Multi-sensor information fusion mechanical equipment safety interlock control method

Abstract

The application discloses a mechanical equipment safety interlock control method based on multi-sensor information fusion. The method collects the motion position, speed, acceleration, force and torque data of the mechanical equipment, and performs time alignment, amplitude normalization and short-time abnormality filtering, estimates the kinetic energy, potential energy and friction energy consumption of each energy node, and constructs an energy transmission matrix between nodes; fusion sensor data and matrix information extracts four types of abnormal characteristics of short-term energy surge, local energy reflection, friction energy abnormal fluctuation and energy accumulation accompanied by micro-vibration, and predicts dangerous energy states including single and complex dangerous energy states accordingly; based on the prediction of dangerous energy state and the current running state of the equipment, the safety interlock action is adaptively triggered, the driver torque, current amplitude limiting and active damping are dynamically adjusted, and the control of backflow energy and accumulated energy along the virtual dissipation path is realized. The method is suitable for high dynamic mechanical equipment and can realize fine safety interlock control of multiple nodes and multiple abnormal characteristics.

Description

Technical Field

[0001] This application relates to the field of mechanical equipment safety control technology, specifically a mechanical equipment safety interlock control method based on multi-sensor information fusion. Background Technology

[0002] Safety interlock control during the operation of mechanical equipment (such as crushers, presses, tunnel boring machines, and large rotating machinery) typically relies on real-time monitoring of key parameters such as position, speed, and force. The current mainstream method is threshold-triggered safety interlocking: when any parameter exceeds a preset safety threshold, the controller triggers an emergency stop or protective action. This method is widely used in engineering practice, for example, for overload protection, overspeed protection, and bearing temperature protection, but it still has the following drawbacks:

[0003] Since threshold values ​​are typically set at safety boundaries where parameters have significantly deviated from normal operating ranges, dangerous conditions such as overload, runaway, and jamming often occur when monitored values ​​exceed these limits. In such cases, interlocking actions can only limit the further escalation of the accident, but cannot achieve preventative intervention. For high-inertia, high-impact machinery (such as large crushers and presses), the time window from parameter exceeding the limit to the complete occurrence of the accident is extremely short, limiting the actual protective effect of threshold-based interlocking.

[0004] Existing methods rely on distributed individual sensors to measure local locations, such as speed sensors mounted on shaft ends. This monitoring approach cannot detect the spatial distribution and dynamic flow of energy in the drivetrain. Typically, microcracks in the drive shaft can cause local reflections of elastic potential energy and standing waves, and bearing wear can lead to abnormal accumulation of frictional dissipation energy in local areas. However, these early physical characteristics are completely ignored before reaching the threshold of a single sensor, and are only triggered when the fault develops to a severe stage.

[0005] It is evident that existing technologies lack a safety interlocking method that can overcome the limitations of threshold-based passive response and achieve proactive early identification of anomalies during the operation of mechanical equipment, which urgently needs further improvement. Summary of the Invention

[0006] The purpose of this application is to provide a method for safety interlock control of mechanical equipment based on multi-sensor information fusion, as well as a computer device, storage medium, and computer program product, to solve the problems mentioned in the background art.

[0007] According to the first aspect of this application, a method for safety interlock control of mechanical equipment based on multi-sensor information fusion is provided, comprising the following steps:

[0008] Collect sensor data of mechanical equipment, including motion position, speed, acceleration, force, and torque, and perform time alignment, amplitude normalization, and short-term anomaly filtering on each sensor data.

[0009] Based on the processed sensor data, the kinetic energy, potential energy, and frictional energy consumption of each energy node of the mechanical equipment are estimated, and an inter-node energy transfer matrix that can reflect the energy flow direction, energy accumulation, and reverse transmission state is constructed accordingly.

[0010] Data from various sensors and the energy transfer matrix between nodes are integrated to extract abnormal energy features, including short-term energy surge features, local energy reflection features, abnormal fluctuation features of frictional energy, and energy accumulation features accompanied by micro-vibration. Based on the abnormal energy features, the prediction results of dangerous energy states are obtained.

[0011] Based on the predicted dangerous energy state and the current operating status of the mechanical equipment, the safety interlock action is adaptively triggered, and the sensor weights and interlock logic are dynamically adjusted to achieve proactive intervention and advanced protection for dangerous energy states.

[0012] According to a second aspect of this application, a computer device is provided, including a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, implements the method as described in any of the preceding claims.

[0013] According to a third aspect of this application, a computer program product is provided, including instructions that, when the computer program product is executed, cause a computer device to perform the method as described in any of the preceding claims.

[0014] According to a fourth aspect of this application, a storage medium is provided that stores a computer program product as described above.

[0015] Compared with existing technologies, this invention collects data from multiple sensors of mechanical equipment and performs time alignment, amplitude normalization, and short-term anomaly filtering. Combined with the calculation of kinetic energy, potential energy, and frictional energy consumption of each energy node, an energy transfer matrix between nodes is constructed. Then, it integrates data from various sensors to extract abnormal energy characteristics such as short-term energy surges, local energy reflections, abnormal fluctuations in frictional energy, and energy accumulation accompanied by micro-vibrations. Based on these characteristics, dangerous energy states are predicted. At the same time, safety interlock actions are adaptively triggered according to the prediction results and the current operating state of the mechanical equipment. Sensor weights and interlock logic are dynamically adjusted to achieve advanced identification and response to dangerous energy states. Thus, it can realize refined safety interlock control under multiple nodes and multiple abnormal characteristics in highly dynamic mechanical equipment, improving the accuracy of dangerous energy state judgment and the reliability of interlock actions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0018] Figure 1 A flowchart illustrating a multi-sensor information fusion-based safety interlock control method for mechanical equipment, provided in an embodiment of this application;

[0019] Figure 2 A schematic diagram illustrating the adaptive triggering of safety interlocking actions provided in an embodiment of this application;

[0020] Figure 3 A schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0021] The technical solution of the present invention will be described in detail below with reference to specific embodiments. It should be noted that the following embodiments are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0022] like Figure 1 As shown in the figure, this application provides a method for safety interlock control of mechanical equipment based on multi-sensor information fusion. The method includes the following steps:

[0023] S1, collect sensor data of mechanical equipment, including motion position, speed, acceleration, force, and torque, and perform time alignment, amplitude normalization, and short-term anomaly filtering on the sensor data.

[0024] Specifically, the following sensors are placed at the critical energy nodes of mechanical equipment. Taking rotating machinery as an example: a three-phase current / voltage sensor is placed at the input end (to measure the electric power of the motor). The system includes a motor shaft torque sensor and a speed sensor; multiple triaxial vibration acceleration sensors and acoustic emission sensors (for detecting energy release from microcracks) are arranged along the transmission chain path; load torque sensors, actuator position / velocity sensors (such as press slide displacement, crusher speed), and strain gauges (mounted on the frame or key structural components to measure elastic deformation energy) are arranged at the actuator end. All sensors are connected to a real-time controller (PLC, FPGA, or industrial computer) via a synchronous data acquisition card (sampling rate at least 10 times the highest mechanical frequency).

[0025] The acquired raw data is time-aligned (interpolated to a unified time axis based on the same global clock) and amplitude-normalized (the dimensions of each sensor are mapped to the [0,1] or [-1,1] interval). Median filtering or Kalman filtering is used for short-term anomaly filtering to remove transient electromagnetic interference or sensor spike noise.

[0026] S2, based on the processed sensor data, estimate the kinetic energy, potential energy and frictional energy consumption of each energy node of the mechanical equipment, and construct an energy transfer matrix between nodes that can reflect the energy flow direction, energy accumulation and reverse transmission state.

[0027] During operation, mechanical equipment experiences energy flow, conversion, and storage between components in the form of kinetic energy, elastic potential energy, and frictional dissipation energy. Traditional methods only monitor whether the amplitude of a single physical quantity exceeds a threshold, failing to detect abnormal changes in the energy flow path. This invention, from the perspective of energy conservation, discretizes the mechanical equipment into several energy nodes. By using multi-sensor data to estimate the energy state of each node and the power exchange between nodes in real time, a digital model of the energy flow of the entire equipment is constructed.

[0028] As an example, the kinetic energy, potential energy, and frictional energy consumption of each energy node in the mechanical equipment are estimated based on the processed sensor data. Based on this, an inter-node energy transfer matrix reflecting energy flow direction, energy accumulation, and reverse transmission status is constructed, including:

[0029] S21, the mechanical equipment is divided into energy nodes. Combining the mass, inertia and structural stiffness of the nodes, the kinetic energy, potential energy and frictional energy consumption of the nodes are calculated using the collected sensor data. At the same time, the connection relationship between the nodes and the energy transfer direction are considered to form an energy distribution.

[0030] Mechanical equipment can be divided into multiple energy nodes along its energy transmission path. Taking a rotating machinery transmission chain as an example, it can be divided into: motor rotor node (node ​​1), coupling node (node ​​2), transmission shaft input end node (node ​​3), transmission shaft output end node (node ​​4), gearbox node (node ​​5), and actuator node (node ​​6). For reciprocating machinery (such as a stamping press), it can be divided into: motor node, crank-connecting rod node, and slider node, etc.

[0031] For each energy node, its inherent physical properties are obtained: mass (for linear motion nodes), moment of inertia (for rotating nodes), and structural stiffness (for elastic deformation nodes). These parameters can be obtained from the design drawings or field calibrations of the mechanical equipment. Then, using preprocessed multi-sensor data, the following energies for each node are calculated:

[0032] Kinetic energy: For rotating nodes (such as motor rotors, drive shafts, and gears), the angular velocity is measured using a speed sensor. Combined with the rotational inertia of the nodes Calculate rotational kinetic energy For linear motion nodes (such as the slide block of a stamping press or the moving jaw of a crusher), the linear velocity measured by a speed sensor is used. Combined with the quality of the nodes Calculate translational kinetic energy If a node contains both rotational and linear motion (such as a crank-connecting rod mechanism), then calculate them separately and then sum them.

[0033] Potential energy: Primarily considers the deformation energy of the elastic element. For the drive shaft section, torque sensors are installed at both ends, or surface strain is measured using strain gauges to obtain the torsional angle of the cross-sections at both ends. and Given the torsional stiffness of the shaft. (Calculated from the material shear modulus and geometric dimensions), the torsional elastic potential energy stored in the shaft segment is: For gear meshing nodes, the elastic potential energy stored by tooth surface contact deformation can be estimated using a tooth surface deformation model: ,in For meshing stiffness, This represents the normal deformation of the tooth surface (which can be calculated from torque and meshing stiffness). For structural components equipped with strain gauges (such as machine frames and bearing housings), the linear strain measured directly using the strain gauges is... Calculate the elastic strain energy density ( The elastic modulus is used as a unit, and then the total potential energy is obtained by integrating over the structural volume. For nodes such as lifting mechanisms where gravitational potential energy changes, the gravitational potential energy also needs to be calculated based on data from height or displacement sensors. .

[0034] Frictional energy dissipation: Frictional energy dissipation is the irreversible process of converting energy from ordered mechanical energy into heat energy, mainly occurring at contact nodes such as bearings, gear meshing, sliding guides, and seals. Frictional energy dissipation cannot be directly measured. This invention uses a neural network model based on vibration and acoustic emission signals to dynamically estimate frictional torque. Then, the frictional energy consumption is obtained by integration. .

[0035] Specifically, the friction nodes of the same type (such as rolling bearings) are calibrated on the experimental platform beforehand; at the same time, the root mean square value of the vibration acceleration signal is collected. Energy rate of acoustic emission signals and frictional torque measured directly by a torque sensor Construct a training dataset. Then train a multilayer perceptron (input: , Rotation speed Output: In actual operation, the vibration and acoustic emission characteristics collected in real time are input into the neural network to obtain the estimated friction torque value of the current node. Then, the friction energy consumption increment in each control cycle is calculated by numerical integration, and the total friction energy consumption is obtained by summing them up.

[0036] After calculating the kinetic energy, potential energy, and frictional energy consumption of each node, the connection relationships between nodes and the direction of energy transfer must also be considered. Connection relationships include rigid connections (such as couplings), flexible connections (such as belts and chains), gear meshing, and sliding contact. The direction of energy transfer is determined by the sign of the power flow. Specifically: when the output power (torque × speed) of the upstream node is positive and the input power of the downstream node is also positive, energy is transferred from upstream to downstream; if the torque direction of the downstream node is opposite to the speed direction (such as energy feedback during deceleration of a large inertial load), the energy flows in the opposite direction. This directional information is recorded as the initial sign of the energy distribution at each node (positive for inflow, negative for outflow). Simultaneously, based on the physical connections between nodes, it is possible to determine which nodes have energy exchange paths, thus forming the energy topology of the entire system.

[0037] Through the above calculations, the total energy (the sum of kinetic and potential energy) of each node, the cumulative value of frictional dissipation energy, and the power flow direction between each node and its adjacent nodes are finally obtained. This information constitutes the energy distribution map of the entire mechanical equipment.

[0038] S22, the energy distribution of each node is mapped and weighted according to the connection between nodes and the direction of energy flow, so as to construct the energy transfer matrix between nodes that can reflect the energy flow direction, energy accumulation and reverse transmission status between nodes.

[0039] After obtaining the energy distribution of each node and the connection relationships between nodes, this scattered information is organized into an inter-node energy transfer matrix. This facilitates subsequent extraction of abnormal features and real-time monitoring. yes The square array, in which This represents the total number of energy nodes. Matrix elements. (No. Line number (Column) represents the time from node Flow to Node The instantaneous power (i.e., energy flow rate). The construction process of this matrix is ​​roughly as follows:

[0040] The energy distribution of each node and the original power flow between nodes, as determined above, are filled into the corresponding positions in the matrix according to the connection relationships. The specific rules are as follows:

[0041] (1) For each directed connection (from node) To the node ), and the results obtained from the above calculations Output to power value Assigned .

[0042] (2) If node With nodes If there is no direct connection between them, then .

[0043] (3) Diagonal elements Represents a node The energy accumulation rate of a node is equal to the total power flowing into the node minus the total power flowing out of the node, minus the power dissipated by the node (friction, vibration radiation, etc.). According to the law of conservation of energy, It is also equal to the time derivative of the energy stored at that node (kinetic energy + potential energy). This can be verified through numerical differentiation.

[0044] (4) For cases of reverse energy transmission (such as load kinetic energy feedback motor), the corresponding connected power value is negative, which is represented by negative elements in the matrix.

[0045] Considering the path loss during energy transmission (such as coupling efficiency) Gear meshing efficiency Belt drive efficiency (etc.), the mapped power values ​​need to be weighted and corrected. Specifically, if from the node To the node There is inefficiency in the transmission path. Then it is actually transmitted to the node The power is ,in This represents the theoretical power transmitted without considering losses (determined by the output power of the upstream node). The efficiency coefficient can be obtained beforehand through experimental calibration or by consulting the equipment manual. Furthermore, for multiple parallel paths (such as multiple belt drives), a weighted distribution must be performed according to the stiffness ratio or load distribution coefficient of each path, ensuring that the sum of the power of each path equals the total transmitted power.

[0046] After completing the above mapping and weighting, a complete inter-node energy transfer matrix can be obtained within each control cycle (e.g., 1ms). To suppress sensor noise and improve estimation accuracy, this invention employs a Kalman filter to perform optimal estimation of matrix elements. The state variables of the Kalman filter include the stored energy values ​​(kinetic energy, potential energy) of each node and the power flow of each path, while the observed variables are the measured values ​​of all sensors (torque, speed, vibration, etc.), and the system model is an energy conservation differential equation. Through filtering iteration, a smooth and interference-resistant matrix sequence can be obtained.

[0047] The columns of the final constructed inter-node energy transfer matrix Indicates from node Net power outflow (a negative value indicates net inflow); row sum of the matrix Indicates the inflow node Net power; diagonal elements The sign of the element indicates whether the node is accumulating (positive) or releasing (negative); the sign of off-diagonal elements indicates the direction of energy flow (positive for flow from column nodes to row nodes, negative for flow in the opposite direction); the antisymmetric part of the matrix. It reflects the circulating flow or local reflection of energy between nodes, and its norm can quantify the degree of anomaly in the energy path.

[0048] S3, integrate the data from each sensor and the energy transfer matrix between nodes to extract abnormal energy features, including short-term energy surge features, local energy reflection features, abnormal fluctuation features of frictional energy and energy accumulation features accompanied by micro-vibration, and predict dangerous energy state prediction results based on the abnormal energy features.

[0049] During the operation of mechanical equipment, anomalies in energy flow paths often manifest in various forms: sudden overloads cause a short-term surge in energy; microcracks or loosening cause stress wave reflection, leading to reverse energy propagation; bearing wear results in an abnormal increase in frictional energy consumption; and micro-vibrations within structural components couple with elastic strain energy, forming localized energy accumulation. A single feature is insufficient to comprehensively characterize early warning signs. Therefore, this invention employs the following method for extracting abnormal energy features and achieves hazardous energy state determination through multi-feature fusion via joint registration. The extraction process for each abnormal energy feature and the prediction method based on these features are described in detail below.

[0050] As an example, fusing the sensor data and the energy transfer matrix between nodes to extract anomalous energy features includes:

[0051] The energy change rate of each node within the time window is calculated, and the change in the rated power change rate where the energy growth rate exceeds a preset multiple within multiple consecutive control cycles is recorded as a short-term energy surge characteristic.

[0052] In practice, the total energy of each energy node is defined. This is the sum of kinetic and potential energy (frictional dissipation energy is not considered reversible energy storage and is not included). Taking the derivative with respect to time yields the rate of change of energy. In practical digital control systems, the backward difference approximation is used: ,in For control period (e.g., 1ms). For discrete time steps.

[0053] Set a sliding time window (e.g., containing 20 consecutive control cycles, i.e., 20 ms), and calculate the average value of the energy change rate within the window. At the same time, the rate of change of rated power of the equipment under normal and stable operating conditions was statistically analyzed in advance. This value can be determined by the 95th percentile of the rate of change of power in long-term operating data. Exceeding three consecutive control cycles If the energy level is 1.5 times the overload capacity of the equipment (this multiple can be adjusted according to the overload capacity of the equipment, and the value range is 1.2-2.0), it is determined to be a short-term energy surge event.

[0054] At the same time, record the characteristic parameters of this event: the number of the sudden increase node. The moment the surge begins Maximum energy change rate and duration These parameters are stored in the short-term energy surge feature vector. This is for use in subsequent prediction of dangerous energy states.

[0055] It should be understood that this characteristic typically corresponds to sudden load impacts (such as a crusher encountering an uncrushable object), transmission jamming, or motor overload starting.

[0056] The proportion of reverse energy components is calculated by using the symmetry decomposition of the energy transfer matrix between nodes, and the values ​​of reverse proportion exceeding the set threshold of forward energy and lasting for at least one mechanical fluctuation cycle are recorded as local energy reflection characteristics.

[0057] In practical implementation, the energy transfer matrix between nodes yes A real matrix can be uniquely decomposed into a symmetric part and an antisymmetric part: ,in , The symmetrical part represents the circulating flow of energy between nodes (without net transfer), while the antisymmetric part represents the directional transfer of net energy flow.

[0058] The proportion of the energy reverse component is defined as the ratio of the Frobenius norm of the antisymmetric part to the total Frobenius norm of the matrix: ,in .when When the value exceeds a preset threshold (e.g., 0.3, which can be set according to the statistical distribution of the circulating energy ratio during normal operation of the equipment, such as taking the mean plus 3 times the standard deviation), it indicates that there is significant reverse energy transmission or local reflection in the system.

[0059] To prevent false alarms caused by transient noise, this state must last for at least one mechanical fluctuation cycle. The definition of the mechanical fluctuation cycle depends on the type of equipment: for rotating machinery, it is one complete rotation cycle (…). , (RPM is the rotational speed); for reciprocating machinery (such as a stamping press), it is one working cycle (the time it takes for the crankshaft to rotate 360°). During this period, continuous monitoring is performed. If the value remains above the threshold, a local energy reflection feature is recorded once at the end of the duration.

[0060] The recorded feature information includes: the location where the reflection occurs (through analysis). (The node pair corresponding to the element with the largest amplitude determines the reflection path) and reflection intensity. And the duration of the reflection. It should be understood that this characteristic typically corresponds to anomalies such as stress wave reflection caused by microcracks in the drive shaft, sudden release of stored potential energy in a flexible coupling, gear meshing impact, or tooth back contact.

[0061] For nodes with frictional energy consumption, the deviation amplitude is obtained by comparing the current frictional energy consumption value of the node with the historical energy dissipation baseline. Frequency band energy analysis is performed on the vibration acceleration signal corresponding to the same node to obtain the high-frequency energy ratio. The abnormal fluctuation characteristics of frictional energy are obtained by combining the deviation amplitude and the high-frequency energy ratio.

[0062] In practice, for pre-marked friction energy consumption nodes (usually including bearing housings, gearboxes, sliding guides, brakes, etc.), a historical energy dissipation baseline is established for each friction energy consumption node. Specifically, during the normal and stable operation phase after the equipment has completed its break-in period, friction energy consumption data is continuously collected for no less than 1 hour. Calculate the friction power in each control cycle. Calculate the average value of this frictional power. and standard deviation , which serves as the baseline for that node.

[0063] Calculate the current friction power during real-time operation. (This can be obtained by averaging through a sliding window, with a window length of, for example, 100ms), then the deviation amplitude is defined as... .when This indicates that the frictional energy consumption is higher than normal.

[0064] Simultaneously, frequency band energy analysis is performed on the vibration acceleration sensor signals installed at the same node. The sampling frequency should satisfy the Nyquist theorem, for example, by taking 2.56 times the highest analysis frequency. A Fast Fourier Transform (FFT) is performed on the vibration signal segment (length, for example, synchronized with the mechanical wave period) within each control cycle to obtain the power spectral density. Based on power spectral density, the energy distribution within different frequency bands can be quantitatively calculated.

[0065] Define gear meshing frequency (If a gearbox is present) or the characteristic frequency of the device (e.g., the frequency at which the bearing passes through). High-frequency energy percentage. The calculation formula is:

[0066]

[0067] in, Desirable or , This is the low-frequency cutoff frequency (e.g., 10Hz). To analyze the upper frequency limit, it should be understood that the numerator of the above formula represents the total energy in the high-frequency band, and the denominator represents the total energy across the entire frequency band. This ratio... It reflects the proportion of high-frequency impact components caused by frictional anomalies in the vibration signal.

[0068] By combining the deviation magnitude and the proportion of high-frequency energy, an abnormal fluctuation index of friction energy is obtained: The weighting coefficients satisfy ,For example (Can be optimized based on on-site debugging). When When the value exceeds a preset threshold (e.g., 2.0, indicating a deviation from the normal level by 2 standard deviations and a high proportion of high-frequency energy), a characteristic of abnormal frictional energy fluctuation is recorded. Recorded information includes: node number. Current friction power value, deviation range, high-frequency energy ratio, and comprehensive index value It should be understood that this characteristic corresponds to progressive failures such as bearing cage damage, gear tooth surface pitting, poor lubrication, and sliding guide rail scoring.

[0069] By analyzing the micro-vibration signals and accumulated energy at nodes or paths, the energy accumulation caused by micro-vibration is recorded as the energy accumulation characteristic accompanied by micro-vibration.

[0070] Micro-vibrations typically refer to elastic waves with frequencies above 1 kHz and amplitudes less than 1 μm, primarily originating from microcrack propagation, dislocation motion, and frictional flutter within materials. Although these vibrations have small amplitudes, the energy they carry, if unable to be released locally, will gradually accumulate and eventually lead to structural failure. This invention integrates acoustic emission (AE) sensor data with strain gauge data to capture this phenomenon.

[0071] Specifically, acoustic emission sensors (resonant frequency 150kHz~1MHz, sampling rate ≥3MHz) and strain gauges (sampling rate at least 10kHz) are simultaneously installed on the surfaces of key structural components (such as bearing housings, gearbox housings, and frames). The acoustic emission signal is preamplified (40dB gain) and bandpass filtered before being synchronously acquired by a data acquisition card. The following two intermediate quantities are defined:

[0072] rate of change of elastic strain energy Real-time strain measured by strain gauges Calculate elastic strain energy density Multiplied by structural volume To obtain the total elastic strain energy The rate of change is obtained by taking the derivative with respect to time. .

[0073] micro-vibration amplitude The root mean square (RMS) value of the acoustic emission signal is calculated using a sliding time window (e.g., a 10 μs window, corresponding to 10 sampling points of the AE signal). .

[0074] Define the energy accumulation index This index is the product of the rate of change of elastic strain energy per unit time and the amplitude of micro-vibration, reflecting the rate of irreversible energy injection accompanying micro-vibration. When the structure is in a stable elastic vibration state, Alternating between positive and negative, The mean is close to zero; when microcrack propagation or plastic deformation is present, Continuously positive (energy is continuously injected and converted into internal energy), and Elevation, leading to Significantly increased.

[0075] Simultaneously monitor the acoustic emission event rate. (The number of hits per unit of time, for example, a threshold is set to count an event as if it exceeds three times the background noise). When Exceeding a preset threshold (e.g., 5% of rated power, a value set based on structural strength and safety margin) and When the synchronous rise occurs (i.e., the event rate exceeds three times the historical baseline), it is determined to be a characteristic of energy accumulation accompanied by micro-vibration. The location of the accumulation node and the current... Value, event rate And duration. It should be understood that this characteristic typically corresponds to early failures such as structural microcrack propagation, composite material delamination, and fretting wear caused by loose bolts.

[0076] As an example, the prediction of dangerous energy states based on the aforementioned anomalous energy characteristics includes:

[0077] S31, jointly registers the characteristics of short-term energy surge, local energy reflection, abnormal fluctuation of frictional energy, and energy accumulation accompanied by micro-vibration in time and space:

[0078] If the intensity of a single feature exceeds its dynamic threshold and shows an upward trend, it is determined to be a single hazardous energy state; if the ratio of the time difference between the occurrence of any two or more features at the same energy node or adjacent nodes to the mechanical impact propagation delay is less than 1 and the feature intensities all show an upward trend, it is determined to be a composite hazardous energy state.

[0079] During joint registration, a feature intensity vector is first constructed for each energy node to quantify the severity of four types of anomalous energy features. Specifically, the feature intensity of a short-term energy surge is defined as the ratio of the maximum rate of energy change to the rate of change of rated power, i.e. The characteristic intensity of local energy reflection is defined as the proportion of the energy reflection component. The characteristic intensity of abnormal fluctuations in frictional energy is the aforementioned comprehensive index. The characteristic intensity of energy accumulation accompanying micro-vibrations is defined as the energy accumulation index. Its threshold ratio The above feature intensities are all continuous values; the larger the value, the more significant the corresponding anomaly.

[0080] To ensure adaptive anomaly detection under different operating conditions, a dynamic threshold is maintained for each feature type. This dynamic threshold is updated online using the Exponentially Weighted Moving Average (EWMA) method. ,in , This represents the 95th percentile of recent historical feature intensity. It should be understood that dynamic thresholds can drift slowly with equipment wear or environmental changes, avoiding false alarms or missed alarms caused by fixed thresholds.

[0081] Based on the aforementioned feature intensity and dynamic threshold, the feature activation state can be further obtained. The activation state is a Boolean value, indicating whether the feature intensity exceeds the corresponding dynamic threshold within the current control cycle. For example, when the intensity of a short-term energy surge exceeds its dynamic threshold, the feature is marked as an activation state (1); otherwise, it is an inactive state (0). Similarly, the same comparison logic is applied to local energy reflection, abnormal fluctuations in frictional energy, and energy accumulation accompanied by micro-vibrations. Finally, each energy node maintains an activation state array containing four Boolean values ​​to quickly determine which abnormal energy features coexist at the current moment.

[0082] For a single hazardous energy state: if any one of the characteristics is activated and its intensity monotonically increases over three consecutive control cycles, it is determined to be the single hazardous energy state corresponding to that characteristic. The specific correspondences are as follows: short-term energy surge corresponds to overload-type hazardous energy state; local energy reflection corresponds to backflow-type hazardous energy state; abnormal fluctuations in frictional energy correspond to progressive wear-type hazardous energy state; and energy accumulation accompanied by micro-vibration corresponds to microcrack propagation-type hazardous energy state.

[0083] For composite hazard energy states: when two or more features of different classes are activated simultaneously (or sequentially activated on adjacent nodes), it is necessary to evaluate their temporal and spatial correlation. Assume the two features appear on nodes... and nodes The activation times are respectively and (If they are the same node) Calculate the shortest propagation path length between two nodes. (Determined by the topology of the mechanical structure), and the propagation speed of mechanical impact in the material. (The longitudinal wave velocity in steel is approximately 5000 m / s, and the transverse wave velocity is approximately 3000 m / s; taking the slower one as a conservative estimate), then the propagation delay... .

[0084] Define time difference like This indicates that the two characteristics occur almost simultaneously in time (the difference is less than the propagation time), satisfying causal consistency. Simultaneously, the intensity of both characteristics must show an upward trend (increasing continuously for three cycles). When the above conditions are met, it is determined to be a composite hazardous energy state, which can be further subdivided based on the combination of characteristics: for example, short-term energy surges and local energy reflections correspond to impact-rebound composite hazards; abnormal fluctuations in frictional energy and micro-vibration accumulation correspond to wear-crack coupled composite hazards.

[0085] S32, based on the dangerous energy state determination result, uses the extrapolation time difference between the current characteristic change rate and the preset dangerous boundary value to output the dangerous energy state prediction result.

[0086] As an example, the dangerous energy state prediction results include the dangerous type, the location of the dangerous node, and the remaining safe time.

[0087] The results of the hazardous energy state prediction include: hazardous type, hazardous node location, and remaining safe time. The calculation methods are explained below.

[0088] Hazard type: The hazardous energy state type determined above includes: overload type, backflow type, progressive wear type, microcrack propagation type (single hazardous energy state), and various composite types (such as rotational overload type, etc.).

[0089] Dangerous node location: For a single dangerous energy state, take the node number with the highest activation intensity of the corresponding feature; for a composite dangerous energy state, take the node with the highest intensity among all participating features, or take the earliest occurrence of the feature as the main dangerous node.

[0090] Remaining safe time: Let the feature intensity at the current moment be... The characteristic change rate (Obtained by linear fitting of the intensity values ​​from the most recent three control cycles), the danger boundary value is (Determined based on the mechanical design safety factor, for example, twice the rated value, or 50% of the material fatigue limit). For cases where the eigenvalue increases ( The remaining safety time is obtained by linear extrapolation. If the eigenvalues ​​exhibit an exponential growth trend (which can be determined by the fitted residuals), then exponential extrapolation is used: ,in The growth rate is exponential. For the composite hazardous energy state, the corresponding values ​​for each participating characteristic are calculated separately. The minimum value is taken as the final remaining safe time.

[0091] The output of the dangerous energy state prediction result can be in the form of a triplet. .when When the time is less than the preset warning time (e.g., 2 seconds), the warning state is activated; when When the time is less than the equipment's inertial time constant (e.g., 0.5 seconds), an emergency interlock is triggered.

[0092] S4. Based on the predicted dangerous energy state and the current operating status of the mechanical equipment, adaptively trigger the safety interlock action, and dynamically adjust the sensor weights and interlock logic to achieve proactive intervention and advanced protection for dangerous energy states.

[0093] Traditional safety interlocking methods typically cut off the power source or execute emergency braking after detecting excessive parameters. However, for complex hazardous energy states (where energy backflow and local accumulation coexist and couple to form a positive feedback loop), simple shutdown may be counterproductive. For example, suddenly cutting off motor power prevents the backflow energy from being absorbed by the drive end, thus exacerbating the impact on the transmission system; while the intervention of mechanical brakes may cause the accumulated elastic potential energy to be released explosively, resulting in secondary damage. Therefore, this invention dynamically reconstructs the energy dissipation path to safely mitigate hazards while maintaining continuous equipment operation.

[0094] As an example, such as Figure 2 As shown, based on the predicted hazardous energy state and the current operating state of the mechanical equipment, an adaptive safety interlock action is triggered, including:

[0095] S41, when the predicted dangerous energy state is a composite dangerous energy state, the energy coupling impedance between the return path and the accumulation node is inverted in real time, and the critical dissipation power required to decouple the positive feedback loop is calculated.

[0096] In practice, detailed information about the composite hazardous energy states is extracted from the output hazardous energy state prediction results: hazard type (e.g., backflow-accumulation coupling type), location of the main hazardous node (node ​​on the backflow path). and accumulation nodes ), and remaining safe time .

[0097] Real-time inversion of the energy coupling impedance between the return path and the accumulation nodes. The return path is the path where the antisymmetric component dominates in the energy transfer matrix between nodes. This is achieved through analysis... The node pair corresponding to the element with the largest absolute value is determined, and the return power is denoted as... (Take the absolute value); the accumulation node is the diagonal element in the energy transfer matrix between nodes. For a node whose energy is consistently positive and its magnitude is increasing, the rate of energy change of that node is denoted as the rate of change of energy. (A positive value indicates a net accumulation of energy).

[0098] Energy coupling impedance The strength of the interaction between the return energy and the accumulated energy is characterized by the ratio of the cross-power spectral density to the self-power spectral density of their fluctuations. To simplify real-time calculations, this invention employs cross-correlation analysis, specifically:

[0099] Acquire return power signal and accumulation node energy change rate signal Calculate the cross-correlation function for data within a time window (the length of which is half of the remaining safe time, but not less than 10 mechanical fluctuation cycles). Find the delay that maximizes cross-correlation. This time delay is the time it takes for energy to travel from the return path to the accumulation node, denoted as . Meanwhile, the peak value of the normalized cross-correlation is the coupling coefficient. The value range is [0,1].

[0100] Calculate the critical power dissipation required to decouple the positive feedback loop. A positive feedback loop means that the return energy accumulates more rapidly, which in turn further enhances the return flow. By introducing sufficient dissipation into the energy transmission path, the net energy gain can be made negative. The critical dissipation power is proportional to the geometric mean of the return power and the rate of change of accumulated energy, and inversely proportional to the propagation delay.

[0101]

[0102] In time delay If the dissipated energy equals the geometric mean of the return energy and the accumulated energy, then the positive feedback loop is broken. It should be noted that when the coupling coefficient... When the value is low (<0.3), the positive feedback is not significant, and a smaller correction coefficient (such as 0.6) can be used to multiply it. To avoid excessive intervention. In specific implementation, it can be based on... value pairs Perform linear scaling: .

[0103] It should be understood that the above formula can ensure that the critical power dissipation matches the dynamic characteristics (time delay) of the system, so that the decoupling control has sufficient feedforward compensation capability.

[0104] S42, dynamically adjust the torque current limit value of the motor driver according to the critical dissipation power, so that it decreases linearly from the rated value to the critical value. At the same time, inject an active damping current with the same frequency but opposite phase as the micro-vibration signal into the controllable damping device corresponding to the accumulation node, so as to safely release the return energy and accumulated energy along the reconstructed virtual dissipation path while maintaining the continuous operation of the mechanical equipment, until the dangerous energy state drops below the warning threshold.

[0105] Motor drives (such as frequency converters or servo drives) typically have torque and current limiting parameters. This parameter is used to limit the maximum electromagnetic torque output of the motor. This invention utilizes this parameter to actively reduce the power injected into the transmission chain from the drive end, thereby weakening the source of return energy.

[0106] First, calculate the current limit value corresponding to the critical power dissipation: ,in This is the torque constant of the motor (unit: Nm / A). This is the current motor speed. If... If the current is less than 10% of the motor's rated current, then 10% should be taken as the lower limit to prevent the motor from losing steps or stopping unexpectedly.

[0107] Then, based on the remaining safe time The slope of the current limiting is determined. A linear descent strategy is adopted: Let the current time be... It is expected that in Limit the current from the rated value within a certain time. linearly decreasing to ,Right now:

[0108]

[0109] Maintain thereafter Until the dangerous energy state is eliminated. Throughout the descent process, the actual torque and current of the motor are monitored in real time. If it appears If the load continues to approach the limit and the motor speed drops significantly, it indicates that the load is too heavy. The descent speed should be slowed down appropriately (by multiplying by a coefficient of 0.8) to avoid causing severe vibration of the equipment.

[0110] It should be understood that by adjusting the torque current limit, the motor's output capacity is reduced, and the electromagnetic power absorbed from the grid is decreased, so that the drive end no longer becomes an energy source in the return path. The feedback energy that might have flowed back into the motor is now forced to dissipate in the transmission chain, thus enabling subsequent active damping.

[0111] A controllable damping device is installed at the location corresponding to the accumulation node. The controllable damping device can be a magnetorheological damper (response time <10ms), an eddy current brake (response time <5ms), or an active piezoelectric actuator (response time <1ms), selected according to the power level and frequency range of the equipment. Taking a magnetorheological damper as an example, its damping force can be continuously adjusted by the excitation current.

[0112] Real-time acquisition of micro-vibration signals at accumulation nodes. These signals originate from acoustic emission sensors or high-frequency accelerometers. The dominant vibration frequency is extracted using a bandpass filter. (For example, the frequency where the micro-vibration energy is most concentrated is typically 1kHz-10kHz). The controller (FPGA or high-speed DSP) generates a reference signal with the same frequency but out of phase as the micro-vibration signal: ,in The phase of the measured micro-vibration signal, Phase shift achieves phase inversion.

[0113] Amplitude of active damping current Closed-loop regulation is performed based on the deviation between the critical dissipation power and the current rate of change of accumulated energy. A proportional-integral controller is used.

[0114]

[0115] Among them, error , For empirical tuning parameters (e.g.) A / W, A / (W·s)). When When, increase This increases the damping force, converting more vibrational energy into heat; conversely, it decreases the damping force. To avoid excessive damping affecting the normal operation of the equipment.

[0116] It should be understood that through the injection of active damping, the vibrational energy of the accumulation node is actively converted into heat energy (through viscous dissipation of the medium inside the damper or heating by eddy currents), thereby preventing the elastic strain energy from continuing to accumulate and reducing the rate of change of accumulated energy. It gradually approaches zero.

[0117] Through the aforementioned parallel actions, the return energy no longer forces its way into the drive end. Instead, it forms the first virtual dissipation path through the motor windings and the braking resistor (when the motor operates in generator mode due to torque current limiting, the energy is fed back to the DC bus via the inverter and dissipated by the braking resistor). The accumulated energy is then converted into heat dissipation through active damping, forming the second virtual dissipation path. Since the equipment remains rotating or moving (without stopping), energy can be continuously and controllably released, avoiding the impact energy release that may be caused by traditional emergency shutdowns.

[0118] During the execution of interlocking actions, the remaining safety time and characteristic change rate are monitored in real time. The calculation is repeated in each control cycle. and If the dangerous energy state continues to deteriorate (e.g.) Accelerate shortening and If the torque current limit increases instead of decreasing, the interlocking strategy will be automatically upgraded: the torque current limit will be further reduced (potentially down to...). (50% of the rated value), while increasing the upper limit of the active damping current amplitude to 1.5 times the rated value. If the upgrade still cannot curb the deteriorating trend and If the time is less than 1 second, the final safety measure will be triggered, which is to urgently cut off the power source and activate the mechanical brake, while releasing the high-pressure stored energy through the explosive bolt or safety valve.

[0119] While performing the above interlocking actions, the sensor weights and interlocking logic are dynamically adjusted. The specific method is as follows:

[0120] Increase the weights of sensors at hazardous nodes (return and accumulation nodes) and on their adjacent paths to 2-5 times their initial weights. For example, in the energy transfer matrix, multiply the weights of torque and vibration sensors on all paths directly connected to hazardous nodes by a coefficient. Furthermore, in subsequent Kalman filtering or feature fusion, the measurements from these sensors are given higher confidence, thus more sensitively capturing subtle changes in dangerous energy states. It should be understood that the specific weighting factor can be determined based on the physical distance between the sensor and the dangerous node: the closer the distance, the higher the factor; if the distance exceeds 10 wavelengths of stress wave propagation, no adjustment is made.

[0121] Under normal operating conditions, to prevent false triggering, safety interlocks typically employ AND gate logic, meaning the interlock is only triggered when multiple sensors simultaneously exceed a threshold. When a hazardous state has been confirmed, to improve response speed, the interlock decision logic can be dynamically switched from AND gate to OR gate triggering: as soon as any sensor associated with the hazardous node (such as torque, vibration, or acoustic emission) exceeds a threshold, the corresponding level of interlock action is immediately executed. Simultaneously, the trigger threshold of this OR gate is temporarily reduced to 80% of its normal value, further shortening the response delay.

[0122] When the dangerous energy state drops below the warning threshold (i.e.) Seconds and After that, the sensor weights are gradually restored to their initial values, the interlock logic is switched back to the AND gate, the torque current limit value slowly rises to the rated value (the rise rate does not exceed 10% / second of the rated value), the active damping current drops to zero, and the equipment smoothly returns to normal operation.

[0123] Compared to traditional tiered shutdown methods, this invention can proactively resolve the positive feedback coupling of energy backflow and accumulation while maintaining continuous equipment operation, significantly reducing the risk of catastrophic accidents, and avoiding the impact of frequent shutdowns on production efficiency.

[0124] like Figure 3 As shown, this application embodiment also provides a computer device, including a processor 201 and a memory 202, wherein the memory 202 stores a computer program, and the computer program, when executed by the processor 201, implements the method as described in any of the preceding claims.

[0125] The computer equipment can be an industrial control computer, an embedded industrial control computer, a programmable logic controller (PLC), an FPGA controller, or an edge computing gateway. The processor 201 can be an x86 architecture processor, an ARM architecture processor, or a RISC-V architecture processor, with a clock frequency that meets real-time control requirements, typically not less than 1 GHz (for complex model inference) or employing an FPGA parallel pipeline architecture to achieve microsecond-level response. The memory 202 includes non-volatile storage media (such as NAND Flash, SSD, eMMC) and volatile memory (such as DDR4 SDRAM). The non-volatile storage media stores an operating system (such as real-time Linux, VxWorks, or bare-metal programs), deep learning runtime libraries (such as TensorFlow Lite Micro, STM32Cube.AI), and the computer programs described in this invention.

[0126] This application also provides a computer program product including instructions that, when executed, cause a computer device to perform the method as described in any of the preceding claims.

[0127] The computer program product can be a software package stored on a physical medium, such as an optical disc (CD / DVD), USB flash drive, flash memory card, or external hard drive; it can also be firmware pre-installed inside a computer device, such as the method logic of this invention already burned into the factory program of a PLC or embedded controller; or it can be an executable file or container image (such as a Docker image) downloaded via the Internet, which the user obtains from a cloud server or software repository and deploys on an edge computing node.

[0128] The instructions include, but are not limited to: data acquisition thread scheduling instructions, Kalman filter matrix operation instructions, neural network forward propagation instructions, energy transfer matrix update instructions, and driver communication protocol stack instructions. These instructions can be written in high-level languages ​​(C / C++, Python, Rust) and cross-compiled to generate target platform machine code; or they can be written in hardware description languages ​​(Verilog / VHDL) and synthesized into FPGA configuration files.

[0129] This application also provides a storage medium storing the computer program product as described above.

[0130] The storage media include, but are not limited to: magnetic storage media (such as hard disk drives (HDDs), floppy disks, and magnetic tapes), optical storage media (such as DVD-Rs and Blu-ray discs), and semiconductor storage media (such as solid-state drives (SSDs), eMMC, UFS, SD cards, NAND flash memory, NOR flash memory, EEPROM, FeRAM, and MRAM). The storage media can be components integrated within a computer device (such as the BIOS Flash chip on the motherboard) or external pluggable components (such as CF cards, SD cards, and USB flash drives). The storage media can also be part of network-attached storage (NAS) or cloud storage, allowing computer devices to remotely access computer program products within it via network protocols (such as iSCSI, NFS, and SMB).

[0131] The above description represents the preferred embodiments of the present invention. It should be noted that, for those skilled in the art, various improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for safety interlocking control of mechanical equipment using multi-sensor information fusion, characterized in that, Includes the following steps: Collect sensor data of mechanical equipment, including motion position, speed, acceleration, force, and torque, and perform time alignment, amplitude normalization, and short-term anomaly filtering on each sensor data. Based on the processed sensor data, the kinetic energy, potential energy, and frictional energy consumption of each energy node of the mechanical equipment are estimated, and an inter-node energy transfer matrix that can reflect the energy flow direction, energy accumulation, and reverse transmission state is constructed accordingly. Data from various sensors and the energy transfer matrix between nodes are integrated to extract abnormal energy features, including short-term energy surge features, local energy reflection features, abnormal fluctuation features of frictional energy, and energy accumulation features accompanied by micro-vibration. Based on the abnormal energy features, the prediction results of dangerous energy states are obtained. Based on the predicted dangerous energy state and the current operating status of the mechanical equipment, the safety interlock action is adaptively triggered, and the sensor weights and interlock logic are dynamically adjusted to achieve proactive intervention and advanced protection for dangerous energy states.

2. The method for safety interlocking control of mechanical equipment based on multi-sensor information fusion according to claim 1, characterized in that, Based on the processed sensor data, the kinetic energy, potential energy, and frictional energy consumption of each energy node in the mechanical equipment are estimated. Based on this, an inter-node energy transfer matrix reflecting energy flow direction, energy accumulation, and reverse transmission status is constructed, including: Mechanical equipment is divided into energy nodes. Combining the mass, inertia and structural stiffness of the nodes, the kinetic energy, potential energy and frictional energy consumption of the nodes are calculated by using the velocity, acceleration, displacement and force or torque data collected by sensors. At the same time, the connection relationship between nodes and the direction of energy transfer are considered to form an energy distribution. The energy distribution of each node is mapped and weighted according to the connection between nodes and the direction of energy flow, so as to construct the energy transfer matrix between nodes that can reflect the energy flow direction, energy accumulation and reverse transmission status between nodes.

3. The method for safety interlocking control of mechanical equipment based on multi-sensor information fusion according to claim 1, characterized in that, The data from each sensor and the energy transfer matrix between nodes are fused to extract anomalous energy features, including: The energy change rate of each node within the time window is calculated, and the change in the rated power change rate where the energy growth rate exceeds a preset multiple within multiple consecutive control cycles is recorded as a short-term energy surge characteristic. The proportion of reverse energy components is calculated by using the symmetry decomposition of the energy transfer matrix between nodes, and the values ​​of reverse proportion exceeding the set threshold of forward energy and lasting for at least one mechanical fluctuation cycle are recorded as local energy reflection characteristics. For nodes with frictional energy consumption, the deviation amplitude is obtained by comparing the current frictional energy consumption value of the node with the historical energy dissipation baseline. Frequency band energy analysis is performed on the vibration acceleration signal corresponding to the same node to obtain the high-frequency energy ratio. The abnormal fluctuation characteristics of frictional energy are obtained by combining the deviation amplitude and the high-frequency energy ratio. By analyzing the micro-vibration signals and accumulated energy at nodes or paths, the energy accumulation caused by micro-vibration is recorded as the energy accumulation characteristic accompanied by micro-vibration.

4. The method for safety interlocking control of mechanical equipment based on multi-sensor information fusion according to claim 3, characterized in that, Based on the aforementioned abnormal energy characteristics, the predicted dangerous energy states are obtained, including: The characteristics of short-term energy surge, local energy reflection, abnormal fluctuation of frictional energy, and energy accumulation accompanied by micro-vibration are jointly registered in time and space: If the intensity of a single feature exceeds its dynamic threshold and shows an upward trend, it is determined to be a single hazardous energy state; if the ratio of the time difference between the occurrence of any two or more features at the same energy node or adjacent nodes to the mechanical impact propagation delay is less than 1 and the feature intensities all show an upward trend, it is determined to be a composite hazardous energy state. Based on the hazardous energy state determination results, the extrapolation time difference between the current characteristic change rate and the preset hazardous boundary value is used to output the hazardous energy state prediction results.

5. The method for safety interlock control of mechanical equipment based on multi-sensor information fusion according to claim 4, characterized in that, The predicted hazardous energy state results include the hazardous type, the location of the hazardous node, and the remaining safe time.

6. The method for safety interlock control of mechanical equipment based on multi-sensor information fusion according to claim 1, characterized in that, Based on the predicted hazardous energy state and the current operating status of the mechanical equipment, adaptive safety interlock actions are triggered, including: When the predicted dangerous energy state is a composite dangerous energy state, the energy coupling impedance between the return path and the accumulation node is inverted in real time, and the critical dissipation power required to decouple the positive feedback loop is calculated. The torque current limit of the motor driver is dynamically adjusted according to the critical dissipation power, so that it decreases linearly from the rated value to the critical value. At the same time, an active damping current with the same frequency but opposite phase as the micro-vibration signal is injected into the controllable damping device corresponding to the accumulation node. Under the premise of maintaining the continuous operation of the mechanical equipment, the return energy and the accumulated energy are safely discharged along the reconstructed virtual dissipation path until the dangerous energy state drops below the warning threshold.

7. The method for safety interlocking control of mechanical equipment based on multi-sensor information fusion according to claim 6, characterized in that, The critical dissipation power is equal to the geometric mean of the return power and the rate of change of accumulated energy, multiplied by the reciprocal of the time delay of the energy path.

8. A computer device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the method as described in any one of claims 1-7.

9. A computer program product, comprising instructions, characterized in that, When the computer program product is executed, the instructions cause the computer device to perform the method as described in any one of claims 1-7.

10. A storage medium, characterized in that, The computer program product as described in claim 9 is stored thereon.

Images