored in the first security chip via a secure offline injection method. and cloud platform server public key The cloud platform server's private key is pre-stored in the second security chip. and edge computing controller public key The key pair is generated using the secp256r1 curve of the elliptic curve cryptosystem, and the key length is 256 bits.

[0076] Before sending status data to the cloud platform server, the edge computing controller performs the following signature process: The first security chip uses the SHA-256 hash algorithm to calculate the message digest of the status data packet to be sent, obtaining a 256-bit hash value; then, it uses the edge computing controller's private key... The hash value is signed using the elliptic curve digital signature algorithm to generate a 512-bit digital signature. This digital signature is appended to the state data packet to form a complete signed data packet, which is then sent to the predetermined port of the cloud platform server via a Transmission Control Protocol (TCP) socket. Upon receiving the signed data packet, the cloud platform server extracts the state data packet portion and calculates the message digest using the SHA-256 hash algorithm. Then, it uses the edge computing controller public key stored in the second security chip... The received digital signature is verified using the Elliptic Curve Digital Signature Verification (ECD) algorithm. The hash value obtained from the verification is compared with the hash value calculated locally. If the two are consistent, the verification is successful, confirming that the data packet in this state has a legitimate source and has not been tampered with during transmission.

[0077] Before issuing control commands to the edge computing controller, the cloud platform server performs a signature process symmetrical to the above: the second security chip calculates a digest of the control command using the SHA-256 hash algorithm and uses the cloud platform server's private key. The signature is attached to the instruction and sent; the edge computing controller receives the instruction and uses the cloud platform server's public key. The signature is verified before the control command is transmitted to the control command generation unit for execution. Through this two-way authentication mechanism, any unsigned data packet or control command that fails signature verification is directly discarded and triggers a security audit log recording.

[0078] After receiving the status data uploaded by the edge computing controller, the cloud platform server parses out the equivalent negative pressure estimate. Estimated equivalent area of ​​leakage orifice negative pressure change rate Operating mode identifier The system stores the warning status flags, alarm status flags, average temperature, and root mean square value of vibration acceleration amplitude in the time-series database InfluxDB. At the same time, it calls the update interface of the cloud copy of the digital twin to synchronously update the operating parameters of the cloud digital twin.

[0079] The remote monitoring terminal is an industrial-grade touchscreen workstation deployed in the production scheduling center, running a 3D visualization monitoring application developed based on WebGL technology. The remote monitoring terminal maintains a persistent connection with the cloud platform server via the WebSocket protocol, receiving status data pushed by the cloud platform server in real time. The display interface of the remote monitoring terminal includes the following view components: the main view is a 3D model rendering interface of the digital twin, displaying the distribution cloud map of the virtual negative pressure field inside the adsorption chamber in a semi-transparent manner, with areas of higher negative pressure values ​​rendered in a darker blue; the sidebar displays a real-time negative pressure value curve, with time on the horizontal axis and negative pressure value on the vertical axis. The curve is drawn from data from the InfluxDB database over the past 5 minutes, with a refresh rate of 10 times per second; the bottom status bar displays a trend graph of the equivalent area change of the leakage orifice, text indicators of the current operating mode, and warning alarm status indicators. When the edge computing controller generates a warning status flag, the remote monitoring terminal receives the flag, and a warning prompt box with a yellow border pops up on the display interface, displaying the current value of the estimated equivalent area of ​​the leakage orifice and the timestamp of triggering the warning, while simultaneously emitting a short, one-second interval prompt sound through the terminal's built-in speaker. The remote monitoring terminal is equipped with a manual intervention interface. After seeing the warning prompt, the operator can input a confirmation command or manually enter the modified negative pressure setting value via the intervention button on the touch screen. The confirmation command or modified negative pressure setting value is sent to the cloud platform server via a WebSocket connection. The cloud platform server triggers the second security chip to perform SHA-256 hashing and elliptic curve digital signature algorithm signing on the command, and then sends it to the edge computing controller. After the edge computing controller verifies the signature, it parses the command content and executes it.

[0080] Comparative Example To more intuitively illustrate the beneficial technical effects of this embodiment, a comparative example is provided here. The comparative example employs a traditional fixed threshold leakage detection method. Its technical solution involves using only one negative pressure sensor installed within the adsorption chamber to monitor the negative pressure value in real time. When the negative pressure value falls below a preset fixed safety threshold of -45 kPa (i.e., the absolute value of the negative pressure is less than 45 kPa), an alarm is triggered and an emergency stop is executed. The comparative example lacks multi-sensor data fusion, operating condition mode recognition, feedforward compensation control, and two-way authentication secure communication functions.

[0081] A comparative experiment was conducted between this embodiment and the comparative example. The experimental conditions were as follows: In the same glass panel handling process, the system of this embodiment and the system of the comparative example were run respectively. A simulated leak was artificially introduced between the adsorption sealing ring and the glass panel, with the equivalent leakage area increasing linearly over time (achieved by inserting metal foils of different thicknesses under the sealing ring). The time interval from the introduction of the leak to the system making an effective response was recorded, as shown in Table 3.

[0082] Table 3 Comparison Results

[0083] As shown in Table 3, this embodiment, by fusing data from dual negative pressure sensors using extended Kalman filtering, estimates the implicit state variable of the equivalent area of ​​the leakage orifice in real time. This allows the system to identify the leakage state and trigger feedforward compensation control when the negative pressure value drops only from -55 kPa to approximately -53 kPa (a decrease of about 3.6%), reducing the response time by 83.3% compared to the comparative example. The comparative example, relying on a fixed threshold trigger, only responds when the negative pressure value drops below -45 kPa (a decrease of about 18.2%), at which point the leakage is already quite severe, necessitating an emergency stop. Furthermore, this embodiment configures differentiated warning thresholds for four operating modes, automatically adjusting the sensitivity of leakage detection based on actual conditions, avoiding the alarm delay problem that might occur with the fixed threshold in the comparative example under high temperature and high vibration conditions. The two-way authentication mechanism provides security for remote monitoring, a feature completely absent in the comparative example.

[0084] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0085] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A smart negative pressure control clamp arm and remote monitoring system based on Internet of Things (IoT) sensors, characterized in that, include: At least one negative pressure control clamping arm unit, the negative pressure control clamping arm unit including a clamping arm body, an adsorption chamber, a vacuum pump, a servo drive module, a first negative pressure sensor, a second negative pressure sensor, a temperature sensor, and a vibration sensor; the first negative pressure sensor is disposed on the inner wall of the adsorption chamber near the adsorption port, and is used to collect a first negative pressure value time series; the second negative pressure sensor is disposed in the vacuum pump outlet pipeline, and is used to collect a second negative pressure value time series; the temperature sensor is disposed on the outer wall of the adsorption chamber, and is used to collect a temperature value time series; the vibration sensor is disposed at the base of the clamping arm body, and is used to collect a vibration acceleration amplitude time series. An edge computing controller is communicatively connected to the first negative pressure sensor, the second negative pressure sensor, the temperature sensor, and the vibration sensor. The edge computing controller is configured as follows: The received first negative pressure value time series and second negative pressure value time series are subjected to outlier removal and low-pass filtering preprocessing to obtain the preprocessed first negative pressure value time series and the preprocessed second negative pressure value time series; The preprocessed first negative pressure value time series and the preprocessed second negative pressure value time series are input into an extended Kalman filter. The state variables of the extended Kalman filter include the equivalent negative pressure estimate, the negative pressure change rate, and the equivalent area of ​​the leakage orifice. The observations of the extended Kalman filter include the preprocessed first negative pressure value and the preprocessed second negative pressure value. The extended Kalman filter outputs the equivalent area estimate of the leakage orifice and the negative pressure change rate. The temperature value time series and the vibration acceleration amplitude time series are input into a pre-trained working condition mode classifier. The working condition mode classifier outputs the current working condition mode identifier based on the mean temperature and root mean square value of vibration acceleration amplitude within the current time window. The estimated equivalent area of ​​the leakage orifice is compared with a pre-stored leakage threshold table corresponding to the current operating condition mode identifier. When the estimated equivalent area of ​​the leakage orifice is greater than the first-level warning threshold and less than or equal to the second-level alarm threshold, a warning status flag is generated. When the estimated equivalent area of ​​the leakage orifice is greater than the second-level alarm threshold, an alarm status flag is generated. The control command generation module is used to, when the warning status flag is generated, calculate the estimated equivalent area of ​​the leakage orifice and the rate of change of negative pressure according to the formula... Calculate the feedforward compensation increment, add the feedforward compensation increment to the current negative pressure setpoint, and generate a vacuum pump speed adjustment command. This is an estimate of the equivalent area of ​​the leakage orifice. The rate of change of the negative pressure. This is the leakage area compensation coefficient. The rate compensation coefficient is used; the control command generation module is also used to generate an emergency stop command and a vacuum pump maximum power operation command when the alarm status flag is generated, the emergency stop command is used to control the servo drive module to stop moving; A cloud platform server is provided, and a secure communication link with the edge computing controller is established between the cloud platform server and the edge computing controller. The cloud platform server is used to receive status data uploaded by the edge computing controller. The remote monitoring terminal is used to receive status data forwarded by the cloud platform server and display the running interface, as well as to receive control commands input by the operator and send the control commands to the edge computing controller via the cloud platform server.

2. The intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors according to claim 1, characterized in that, The sampling frequency of the first negative pressure sensor is set to 50Hz to 200Hz, the sampling frequency of the second negative pressure sensor is set to 10Hz to 50Hz, the sampling frequency of the temperature sensor is set to 1Hz, and the sampling frequency of the vibration sensor is set to 100Hz to 500Hz. The negative pressure control clamping arm unit also includes a data acquisition module, which is electrically connected to the first negative pressure sensor, the second negative pressure sensor, the temperature sensor, and the vibration sensor. The data acquisition module is used to read the data collected by each sensor at a fixed period, attach a timestamp synchronized based on the NTP protocol to each sampling data point, encapsulate the timestamp-aligned data within the same time window into an Ethernet frame, and send it to the edge computing controller via the EtherNet / IP protocol. The edge computing controller receives the Ethernet frame, unpacks it, and recovers the time sequence of each sensor based on the timestamp.

3. The intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors according to claim 1, characterized in that, The specific process of the edge computing controller performing outlier removal and low-pass filtering preprocessing on the received first and second negative pressure value time series includes: using a sliding window method, with the window width set to the number of sampling points obtained by multiplying the sampling frequency by 0.5 seconds, and the sliding step size set to half the window width; for each data point within the window, calculating the median and median absolute deviation of the data points within the window, identifying data points that deviate from the median by more than 3 times the median absolute deviation as outliers and removing them; for the removed data points, using linear interpolation of the previous and next valid values ​​of the removed data points to fill them; and smoothing the sequences after outlier removal and filling using an infinite impulse response low-pass filter with a cutoff frequency of 5Hz to obtain the preprocessed first and second negative pressure value time series.

4. The intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors according to claim 1, characterized in that, The state vector of the extended Kalman filter is defined as follows: ,in This is an estimate of the equivalent negative pressure. The rate of change of negative pressure. The equivalent area of ​​the leakage orifice; the state transition equation uses a nonlinear function. , where nonlinear functions Based on the dynamic differential equation of the pressure in the adsorption chamber Discretization yields, Here, is the natural leakage coefficient, and is the leakage flow rate coefficient. Let be the process noise vector, which follows a zero-mean Gaussian distribution, and its covariance matrix is... The observation vector is defined as follows: ,in This refers to the first negative pressure value after pretreatment. The second negative pressure value after preprocessing is given; the observation equation is... The observation matrix, This is the static gain ratio between the first negative voltage value and the second negative voltage value. The coefficient representing the impact of leakage on the second negative pressure value. The observed noise vector follows a zero-mean Gaussian distribution, and its covariance matrix is... ; The extended Kalman filter recursively calculates the estimated equivalent area of ​​the leakage orifice through time update and measurement update steps. and the rate of change of negative pressure .

5. The intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors according to claim 1, characterized in that, The operating condition mode classifier employs a pre-trained support vector machine (SVM) model. The input feature vector of the SVM model consists of the average temperature and the root mean square (RMS) value of vibration acceleration amplitude within the current time window. The SVM model uses a radial basis function (RBF) kernel. The training dataset includes calibrated samples of average temperature and RMS values ​​of vibration acceleration amplitude under known operating conditions. During training, cross-validation is used to determine the penalty parameters and kernel function parameters. The operating condition mode categories output by the classifier include "normal temperature - low vibration," "normal temperature - high vibration," and "high temperature." - Low vibration mode and high temperature-high vibration mode; the pre-stored leakage threshold table is configured with corresponding first-level warning threshold and second-level alarm threshold for each working condition mode category, and the first-level warning threshold corresponding to the "high temperature-low vibration" mode, the first-level warning threshold corresponding to the "normal temperature-high vibration" mode and the first-level warning threshold corresponding to the "high temperature-high vibration" mode are all less than the first-level warning threshold corresponding to the "normal temperature-low vibration" mode. The threshold for judging high or low temperature is a preset temperature value, and the threshold for judging vibration intensity is a preset root mean square value of vibration acceleration amplitude.

6. The intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors according to claim 1, characterized in that, The control command generation module is further configured to obtain the corresponding leakage area compensation coefficient from a pre-established operating condition-coefficient mapping table based on the current operating condition mode identifier. and the rate compensation coefficient In the operating condition-coefficient mapping table, the leakage area compensation coefficient corresponding to the "normal temperature-low vibration" mode is set to a value of [value missing]. The leakage area compensation coefficient for the "normal temperature-high vibration" mode is [value], and the leakage area compensation coefficient for the "high temperature-low vibration" mode is [value]. The leakage area compensation coefficient corresponding to the "high temperature-high vibration" mode is taken as: In the operating condition-coefficient mapping table, the rate compensation coefficient corresponding to the "normal temperature-low vibration" mode is set to a value of [value missing]. The rate compensation coefficient for the "normal temperature - high vibration" mode is set to a value of The rate compensation coefficient corresponding to the "high temperature-low vibration" mode is set to a value of The rate compensation coefficient corresponding to the "high temperature-high vibration" mode is set to a value of .

7. The intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors according to claim 1, characterized in that, The edge computing controller also deploys a digital twin model unit, which is used to construct a digital twin of the negative pressure control clamping arm unit; the digital twin receives the equivalent negative pressure estimate output by the extended Kalman filter. The estimated equivalent area of ​​the leakage orifice. The rate of change of negative pressure The current operating mode identifier, the average temperature, and the root mean square value of the vibration acceleration amplitude are used to update the virtual negative pressure value and virtual leakage status parameters within the virtual adsorption cavity of the digital twin, synchronizing the operating status of the digital twin with the real-time physical status of the negative pressure control clamping arm unit. A cloud copy of the digital twin is stored in the cloud platform server, which updates the cloud copy synchronously after receiving status data uploaded by the edge computing controller. The remote monitoring terminal displays a three-dimensional visualization interface of the digital twin by accessing the cloud copy. This three-dimensional visualization interface includes a real-time negative pressure curve, a trend chart of the equivalent area change of the leakage orifice diameter, and early warning alarm status indicators.

8. The intelligent negative pressure control clamping arm and remote monitoring system based on IoT sensors according to claim 1, characterized in that, The two-way authentication mechanism of the secure communication link specifically includes: the edge computing controller has a built-in first security chip, which stores the edge computing controller's private key and the cloud platform server's public key; the cloud platform server has a built-in second security chip, which stores the cloud platform server's private key and the edge computing controller's public key; before sending status data to the cloud platform server, the edge computing controller uses its private key to generate a digital signature for the status data packet using the SHA-256 hash algorithm and the elliptic curve digital signature algorithm, and attaches the digital signature to the status data packet before sending it; after receiving the data, the cloud platform server uses the edge computing controller's public key to verify the digital signature, and if the verification is successful, it confirms that the source of the status data packet is legitimate; before issuing control commands to the edge computing controller, the cloud platform server uses its private key to generate a digital signature for the control command using the SHA-256 hash algorithm and the elliptic curve digital signature algorithm, and attaches the digital signature to the control command before sending it; after receiving the control command, the edge computing controller uses the cloud platform server's public key to verify the digital signature, and if the verification is successful, it executes the control command.

9. The intelligent negative pressure control clamp arm and remote monitoring system based on IoT sensors according to claim 1, characterized in that, When the warning status flag is generated, the edge computing controller packages the estimated equivalent area of ​​the leakage orifice, the rate of change of negative pressure, the calculated feedforward compensation increment, and the adjusted vacuum pump speed command into a warning event record, and uploads it to the cloud platform server through the secure communication link. After receiving the warning event record, the remote monitoring terminal pops up a warning prompt box on the display interface and enables an audible alarm. The remote monitoring terminal provides a manual intervention interface, through which the operator inputs a confirmation command or a modified negative pressure setting value. The confirmation command or the modified negative pressure setting value is signed by the cloud platform server using the cloud platform server's private key and then sent to the edge computing controller. The edge computing controller executes the command after the signature is verified.

Images