Software and hardware layered isolation redundancy safety control and diagnosis method for engineering machinery
By aligning and analyzing the control data of engineering machinery, and using the ant colony algorithm to search for the compensation responsibility rotation sequence, the abnormal association layer is determined and the compensation amount is reallocated. This solves the problem of difficulty in distinguishing the source of fault in the existing technology, and achieves more accurate fault diagnosis and stable redundant control.
Patent Information
- Application Number
- CN202611314370.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-27
- Publication Date
- 2026-11-17
AI Technical Summary
Existing engineering machinery control systems, under multi-layer compensation and redundant control conditions, have difficulty accurately distinguishing the source of faults. This causes abnormal compensation relationships to continue to affect the target actuators after channel switching, making it difficult to achieve accurate fault diagnosis and control.
By acquiring the target control quantities, hierarchical compensation quantities, and operational response data of the main control channel and backup control channel, performing time alignment and validity marking, establishing a candidate mechanical state set and an initial compensation responsibility association matrix, using the ant colony algorithm to search for the compensation responsibility rotation sequence, calculating residuals and determining the abnormal association layer, constructing a safety control space, selecting and identifying actions and reallocating compensation quantities, and generating safety control instructions.
It improves the accuracy of fault diagnosis and the reliability of redundant control, reduces the transmission of fault misjudgment and abnormal compensation, and ensures the consistency and stability of control.
Smart Images

Figure CN123113284A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of engineering machinery safety control technology, and in particular to a method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation. Background Technology
[0002] During the operation of construction machinery, actuators such as hydraulic cylinders and hydraulic motors are typically driven by a controller based on the target control quantity. The controller then uses operational response data such as hydraulic pressure, actuated displacement, drive current, and mechanical posture to determine whether the actual action meets the requirements. To reduce control errors, existing control systems also compensate or correct the control quantity at the application layer, interface layer, and firmware layer. When the main control channel experiences a significant fault, communication anomaly, or operating parameters exceed limits, the system typically switches to a backup control channel after fault diagnosis to maintain control of the target actuator.
[0003] However, in actual operation, construction machinery is often simultaneously affected by factors such as load changes, hydraulic response lag, sensor drift, interface transmission delay, and hydraulic circuit leakage. These anomalies may produce similar changes in hydraulic pressure, actuator displacement, or drive current. For example, a slowdown in actuator movement could be caused by hydraulic circuit leakage, control command transmission delay, or actuator lag itself. Furthermore, application-layer compensation, interface-layer correction, and firmware-layer drive compensation all work together to affect the actual output control quantity. When an anomaly occurs at one level, other levels may continue to increase compensation, temporarily masking the anomaly. Therefore, relying solely on control deviation at a single moment or a single operational response data point can easily lead to difficulties in identifying the source of the fault.
[0004] When the source of the fault is not accurately determined, even after switching from the main control channel to the backup control channel, the backup control channel may still continue to use compensation parameters, feedback data, or execution links related to the anomaly, causing the original anomaly to continue acting on the target actuator after the control channel switch. The fundamental reason is that existing processing methods struggle to distinguish the correspondence between different compensation levels and mechanical operating responses, and also struggle to determine which compensation relationships should be retained and which should be isolated before switching redundant control channels.
[0005] Therefore, how to accurately determine the actual source of control anomalies when multiple layers of compensation and redundant control coexist, and how to prevent the compensation relationship related to the anomaly from continuing to affect the target actuator after the redundant control channel is switched, is a technical problem that needs to be solved. Summary of the Invention
[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a hardware and software layered isolation method for redundant safety control and diagnosis of engineering machinery, thereby solving the technical problems mentioned in the background section.
[0007] To achieve the above objectives, the present invention provides the following technical solution: A hardware and software layered isolation redundancy safety control and diagnosis method for engineering machinery includes the following steps: S1. Obtain the target control quantity, hierarchical compensation quantity and operation response data of the main control channel and the backup control channel, perform time alignment and validity marking, and establish a candidate mechanical state set and an initial compensation responsibility association matrix; S2. Determine the motion safety boundary and target total compensation amount based on the candidate mechanical state set and operation response data. Within the motion safety boundary, transfer the compensation share of the application layer, interface layer and firmware layer in sequence, and provide equal compensation by the other compensation layers to form a compensation responsibility rotation state set. S3. Calculate the motion residual, drive residual, energy transfer residual and time delay residual of the compensation responsibility rotation state, determine the abnormal association layer or abnormal association component according to the residual migration relationship, update the candidate mechanical state credibility and compensation responsibility association matrix, and form a corrected candidate mechanical state set and fault compensation isolation relationship. S4. Construct a safety control space based on the modified candidate mechanical state set, fault compensation isolation relationship and motion safety boundary, select and execute identification actions, match the actual identification response with the predicted identification response, and determine the credible mechanical state and the credibility ranking result of redundant control channels. S5. Select the target redundant control channel based on the trusted mechanical state, the redundancy control channel trust ranking results, and the fault compensation isolation relationship. Reallocate the compensation amount of each compensation level, generate safety control commands, execute restricted normal control or safety degradation control, and verify the redundant safety takeover results.
[0008] S1 specifically includes: edge computing nodes of the main control channel and backup control channel collect target control quantity, hierarchical compensation quantity, actual output control quantity and operation response data according to a unified control cycle, forming raw hierarchical control data and raw operation response data; based on the control cycle identifier and the actual output control quantity transmission time, perform time alignment, abnormal data identification, unit conversion and validity marking on the data, forming a hierarchical control alignment dataset and continuous control interval identifier; establish a candidate mechanical state set with initial credibility based on the hierarchical control alignment dataset, and establish an initial compensation responsibility correlation matrix according to the correlation degree between each compensation level and the operation response.
[0009] S2 specifically includes: determining the motion safety boundary, target total compensation amount, and compensation rotation constraints based on the hierarchical control alignment dataset, continuous control interval identifiers, candidate mechanical state set, and equipment calibration data; based on the compensation rotation constraints, using the application layer, interface layer, and firmware layer as compensation nodes, using the ant colony algorithm to search for the compensation transfer source layer, compensation transfer target layer, compensation transfer step size, state holding time, and recovery order, generating a compensation responsibility rotation sequence that meets the rotation coverage conditions; the execution control channel transfers the compensation amount of each compensation level in equal amounts according to the compensation responsibility rotation sequence, and the shadow control channel performs independent calculations synchronously, collecting the actual output control amount and operation response data corresponding to each transfer state, and restoring the previous effective allocation state when the safety parameters exceed the limit, forming a compensation responsibility rotation state set.
[0010] S3 specifically includes: reading the set of compensation responsibility rotation states; calculating motion residuals, drive residuals, energy transfer residuals, and time delay residuals based on the hierarchical compensation amount and operation response data to form a rotation residual feature sequence; matching the rotation residual feature sequence with the compensation responsibility rotation sequence to determine whether the residuals migrate with the compensation share, and determining the abnormal association layer or abnormal association component based on the residual migration relationship and abnormal response to form the abnormal migration path and fault source location result; updating the candidate mechanical state credibility based on the abnormal migration path and fault source location result to form a corrected candidate mechanical state set, and setting a prohibition compensation flag in the initial compensation responsibility association matrix to obtain the fault compensation isolation relationship.
[0011] S4 specifically includes: constructing a safety control space based on the modified candidate mechanical state set, fault compensation isolation relationship, and motion safety boundary; generating candidate identification actions; and using the ant colony algorithm to select identification actions with high state distinguishability and low control risk; inputting the identification actions into the main control channel and the backup control channel to generate predicted identification responses corresponding to each modified candidate mechanical state; having the control channel with actual output authority execute the identification actions; collecting actual identification responses and marking data validity; calculating the matching error between the actual identification response and each predicted identification response; determining the credible mechanical state based on the fault source location result; and generating a redundancy control channel credibility ranking result based on the prediction deviation, response delay, and fault compensation isolation relationship verification results of the main control channel and the backup control channel.
[0012] S5 specifically includes: verifying the takeover qualification of the control channel based on the trusted mechanical state, the credibility ranking results of redundant control channels, the fault compensation isolation relationship, and the compensation responsibility association matrix; selecting the target redundant control channel; reallocating the compensation amount of each compensation level to form the target compensation responsibility allocation result; generating safety control commands by the target redundant control channel based on the target compensation responsibility allocation result and the motion safety boundary; executing restricted normal control when the predicted response meets the motion safety boundary, otherwise executing safety degradation control to form the redundant safety control result; collecting the running response after executing the safety control command within the takeover verification window; calculating the control verification error between it and the expected safety control response; confirming the redundant safety takeover result according to the control verification conditions; and re-identifying or executing controlled shutdown when verification fails.
[0013] The beneficial effects of this invention are as follows: This invention distinguishes the sources of anomalies caused by the mutual masking of multiple layers of compensation by associating hierarchical control data, operational responses, and fault compensation relationships between the main control channel and the backup control channel. This reduces fault misjudgment and the continued transmission of anomaly compensation during channel switching, thereby improving the accuracy of fault diagnosis and the reliability of redundant control. By unifying the control cycle, the invention achieves time alignment, validity marking, and establishment of candidate mechanical states for hierarchical control data and operational responses. This ensures that all types of control quantities and responses maintain the same time reference, reducing timing misalignment and invalid data interference, and improving the consistency of state judgment and the accuracy of subsequent diagnostic data.
[0014] This invention utilizes an ant colony algorithm to search for a compensation responsibility rotation sequence within the motion safety boundary and implements equal compensation transfer between the application layer, interface layer, and firmware layer. This keeps the actual output control quantity under control, obtains comparable responses under different compensation assumption states, and improves identification safety and state distinguishability. By calculating motion residuals, drive residuals, energy transfer residuals, and time delay residuals, and analyzing the migration relationship of residuals with compensation share, it distinguishes control-level anomalies from anomalies in sensors, actuators, communication interfaces, and hydraulic circuits, reducing fault source location errors and improving the consistency of candidate mechanical state updates.
[0015] This invention improves the accuracy of reliable mechanical state confirmation and the reliability of redundant control channel sequencing by selecting identification actions within the safety control space commonly satisfied by modified candidate mechanical states and comparing the actual identification response with the predicted identification response, thus eliminating similar fault states without exceeding the motion safety boundary. By verifying the takeover eligibility of redundant control channels, isolating prohibited compensation relationships, and reallocating the allowed compensation amount, combined with restricted normal control, safety degradation control, and takeover verification at the motion safety boundary, the risk of fault compensation inheritance and control mutation is reduced, improving takeover stability and control consistency. Attached Figure Description
[0016] Figure 1 This is a flowchart of the engineering machinery redundancy safety control and diagnosis method with hardware and software layered isolation according to the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Example: Figure 1 As shown, this embodiment provides a method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation, including the following steps: S1. Obtain the target control quantity, hierarchical compensation quantity and operation response data of the main control channel and the backup control channel, perform time alignment and validity marking, and establish a candidate mechanical state set and an initial compensation responsibility association matrix; S2. Determine the motion safety boundary and target total compensation amount based on the candidate mechanical state set and operation response data. Within the motion safety boundary, transfer the compensation share of the application layer, interface layer and firmware layer in sequence, and provide equal compensation by the other compensation layers to form a compensation responsibility rotation state set. S3. Calculate the motion residual, drive residual, energy transfer residual and time delay residual of the compensation responsibility rotation state, determine the abnormal association layer or abnormal association component according to the residual migration relationship, update the candidate mechanical state credibility and compensation responsibility association matrix, and form a corrected candidate mechanical state set and fault compensation isolation relationship. S4. Construct a safety control space based on the modified candidate mechanical state set, fault compensation isolation relationship and motion safety boundary, select and execute identification actions, match the actual identification response with the predicted identification response, and determine the credible mechanical state and the credibility ranking result of redundant control channels. S5. Select the target redundant control channel based on the trusted mechanical state, the redundancy control channel trust ranking results, and the fault compensation isolation relationship. Reallocate the compensation amount of each compensation level, generate safety control commands, execute restricted normal control or safety degradation control, and verify the redundant safety takeover results.
[0019] S1 specifically includes the following sub-steps: S110: Independent edge computing nodes for the main control channel and backup control channel collect raw hierarchical control data and raw operation response data, using a unified control cycle as the basic processing unit. The unified control cycle is a fixed processing time interval set by the edge computing nodes according to the safety control tasks of the engineering machinery, and each unified control cycle has a unique control cycle identifier.
[0020] The original hierarchical control data includes target control quantity, basic control quantity, application layer compensation quantity, interface layer correction quantity, firmware layer drive compensation quantity, and actual output control quantity. Among them, the target control quantity is generated by the application layer based on the control handle command, work task command, or safety control command, and is used to characterize the target position, target speed, target pressure, or target torque. The basic control quantity is calculated from the target control quantity and the effective operating response of the previous unified control cycle. The application layer compensation quantity is calculated from the deviation between the target control quantity and the controlled response quantity. The interface layer correction quantity is determined by the sensor calibration coefficient, message integrity, and sampling quality. The firmware layer drive compensation quantity is determined by the proportional valve dead zone, drive current deviation, and target actuator response hysteresis.
[0021] The above three compensation quantities are converted into percentages of the rated control range according to the equipment calibration data. The converted values are collectively referred to as the equivalent quantity of the control command, and the unit is .
[0022] The actual output control quantity for the kth unified control cycle is:
[0023] In the formula, This represents the actual output control quantity in the k-th unified control cycle; Indicates the basic control quantity; Indicates the application layer compensation amount; Indicates the interface layer correction amount; This indicates the amount of compensation for the firmware layer driver.
[0024] The raw operational response data includes hydraulic pressure, actuated displacement, drive current, mechanical attitude, and response delay, which originate from pressure sensors, displacement or angle sensors, current detection units, inertial measurement units, and communication monitoring units, respectively. The response delay is the time difference between the actual output control quantity being sent and the moment when the corresponding mechanical response first exceeds the upper limit of the measurement noise, measured in milliseconds (ms). Each data point is accompanied by a data source identifier, acquisition timestamp, and initial data validity flag.
[0025] For the same target actuator, pressure, displacement, and current data are all bound by the execution link identifier to avoid cross-attribution of response data between the main control channel and the backup control channel.
[0026] The original hierarchical control data and original operation response data are generated and sent to S120.
[0027] S120: The edge computing nodes of the main control channel and the backup control channel respectively perform time alignment, abnormal data identification, unit conversion and data validity update on the original hierarchical control data and the original operation response data. The two edge computing nodes do not share the status judgment results.
[0028] Using the actual output control signal transmission time and control cycle identifier as alignment references, hydraulic pressure, actuated displacement, drive current, mechanical attitude, and response delay are mapped to the corresponding unified control cycle. When the target time is between two valid sampling times and the adjacent sampling interval is no greater than twice the rated sampling cycle, linear interpolation is used to obtain the alignment value.
[0029] In the formula, This represents the alignment value for the k-th unified control cycle; and These represent the valid sampled values before and after the target time; This indicates the target time corresponding to the unified control cycle; and These represent the sampling times before and after.
[0030] When the sampling interval is greater than twice the rated sampling period, no interpolation is performed, and the corresponding data is marked as unavailable. When the hydraulic pressure exceeds the sensor's physical range, the change in displacement exceeds the maximum allowable change of the target actuator, the timestamp is earlier than the previous sampling time, or the control cycle identifier is inconsistent, it is marked as exceeding the range, jumping, or timing abnormality, respectively. A single communication failure data can be compensated based on the valid data before and after it, but no compensation is made for two or more consecutive abnormal data.
[0031] Hydraulic pressure drop, sluggish displacement, and continuous deviation of drive current, which may reflect real faults, are only marked as abnormal features and are not deleted. Hydraulic pressure, displacement, drive current, mechanical attitude, and response delay are standardized to MPa, mm or °, A, °, and ms, respectively.
[0032] A continuous control interval is defined as a continuous control cycle in which the target action type remains unchanged, the target action direction is consistent, the control channel is not switched, and the direction of the executed displacement is consistent with the direction of the target action. The start and end identifiers of the interval are recorded. Finally, a hierarchical control alignment dataset and continuous control interval identifiers are formed and sent to S130 and S210, while also being saved for use by S310 and S530.
[0033] S130. Within each continuous control interval, establish a set of candidate mechanical states and an initial compensation responsibility correlation matrix using the hierarchical control alignment dataset. Candidate mechanical states are state hypotheses used to explain the source of the current control deviation, but which have not yet been confirmed by identified actions. Each candidate mechanical state includes a state type, associated physical object, state characteristic parameters, expected response relationship, and initial confidence level. State types include at least sensor drift, target actuator hysteresis, target hydraulic circuit leakage, interface transmission delay, and actual load variation.
[0034] The conditions for establishing the sensor drift state are that the actual output control quantity and drive current remain stable, while the sensor baseline under the same mechanical posture continues to shift; the conditions for establishing the target actuator hysteresis state are that the drive current and hydraulic pressure have changed effectively, while the moment when the executed displacement begins to change exceeds the normal response delay; the conditions for establishing the target hydraulic circuit leakage state are that the actual output control quantity and drive current are stable, the hydraulic pressure continues to decrease, and the executed displacement changes along the load direction; the conditions for establishing the interface transmission delay state are that the application layer output time is normal, the interface message arrival time is delayed, and the firmware layer drive compensation quantity and drive current are synchronously delayed; the conditions for establishing the actual load change state are that the interface delay and the target actuator response timing are normal, while the hydraulic pressure and drive current change synchronously along the direction that maintains the same target speed.
[0035] Normal response delay, normal pressure range, and normal current range are all read from the normal reference dataset, which includes equipment factory calibration data, fault-free recalibration data after maintenance, and historical normal operation data verified by the takeover team. Initial confidence is determined by the proportion of the number of valid unified control cycles that meet the conditions for establishing the corresponding state to the total number of valid unified control cycles in the continuous control interval. Multiple candidate mechanical states are allowed for the same abnormal phenomenon, and the source of the fault is not uniquely determined in this step.
[0036] The rows of the initial compensation responsibility association matrix correspond to candidate mechanical states, and the columns correspond to the application layer, interface layer, and firmware layer, respectively. The matrix elements are calculated by normalizing and weighting the corresponding level compensation ratio, the consistency rate between compensation changes and operational response changes, and the degree of deviation from the historical normal compensation range. The weight coefficient of the corresponding level compensation ratio is preferably set to 0.2~0.3, the weight coefficient of the consistency rate is preferably set to 0.4~0.5, and the weight coefficient of the degree of deviation is preferably set to 0.2~0.3. The sum of the three weight coefficients is 1. In a specific embodiment, the weights of the above three items can be 0.3, 0.5, and 0.2, respectively.
[0037] The candidate mechanical state set is sent to S210 and S220, and the initial compensation responsibility association matrix is saved to the edge computing node for S330 to call.
[0038] S2 specifically includes the following sub-steps: S210. Based on the continuous control interval identifier and hierarchical control alignment dataset output by S120, the candidate mechanical state set output by S130, and the equipment calibration data, determine the motion safety boundary, total target compensation amount, and compensation rotation constraints for the current continuous control interval. The motion safety boundary is the range of multiple parameters allowed during the compensation responsibility rotation period, including actual speed, hydraulic pressure, stabilizing torque margin, expected braking distance, target actuator position, and end-effector motion deviation. The stabilizing torque margin is the difference between the anti-overturning moment and the overturning moment, expressed in N·m.
[0039] Rated speed, rated hydraulic pressure, mechanical position limit, braking deceleration, structural dimensions, and rated load are read from the equipment calibration data; the actual speed is obtained by dividing the difference in executed displacement between adjacent unified control cycles by the duration of the unified control cycle; the current working load is converted based on the hydraulic pressure, the effective working area of the hydraulic cylinder, and the transmission ratio of the mechanism; the stabilizing torque margin is calculated based on the current working load, mechanical posture, support position, and structural dimensions; when the current target action involves walking, the expected braking distance is calculated based on the actual speed, braking deceleration, slope, and response delay; when the target action only involves the working device under fixed chassis conditions, the expected braking distance is marked as a non-participating parameter.
[0040] The edge computing node uses the rated safety parameters as its outer boundary and shrinks the corresponding range according to the candidate mechanical state: when there is a candidate mechanical state of leakage in the target hydraulic circuit, the upper limit of the allowable pressure change rate is reduced by a shrinkage coefficient of 0.6~0.8 (preferably 0.7); when there is a candidate mechanical state of hysteresis in the target actuator, the single action amplitude is limited to 50%~70% (preferably 60%) of the rated action amplitude; when there is a candidate mechanical state of sensor drift, the allowable tolerance range of cross-validation of related sensors is reduced by 20%~30% to improve the validation requirements.
[0041] The target total compensation amount for the kth unified control cycle is:
[0042] In the formula, This represents the total target compensation amount for the k-th unified control cycle; , and These represent the application layer compensation amount, the interface layer correction amount, and the firmware layer driver compensation amount, respectively. Indicates the actual output control quantity; This represents the basic control quantity.
[0043] The compensation rotation constraints include the allowable deviation of the total target compensation amount, the allowable deviation of the actual output control amount, the motion safety boundary, the upper limit of a single compensation transfer, the state holding time, the longest rotation duration, and the candidate mechanical state identification conditions.
[0044] The above results are sent to S220 and S230, and the motion safety boundary and total target compensation are saved for S410, S510, S520 and S530 to call.
[0045] S220. Based on the compensation rotation constraints, generate a compensation responsibility rotation sequence using the ant colony algorithm. A compensation node is the compensation-bearing position in the application layer, interface layer, or firmware layer; the compensation transfer source layer is the layer that reduces the compensation amount; the compensation transfer target layer is the layer that increases the compensation amount by the same amount; the compensation transfer path is a directed connection from the compensation transfer source layer to the compensation transfer target layer; the compensation transfer step size is the equivalent amount of the control command corresponding to one transfer; each ant corresponds to one candidate compensation responsibility rotation scheme consisting of the compensation transfer source layer, compensation transfer target layer, compensation transfer step size, state maintenance time, and recovery order.
[0046] First, candidate compensation transition step sizes are generated according to a preset ratio of the target total compensation amount. If a candidate compensation transition step size is less than the minimum identifiable compensation amount, it is adjusted to the minimum identifiable compensation amount. If the adjusted step size is greater than the upper limit of a single compensation transition, the rotation search of the current continuous control interval is stopped. The minimum identifiable compensation amount is determined based on the expected response relationship of the candidate mechanical states, ensuring that the difference in predicted response between at least two candidate mechanical states exceeds the larger of the corresponding sensor measurement error and the upper limit of historical normal fluctuations.
[0047] Calculate the rotation cost for the r-th candidate compensation responsibility rotation scheme:
[0048] In the formula, Let r represent the rotation cost of the r-th candidate compensation responsibility rotation scheme; Indicates the normalized terminal motion bias; This represents the normalized safety out-of-bounds penalty value, which is 0 if no out-of-bounds violation occurs. Indicates the number of normalized compensation transfers; This indicates the difference in predicted response of the normalized candidate mechanical state. , , and This indicates non-negative weights, which are selected to prioritize security and discriminative power in the search, based on the condition of meeting the criteria. > > > In a specific engineering debugging example, take =0.3, =0.4, =0.1, =0.2; This indicates that a positive number with a denominator of 0 should be avoided, for example, a value of 0.001.
[0049] The end-effector deviation is divided by the allowable end-effector deviation, the safety overrun is divided by the corresponding safety boundary, the number of compensated transfers is divided by the maximum allowable number of transfers, and the predicted response difference is divided by the larger of the measurement error and the upper limit of normal fluctuation, in order to eliminate the influence of dimensions.
[0050] Each candidate path initializes the pheromone with the same positive value, and the heuristic information is provided by... and and The ratio of the sums is determined; after each round of search, the pheromone of all paths is reduced according to the volatility coefficient, and the pheromone is increased on paths with lower rotation costs. The search stops when the same minimum rotation cost is obtained for consecutive preset rounds, the maximum number of iterations is reached, or a scheme that satisfies the rotation coverage condition is formed. The rotation coverage condition is that each compensation level participating in the diagnosis is at least one compensation transfer source layer and one compensation transfer target layer. The compensation responsibility rotation sequence is output and sent to S230.
[0051] S230. The compensation responsibility rotation sequence is simultaneously distributed to the edge computing nodes of the main control channel and the backup control channel. The hardware arbitration unit determines the control channel currently holding actual output authority as the execution control channel, and the other control channel as the shadow control channel. The execution control channel actually adjusts the hierarchical compensation amount and outputs it to the target execution mechanism. The shadow control channel independently calculates the prediction result based on the same sequence. At any given time, only one control channel is allowed to hold actual output authority.
[0052] Before each transfer of compensation responsibility, the execution control channel confirms that all safety parameters are within the motion safety boundary; then, within the same unified control cycle, the compensation amount of the compensation transfer source layer is reduced, and the compensation amount of the compensation transfer target layer is increased by an equal amount, so that the total target compensation amount is kept within the allowable deviation.
[0053] After the transfer is completed, the current hierarchical compensation allocation is maintained at no less than the maximum calibrated response time of the target hydraulic circuit, and the actual output control quantity, hydraulic pressure, actuated displacement, drive current, mechanical attitude, and response delay are collected cycle by cycle. The compensation responsibility rotation state is the effective control state after completing one compensation responsibility transfer and passing the motion safety boundary verification. It includes the rotation state identifier, compensation transfer source layer, compensation transfer target layer, compensation transfer step size, three hierarchical compensation quantities, running response data, state start and end times, and data validity flags.
[0054] If any safety parameter exceeds its limit, critical data becomes unavailable, the actual output control deviation exceeds the allowable value, or the safety judgments of two edge computing nodes are inconsistent, the execution control channel restores the previous effective hierarchical compensation allocation in reverse order and marks the current state as invalid. For example, if the maximum calibrated response time of the target hydraulic circuit is 60ms and the unified control cycle is 20ms, at least three unified control cycles should be maintained after the transfer before determining the pressure and displacement response to avoid mistakenly using data before the transfer as the response after the transfer.
[0055] The set of compensation responsibility rotation states, along with the hierarchical compensation amount, operational response data, and data validity flags corresponding to each state, are sent to S310. If the rotation coverage conditions are not met or the state difference does not exceed the measurement error, the system returns to S220 to search again.
[0056] S3 specifically includes the following sub-steps: S310 reads the set of compensation responsibility rotation states output by S230, and calculates the rotation residual only for compensation responsibility rotation states where the data validity flag is valid, the state holding time is not less than the maximum calibrated response time of the target hydraulic circuit, and no over-limit backoff is triggered. The edge computing node reads the target control quantity, actual output control quantity, three-level compensation quantity, hydraulic pressure, execution displacement, drive current, mechanical attitude, and response delay in each state, without re-collecting data outside the state holding interval.
[0057] The actual operating response with the same physical meaning as the target control quantity is defined as the controlled response quantity: when the target control quantity is position, the controlled response quantity is the executed displacement; when the target control quantity is speed, the controlled response quantity is the ratio of the difference in executed displacement between adjacent unified control cycles to the duration of the unified control cycle; when the target control quantity is pressure, the controlled response quantity is hydraulic pressure; when the target control quantity is torque, the controlled response quantity is converted from hydraulic pressure, the effective working area of the hydraulic cylinder, and the lever arm of the mechanism.
[0058] Normal reference values are selected from the normal reference dataset. When there are no completely identical working conditions, the position difference, mechanical attitude difference, action direction difference, and load range difference of the target actuator are calculated separately. The continuous parameters are normalized according to the corresponding allowable variation range and then summed. The normal working condition data with the smallest sum of normalized differences is selected.
[0059] The j-th type of signed normalized residual and its magnitude in the q-th effective compensation responsibility rotation state are:
[0060] In the formula, This represents the signed normalized residual; Indicates the normalized residual magnitude; Indicates the actual response value; Indicates the target value or normal reference value; This indicates the permissible deviation specified in the equipment calibration data; This indicates that positive numbers with a denominator of 0 should be avoided.
[0061] Motion residuals are calculated from the controlled response quantity and the target control quantity; drive residuals are calculated from the actual drive current and the expected drive current obtained from the actual output control quantity, drive temperature, and firmware parameters; energy transfer residuals are calculated from the actual hydraulic pressure change and the expected hydraulic pressure change obtained from the drive current, mechanical attitude, and load range; time delay residuals are calculated from the actual response time delay and the normal response time delay of the same target actuator, action direction, and load range.
[0062] The rotation status identifier, compensation transfer source layer, compensation transfer target layer, compensation transfer step size, three-item hierarchical compensation amount, signed normalized residual and normalized residual amplitude are arranged according to the compensation responsibility rotation sequence to form a rotation residual feature sequence and sent to S320.
[0063] S320. Match the rotation residual characteristic sequence with the compensation responsibility rotation sequence item by item to identify whether the abnormal characteristics migrate with the hierarchical compensation responsibility. The compensation share is the proportion of the equivalent quantity of the control instruction at a certain compensation level to the absolute value of the total target compensation amount; when the total target compensation amount is less than the minimum effective compensation amount, the change in compensation amount at that level is directly used.
[0064] Edge computing nodes calculate the changes in compensation share, signed normalized residual change, and normalized residual amplitude between adjacent effective compensation responsibility rotation states; the change in residual amplitude that is greater than the larger value between the sensor measurement error and the upper limit of historical normal fluctuation is defined as the effective residual change.
[0065] For the same compensation level, if it participates in at least two effective compensation responsibility transfers and in no less than 70% of the effective state transfers, the signed normalized residual maintains the same monotonic relationship with the compensation share of that level, and returns to the allowable fluctuation range before the rotation after restoring the initial tiered compensation allocation, then the corresponding abnormal characterization is determined to have migrated with the compensation responsibility of that level. The 70% directional consistency rate is determined by the trade-off between the mislocation rate and the missed location rate in known fault injection tests.
[0066] When the motion residual changes only with the application layer compensation share, the application layer is identified as an abnormal association layer; when the delay residual changes with the interface layer correction amount and the interface message arrival time changes synchronously, the interface layer is identified as an abnormal association layer; when the drive residual or energy transfer residual changes with the firmware layer drive compensation amount and the interface message delay is normal, the firmware layer is identified as an abnormal association layer.
[0067] When residuals persist but migration conditions are not met, abnormal associated components are identified based on the combination of operational responses: if the hydraulic pressure change exceeds the pressure measurement error, but the drive current, actuated displacement, and mechanical attitude do not show matching changes, the corresponding pressure sensor is identified as an abnormal associated component; if the drive current and hydraulic pressure have changed, but the actuated displacement is continuously delayed beyond the normal response time, the target actuator is identified as an abnormal associated component; if the actual output control quantity and drive current are stable, the hydraulic pressure continues to decrease, and the actuated displacement changes along the load direction, the target hydraulic circuit is identified as an abnormal associated component; if the application layer output is always normal, but the interface message is continuously delayed, the communication interface is identified as an abnormal associated component.
[0068] Record the initial compensation level, changes in compensation share, the type of residual that first changes effectively, subsequent operational response and recovery results in rotation order, forming an abnormal migration path and fault source location results including abnormal association layer, abnormal association component and supporting residual type, and send them to S330.
[0069] S330. Update the credibility of candidate mechanical states based on the abnormal migration path and fault source location results. For the i-th candidate mechanical state, calculate the residual inconsistency degree between its expected residual change direction and the residual feature sequence of the rotation. The degree of inconsistency between the state type, association level, and associated physical object and the fault source location results. The percentage of effective state transitions that do not conform to the expected direction is used as... When the fault source category, the abnormal association layer, and the abnormal associated component are all consistent The value is 0. If there is one inconsistency, the value is 0.5. If there are two or more inconsistencies, the value is 1.
[0070] The updated credibility is:
[0071] In the formula, This represents the update confidence level of the i-th candidate machine state; Indicates the initial credibility; and This indicates non-negative weights. Since residual inconsistency has a stronger indicative effect on confidence decay, it is preferable to set it to... =1.5~2.5 (for example, take 2.0), =0.8~1.2 (for example, take 1.0); and Let represent the residual inconsistency and the source inconsistency of the i-th candidate mechanical state, respectively; , and represents the initial confidence, residual inconsistency, and source inconsistency of the nth candidate mechanical state, respectively; n represents the summation index of the candidate mechanical states; N represents the total number of candidate mechanical states.
[0072] Candidate mechanical states with a confidence level not lower than the state retention threshold are added to the revised candidate mechanical state set; those below the state exclusion threshold are excluded; those between the two thresholds are temporarily retained and their lower identification priority is recorded. In this embodiment, the state retention threshold is preferably set to 0.6~0.75 (e.g., 0.70), and the state exclusion threshold is preferably set to 0.2~0.3 (e.g., 0.25). When all candidate mechanical states are below the state retention threshold, the abnormal migration path, fault source location result, and exclusion reason are returned to S130 for expanding the state type or adjusting the state characteristic parameters.
[0073] In the initial compensation responsibility association matrix, the edge computing node sets the matrix elements corresponding to the abnormal association layer and the matrix elements that depend on the feedback of the abnormal association components as the prohibition of compensation flags, and obtains the updated compensation responsibility association matrix; based on the prohibition of compensation flags, the allowed feedback data and the allowed execution links, a fault compensation isolation relationship is formed.
[0074] The corrected candidate mechanical state set, the update reliability of each state, the fault compensation isolation relationship and the fault source location result are sent to S410-S430. The updated compensation responsibility association matrix and the abnormal migration path are sent to S510 and saved for S530 to update.
[0075] S4 specifically includes the following sub-steps: S410: Based on the motion safety boundary saved in S210, the set of corrected candidate mechanical states output in S330, the update reliability of each state, and the fault compensation isolation relationship, construct the safety control space and select the identification actions. The safety control space is the set of identification actions that, under the constraint of the fault compensation isolation relationship, after executing any corrected candidate mechanical state, predict the actual speed, hydraulic pressure, stabilizing torque margin, expected braking distance, and target actuator position will not exceed the motion safety boundary.
[0076] Action identification is a constrained control sequence temporarily superimposed on the current actual output control quantity to distinguish and correct candidate mechanical states. It includes action object, action direction, action amplitude, action duration, state holding time, and recovery method.
[0077] The edge computing node reads the rated control range, minimum executable control increment, maximum control change rate, and mechanical position limit of the target actuator from the device calibration data. Using the latest valid operating response saved in S230 and S310 as the initial state, it generates positive and negative control increments within the allowable control range and performs discrete combinations of action amplitude, action duration, and recovery change rate. Combinations of hierarchical compensation quantities, feedback data, or execution links that rely on prohibited use are directly deleted.
[0078] For each candidate identified action, the hydraulic pressure, execution displacement, drive current, actual speed, stabilizing torque margin, and expected braking distance during the execution and recovery periods are predicted using the expected response relationship of all corrected candidate mechanical states. The action is only included in the safety control space if all prediction results are within the motion safety boundary.
[0079] Ant colony algorithm is used to select identification actions from the safety control space. Each candidate identification node corresponds to one complete identification action, and each ant corresponds to an identification action sequence consisting of an excitation segment, a hold segment, and a recovery segment. All candidate identification nodes are initialized with the same positive value for pheromone. The heuristic information is determined by the ratio of state discrimination to normalized control risk.
[0080] The minimum state discriminant of the l-th candidate action is:
[0081] In the formula, The minimum state discrimination of the l-th candidate action is represented; i and n represent the indices of different corrected candidate mechanical states; m represents the response type index; M represents the total number of response types; and Let represent the m-th predicted identification response of the i-th and n-th modified candidate mechanical states under the l-th identification action, respectively; The non-negative weights represent the m-th type of response. Preferably, larger weights can be assigned to the displacement and hydraulic pressure, for example, a displacement weight of 0.4, a pressure weight of 0.4, and a drive current weight of 0.2, and all... The sum is 1; This represents the larger of the sensor measurement error and the upper limit of historical normal fluctuations. This indicates that positive numbers with a denominator of 0 should be avoided.
[0082] Risk control is calculated by normalizing the predicted response to the motion safety boundary, the motion amplitude, and the motion duration. Specifically, risk control can be defined as the weighted sum of the proportion of the predicted response value to the safety boundary value, the proportion of the motion amplitude to the maximum allowable amplitude, and the proportion of the motion time to the maximum allowable time. After each search round, the pheromone of all candidate identification nodes is reduced according to the volatility coefficient, and the pheromone is increased for candidate identification nodes with high state distinguishability and low risk. The search stops when the same preferred identification action is obtained for a preset number of consecutive rounds, the maximum number of iterations is reached, or the state distinguishability meets the identification requirements.
[0083] A higher identification weight is assigned to the state pairs with high update credibility that have not yet been ruled out, and the security control space and identification action are obtained and sent to S420.
[0084] S420. The identification action is simultaneously input into the edge computing nodes of the main control channel and the backup control channel, and the two edge computing nodes independently generate the predicted identification response. The identification response window is the time interval from the start of the identification action to the completion of the recovery through the maximum calibrated response time of the target hydraulic circuit; the reference value before identification is the average value of the corresponding running response within the continuous effective unified control cycle before the identification action is executed, and the number of sampling cycles is determined based on the response fluctuation stabilization time in the normal reference dataset.
[0085] Each edge computing node takes the actual output control quantity, hydraulic pressure, execution displacement, drive current, mechanical attitude and response delay before identification as initial values, and takes the direction, amplitude, duration and recovery method of the identified action as input. According to the expected response relationship of each corrected candidate mechanical state, it generates the predicted pressure change, predicted displacement change, predicted drive current change and predicted response delay change relative to the reference value before identification on a cycle basis. These are collectively referred to as the predicted identification response.
[0086] The execution control channel authorized by the hardware arbitration unit actually outputs the identified action through the execution link allowed in the fault compensation isolation relationship. The shadow control channel only performs prediction and does not output to the target actuator. In each unified control cycle, the execution control channel collects hydraulic pressure, execution displacement, drive current, mechanical attitude and response delay, and converts them into changes relative to the reference value before identification to form the actual identification response.
[0087] During the identification action, if the actual speed, hydraulic pressure, stabilizing torque margin, expected braking distance, or target actuator position enters the warning range of the motion safety boundary, or if key data is marked as unavailable, the identification action is immediately terminated, and the output control quantity is restored to the actual output before identification according to the preset recovery rate of change; the aborted data is marked as invalid, and the aborted reason is returned to S410 to delete the corresponding candidate identification action.
[0088] During full execution, the identification action identifier, execution control channel identifier, execution link identifier, pre-identification baseline value, predicted identification response, actual identification response, identification response window start and end times, and data validity flag will be sent to S430.
[0089] For example, if the hydraulic pressure before identification is 18MPa and rises to 18.5MPa after the identification action is executed, the actual pressure identification response is recorded as 0.5MPa, thus avoiding the initial pressure base value from masking the differences between candidate mechanical states.
[0090] S430. Determine the reliable mechanical state based on the predicted and actual identification responses, and generate a ranking result of the reliability of redundant control channels. For the i-th corrected candidate mechanical state, extract the maximum change, steady-state change, start response time, and recovery time of hydraulic pressure, actuated displacement, drive current, and response delay within the identification response window, and calculate the matching error:
[0091] In the formula, This represents the matching error of the i-th corrected candidate mechanical state; This represents the m-th type of response feature in the actual identification response; This indicates the corresponding predictive identification response features; This indicates the permissible prediction error determined by equipment factory fault injection tests, post-maintenance verification tests, or known operating data. Indicates non-negative weights; M represents the total number of response types; This indicates that positive numbers with a denominator of 0 should be avoided.
[0092] If the matching error of the candidate mechanical state is not greater than the state matching threshold (preferably 0.15~0.25 in this embodiment, for example 0.2), the difference between the candidate mechanical state and the second smallest matching error is not less than the minimum confirmation difference (preferably 0.1~0.15 in this embodiment, for example 0.12), the fault source is consistent with the fault source location result of S320, and the identification data is valid, then the candidate mechanical state is determined as a reliable mechanical state.
[0093] If at least two corrective candidate mechanical states meet the state matching threshold but cannot be uniquely confirmed, return to S410 to select a new identification action; if all corrective candidate mechanical states do not meet the state matching threshold, return the actual identification response and the reason for elimination to S130 to re-establish the candidate mechanical states; if the maximum number of identifications is reached but unique confirmation is still not possible, send the remaining states as unresolved mechanical states to S510 to perform control within the common safety range.
[0094] For the primary and backup control channels, first verify whether they meet the prohibited compensation flag, prohibited feedback data, or prohibited execution link in the fault compensation isolation relationship, and verify the parameter version, self-test results, command check value, and actual output permission status; control channels that do not meet any of these requirements are disqualified from taking over. For the remaining control channels, calculate the channel reliability based on its prediction deviation of the trusted machine state, prediction calculation delay, control response delay, and communication integrity rate of the most recent effective unified control cycle; the higher the value, the earlier the takeover order.
[0095] The system generates a ranking result of the redundancy control channel credibility, which includes control channel identifier, prediction deviation, prediction calculation delay, control response delay, configuration verification result, self-test result, channel credibility, and takeover qualification. The system then sends the credible mechanical status or unresolved mechanical status and the ranking result of the redundancy control channel credibility to S510.
[0096] S5 specifically includes the following sub-steps: S510. Based on the trusted mechanical state or unresolved mechanical state output by S430, the redundancy control channel trust ranking result, and the fault compensation isolation relationship and updated compensation responsibility association matrix output by S330, determine the target redundant control channel, trusted takeover data packet, and target compensation responsibility allocation result.
[0097] The primary and backup control channels each generate a control configuration snapshot. This snapshot includes the three currently loaded hierarchical compensation parameters, feedback data source identifier, execution link identifier, compensation responsibility association matrix version, and verification value. The control configuration snapshot is compared item by item with the fault compensation isolation relationship. If a prohibited compensation flag, prohibited feedback data, or prohibited execution link is detected, or if the compensation responsibility association matrix version is inconsistent, self-test fails, or verification value is incorrect, the corresponding control channel is deemed ineligible to take over. If neither control channel is ineligible to take over, the system directly enters the S520 Level 4 safety degradation control.
[0098] A trusted takeover data packet is a set of state data that is allowed to be transmitted to the target redundant control channel and has excluded untrusted data. This includes trusted or unresolved mechanical states, fault source location results, fault compensation isolation relationships, updated compensation responsibility association matrices, current target control variables, latest basic control variables, latest actual output control variables, latest effective operating responses, motion safety boundaries, control cycle identifiers, and verification values. It does not include parameters with prohibited compensation flags, prohibited feedback data, or unconfirmed integral states. The current total target compensation is recalculated based on the latest actual output control variable and the basic control variable.
[0099] For the k-th unified control cycle, the first... There are one compensation level, and the redistributed compensation amount is as follows:
[0100] In the formula, This indicates the amount of compensation after redistribution; This indicates the current total target compensation amount; This indicates a compensation permission mark; a value of 1 indicates permitted use, and a value of 0 indicates prohibited use. Indicates the first The credibility of each compensation level; and These represent the compensation permission flag and trust level of the g-th compensation level, respectively; g represents the summation index of the application layer, interface layer, and firmware layer.
[0101] The reliability of the compensation level is calculated based on the normalized weighted average of the corresponding matrix elements, abnormal migration paths, and recent data effectiveness. The optimal weighting ratio for these three factors is 4:4:2, i.e., the weight of the matrix element is 0.4, the weight of the abnormal migration path matching degree is 0.4, and the weight of the recent data effectiveness is 0.2. When the denominator is 0, the three layered compensation quantities are set to 0, and the system enters the fourth level of security degradation control.
[0102] Before acquiring actual output authority, the target redundant control channel undergoes independent pre-run calculations. If the difference between the calculated actual output control quantity and the current effective output control quantity is not greater than the takeover command jump threshold, the hardware arbitration unit first revokes the actual output authority of the original execution control channel and then grants actual output authority to the target redundant control channel. The takeover command jump threshold is obtained by multiplying the maximum allowable control change rate of the target actuator by the uniform control cycle duration, and it is not greater than the maximum single-cycle control increment specified in the equipment calibration data.
[0103] Send the target redundant control channel, trusted takeover data packet, target compensation responsibility allocation result, and actual output permission switching result to S520.
[0104] S520: The target redundant control channel generates safety control instructions based on the trusted takeover data packet, the target compensation responsibility allocation result, the fault compensation isolation relationship, and the motion safety boundary. The safety control instructions are instruction data packets containing the actual output control quantity, control cycle identifier, control parameter version, execution link identifier, and checksum.
[0105] The target redundancy control channel reads the current target control quantity, recalculates the basic control quantity based on the reliable mechanical state, adds the hierarchical compensation quantity allowed in the target compensation responsibility allocation result, deletes the correction item that depends on prohibited feedback data, limits the change of instructions according to the maximum control change rate in the equipment calibration data, and determines the upper and lower limits of the actual output control quantity based on the motion safety boundary.
[0106] Before outputting safety control commands, the corresponding hydraulic pressure, driven displacement, drive current, actual speed, stabilizing torque margin, and expected braking distance are predicted based on the reliable mechanical state. When the mechanical state is undecided, predictions are made based on all undecided mechanical states, and their common allowable control range is adopted. If all prediction results are within the motion safety boundary and no prohibited items are invoked, restricted normal control is executed. Restricted normal control maintains the target action under the condition that the fault compensation isolation relationship remains effective, and does not indicate the release of fault isolation.
[0107] When safety cannot be maintained, safety degradation control is implemented sequentially according to the degree of constraint: Level 1 reduces the target speed and the rate of change of safety control commands; Level 2 further limits the working load and prohibits multiple actuators from operating in parallel; Level 3 only allows the target actuator to move in the direction that reduces the degree of danger; Level 4 smoothly reduces the actual output control quantity to a safe value according to the shutdown change rate calibrated by the equipment and cuts off the power output.
[0108] The target redundant control channel reads the operational response in each unified control cycle. If any safety parameter reaches the motion safety boundary, critical data becomes unavailable, a safety control command hits a prohibited item, or the actual output permission is abnormal, the system immediately switches to a more stringent safety degradation control level. For example, if the trusted mechanical state is that the target hydraulic circuit is leaking and the boom can only safely descend, lifting and increasing the firmware layer drive compensation are prohibited, and level 3 safety degradation control is executed. If the rate of hydraulic pressure decrease continues to increase during descent, the system switches to level 4 safety degradation control.
[0109] The redundant safety control results, which include a sequence of safety control instructions, target redundancy control channels, target compensation responsibility allocation results, safety degradation control levels, permitted feedback data and execution links, and actual operational responses, are generated and sent to S530.
[0110] S530. Set up the takeover verification window and verify the redundant safety control results. The takeover verification window starts when the target redundant control channel obtains actual output permission, and its duration is the larger of the target hydraulic circuit's maximum calibrated response time and the preset verification period; the preset verification period is determined by the time required for the running response to reach stability during the takeover test.
[0111] Edge computing nodes read safety control commands, hydraulic pressure, executed displacement, drive current, mechanical attitude, response latency, actual output permission status, safety degradation control level, and data validity flags according to a unified control cycle within the takeover verification window. The expected safety control response is generated cycle-by-cycle based on the trusted mechanical state, the actual output safety control commands, the initial takeover operation response, equipment calibration data, and the current safety degradation control level.
[0112] The control verification error for the kth unified control cycle is:
[0113] In the formula, This indicates the control of verification error; This represents the actual runtime response of the m-th type; This indicates the expected response to the corresponding security control measures; This indicates the permissible response error determined by the acceptance testing. Indicates non-negative weights; M represents the total number of response types; This indicates that positive numbers with a denominator of 0 should be avoided.
[0114] The redundant safety takeover is confirmed to be successful when the number of consecutive effective unified control cycles is not less than the rounded-up value obtained by dividing the takeover verification window length by the unified control cycle duration, and the control verification error of each unified control cycle is not greater than the control verification threshold, all safety parameters are within the motion safety boundary, the actual output control quantity does not undergo a sudden change exceeding the takeover command jump threshold, no prohibited items are invoked, and the target redundant control channel uniquely holds the actual output authority. The control verification threshold is used to determine whether the actual operation deviates from the safety expectation. Its specific value is set according to the allowable tolerance of the equipment. In this embodiment, the control verification threshold is preferably set to 0.10~0.20 (for example, 0.15).
[0115] The verified data is written into the hierarchical control alignment dataset as a verified fault control sample, but not into the normal reference dataset; only the matrix elements without the prohibited compensation flag and the state feature parameters and confidence of the trusted mechanical states are updated.
[0116] When verification fails, the target redundant control channel maintains the actual output authority and outputs a safety hold command. At the same time, it returns the control verification error, failure response type, latest running response, takeover failure reason and remaining re-identification count to S410 to update the safety control space and exclude the invalid identification actions that have been verified. When the maximum number of re-identification counts determined by the equipment safety test is reached, or when any safety parameter exceeds the motion safety boundary during verification, re-identification stops and Level 4 safety degradation control is executed.
[0117] The final result is a redundant safety control and fault diagnosis result for engineering machinery, including the trusted mechanical state or unresolved mechanical state, fault source location result, fault compensation isolation relationship, target redundancy control channel, actual output permission switching record, target compensation responsibility allocation result, safety degradation control level, control verification error and takeover verification conclusion. The updated hierarchical control alignment dataset, compensation responsibility association matrix and trusted mechanical state are saved as historical inputs when executing S120, S130 and S510 in the next continuous control interval.
[0118] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.
[0119] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0120] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0121] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for redundant safety control and diagnosis of engineering machinery with hardware and software layering and isolation, characterized in that, Includes the following steps: S1. Obtain the target control quantity, hierarchical compensation quantity and operation response data of the main control channel and the backup control channel, perform time alignment and validity marking, and establish a candidate mechanical state set and an initial compensation responsibility association matrix; S2. Determine the motion safety boundary and target total compensation amount based on the candidate mechanical state set and operation response data. Within the motion safety boundary, transfer the compensation share of the application layer, interface layer and firmware layer in sequence, and provide equal compensation by the other compensation layers to form a compensation responsibility rotation state set. S3. Calculate the motion residual, drive residual, energy transfer residual and time delay residual of the compensation responsibility rotation state, determine the abnormal association layer or abnormal association component according to the residual migration relationship, update the candidate mechanical state credibility and compensation responsibility association matrix, and form a corrected candidate mechanical state set and fault compensation isolation relationship. S4. Construct a safety control space based on the modified candidate mechanical state set, fault compensation isolation relationship and motion safety boundary, select and execute identification actions, match the actual identification response with the predicted identification response, and determine the credible mechanical state and the credibility ranking result of redundant control channels.
2. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation as described in claim 1, characterized in that, Also includes: S5. Select the target redundant control channel based on the trusted mechanical state, the redundancy control channel trust ranking results, and the fault compensation isolation relationship. Reallocate the compensation amount of each compensation level, generate safety control commands, execute restricted normal control or safety degradation control, and verify the redundant safety takeover results.
3. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation as described in claim 1, characterized in that, S1 specifically includes: The edge computing nodes of the main control channel and the backup control channel collect target control quantity, hierarchical compensation quantity, actual output control quantity and operation response data according to a unified control cycle, forming raw hierarchical control data and raw operation response data; Based on the control cycle identifier and the actual output control quantity transmission time, the data is time-aligned, abnormal data is identified, units are converted, and validity is marked to form a hierarchical control alignment dataset and a continuous control interval identifier.
4. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation as described in claim 3, characterized in that, Also includes: A set of candidate mechanical states with initial credibility is established based on the hierarchical control alignment dataset, and an initial compensation responsibility correlation matrix is established based on the degree of correlation between each compensation level and the operation response.
5. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation as described in claim 1, characterized in that, S2 specifically includes: Based on the hierarchical control alignment dataset, continuous control interval identifiers, candidate mechanical state sets, and equipment calibration data, determine the motion safety boundary, total target compensation amount, and compensation rotation constraints. Based on the compensation rotation constraints, with the application layer, interface layer, and firmware layer as compensation nodes, the ant colony algorithm is used to search for the compensation transfer source layer, compensation transfer target layer, compensation transfer step size, state maintenance time, and recovery order, generating a compensation responsibility rotation sequence that meets the rotation coverage conditions; The execution control channel transfers the compensation amount of each compensation level in equal amounts according to the compensation responsibility rotation sequence, and the shadow control channel performs independent calculations synchronously, collecting the actual output control amount and operation response data corresponding to each transfer state. When the safety parameters exceed the limit, the previous effective allocation state is restored, forming a set of compensation responsibility rotation states.
6. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation according to claim 1, characterized in that, S3 specifically includes: Read the set of compensation responsibility rotation statuses, and calculate motion residuals, drive residuals, energy transfer residuals, and time delay residuals based on the hierarchical compensation amount and operation response data to form a rotation residual feature sequence; By mapping the rotation residual feature sequence to the compensation responsibility rotation sequence, it is determined whether the residual migrates with the compensation share. Based on the residual migration relationship and abnormal response, the abnormal association layer or abnormal association component is determined, forming the abnormal migration path and fault source location result.
7. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation according to claim 6, characterized in that, Also includes: The credibility of candidate mechanical states is updated based on the abnormal migration path and fault source location results, forming a corrected candidate mechanical state set. A prohibition flag for compensation is set in the initial compensation responsibility association matrix to obtain the fault compensation isolation relationship.
8. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation according to claim 1, characterized in that, S4 specifically includes: A safety control space is constructed based on the modified candidate mechanical state set, fault compensation isolation relationship and motion safety boundary, candidate identification actions are generated, and the ant colony algorithm is used to select identification actions with high state discrimination and low control risk. The identification action is input into the main control channel and the backup control channel to generate the predicted identification response corresponding to each corrected candidate mechanical state. The identification action is executed by the control channel with actual output authority, the actual identification response is collected and the data validity is marked. The matching error between the actual identification response and each predicted identification response is calculated. The credible mechanical state is determined by combining the fault source location results. Based on the prediction deviation, response delay and fault compensation isolation relationship verification results of the main control channel and the backup control channel, the credibility ranking results of the redundant control channels are generated.
9. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation according to claim 2, characterized in that, S5 specifically includes: Based on the reliable mechanical state, the reliability ranking results of redundant control channels, the fault compensation isolation relationship and the compensation responsibility association matrix, the takeover qualification of the control channel is verified, the target redundant control channel is selected, the compensation amount of each compensation level is reallocated, and the target compensation responsibility allocation result is formed. The target redundancy control channel generates safety control commands based on the target compensation responsibility allocation results and motion safety boundaries. When the predicted response meets the motion safety boundaries, restricted normal control is executed; otherwise, safety degradation control is executed, resulting in redundant safety control.
10. The method for redundant safety control and diagnosis of engineering machinery with hardware and software layered isolation according to claim 9, characterized in that, Also includes: Within the takeover verification window, the operational response after executing the safety control command is collected, the control verification error between the response and the expected safety control response is calculated, the redundant safety takeover result is confirmed based on the control verification conditions, and the controlled shutdown is re-identified or executed when the verification fails.
