Robot remote monitoring and data management system based on cloud platform

Abstract

The invention discloses a robot remote monitoring and data management system based on a cloud platform, and the system carries out the real-time monitoring of the product quality detection data and process operation parameters of a corresponding process of a target robot, obtains a product defect rate according to a monitoring result, and marks the target robot as a fault state or a critical estimation state. When the product defect rate does not conform to a preset defect rate threshold value, the operation parameters of the target robot perform self-adaptive operation; performing multi-dimensional feature extraction on the vibration data of the target robot in the critical pre-estimation state, and marking the target robot as a normal state or a fault type early warning state; load rate analysis is carried out on the target robot in the normal state, cross-plant distributed cooperative scheduling is carried out on the target robot in the high-load or fault state, and maintenance duration analysis is carried out on the target robot in the fault type early warning state. Maintenance requirements and production plans are effectively balanced, the flexibility and rationality of scheduling decisions are ensured, and the overall production and operation efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of robot operation and maintenance technology, specifically a cloud-based robot remote monitoring and data management system. Background Technology

[0002] With the rapid development of intelligent manufacturing technology, robots have been widely used in various fields such as industrial production and logistics warehousing, becoming core equipment for improving production efficiency and ensuring product quality. However, in industrial production, existing robot remote monitoring and data management systems still face many technical bottlenecks in practical applications. Robot operation efficiency is affected by many factors (such as robot failure, unreasonable processes, material supply delays, and production line rhythm mismatch). Existing systems can only monitor single-point status, and bottleneck diagnosis relies on human experience, which is time-consuming (>8 hours) and has low accuracy. Furthermore, large manufacturing enterprises often have cross-plant layouts (such as main plant + branch plants, multiple production bases). Existing systems only support robot scheduling in a single plant area, resulting in: when a robot in a certain plant area fails or is fully loaded, it is impossible to call on idle robots in other plants to provide support, causing task backlog; and uneven resource allocation across multiple plants (some plants have idle robots, while others are overloaded). Summary of the Invention

[0003] To address the aforementioned technical problems, the present invention aims to provide a cloud-based robot remote monitoring and data management system, comprising a cloud monitoring platform, a data acquisition terminal, a sample database, a lightweight module, an adaptive operating parameter module, a fault early warning module, and a collaborative scheduling module. The lightweight module is used to monitor the product quality inspection data of the corresponding process of the target robot in real time, obtain the product defect rate, and monitor the process operation parameters of the target robot in real time. Based on the monitoring results, the target robot is marked as a fault state or a critical prediction state. The adaptive operation parameter module is used to build a parameter-state-quality lookup table. When the product defect rate does not meet the preset defect rate threshold, the adaptive operation parameter of the target robot is performed based on the target robot's process operation parameters, environmental interference data, and parameter-state-quality lookup table. The fault early warning module is used to build a fault prediction model, extract multi-dimensional features from the vibration data of the target robot in the critical prediction state, obtain time domain features, frequency domain features and time-frequency domain features and input them into the fault prediction model, and mark the target robot as normal state or fault type early warning state according to the fault prediction model. The collaborative scheduling module is used to analyze the load rate of target robots in normal state, extract the tag information, load rate, and corresponding task tags and material tags of each robot within the factory area, perform cross-factory distributed collaborative scheduling of target robots in high load or fault state based on tag information, load rate, task tags and material tags, and analyze the maintenance time of target robots in fault type warning state, and determine whether to perform cross-factory distributed collaborative scheduling based on the analysis results.

[0004] Furthermore, data acquisition terminals and lightweight modules are deployed on the target robot and the product production line. The cloud monitoring platform communicates with the data acquisition terminals and lightweight modules within the pre-defined factory area. The data acquisition terminals are used to transmit the collected data to the lightweight modules and the cloud monitoring platform. The cloud monitoring platform is equipped with a sample database, an adaptive operating parameter module, a fault early warning module, and a collaborative scheduling module. The sample database, the adaptive operating parameter module, the fault early warning module, and the collaborative scheduling module communicate with each other.

[0005] Furthermore, the process of obtaining product defect rates includes: Acquire the process operation parameters, vibration data, environmental interference data, and product quality inspection data of the corresponding process of the target robot in the current data collection cycle; Set a time window, perform statistical analysis on the product quality inspection data of the target robot within the time window, obtain the product defect rate. If the product defect rate within the time window is greater than the preset defect rate threshold, then perform adaptive operation of the running parameters.

[0006] Furthermore, the process of performing adaptive operations on the target robot's operating parameters includes: The sample database is used to extract the real-time operating parameters, environmental interference data and corresponding product quality inspection data of the target robot in several historical collection periods as sample data. A parameter-state-quality comparison table is constructed based on the sample data. The parameter-state-quality comparison table includes the product defect rate corresponding to different process operating parameters under different environmental interference data conditions. A state space is constructed based on the process operating parameters and environmental interference data of the current acquisition cycle and the parameter-state-quality comparison table. Several target operating parameters are randomly generated. A fitness function is constructed based on the state space and the target operating parameters. Chromosome encoding is performed on the several target operating parameters and the population is initialized to generate an initial population. Constraints are introduced. Based on the initial population, fitness function and constraints, the optimal process operating parameters are obtained through a multi-objective genetic algorithm. The optimal process operating parameters are marked as standard process operating parameters and fed back to the lightweight module.

[0007] Furthermore, the process of real-time monitoring of the target robot's operational parameters includes: When the lightweight module receives the standard process operating parameters, it removes the original standard process operating parameters and adjusts the process operating parameters of the target robot to the standard process operating parameters. Then, it compares the process operating parameters of the target robot with the standard process operating parameters in real time to obtain the parameter deviation time series. It presets the error range, compares the parameter deviation time series with the error range to obtain the cumulative time when the parameter deviation is not within the error range, and presets the cumulative time upper limit. When the cumulative time is greater than the cumulative time upper limit, it marks the target robot as a fault state, obtains the fault type, marks the process operating parameters, vibration data and environmental interference data of the target robot in the collection period as fault data, associates the fault data with the fault type and uploads them to the sample database, and performs cross-plant distributed collaborative scheduling for the target robot. When the cumulative time is less than or equal to the upper limit of the cumulative time, the target robot is marked as a critical prediction state.

[0008] Furthermore, a fault prediction model is constructed, and multi-dimensional feature extraction is performed on the vibration data of the target robot in the critical prediction state. Time-domain features, frequency-domain features, and time-frequency-domain features are obtained and input into the fault prediction model. The process of marking the target robot as a normal state or a fault type warning state according to the fault prediction model includes: Multi-dimensional feature extraction is performed on the vibration data of each component of the target robot in the current acquisition cycle to obtain time-domain features, frequency-domain features, and time-frequency-domain features. At the same time, auxiliary features are obtained based on the process operation parameters and environmental interference data of the target robot in the current acquisition cycle. A fault prediction model is constructed by extracting time-domain features, frequency-domain features, time-frequency-domain features, and auxiliary features of fault data and normal data from the sample database as training data. The fault prediction model is trained using the training data to obtain the completed fault prediction model. The time-domain features, frequency-domain features, time-frequency-domain features, and auxiliary features of the target robot during the current acquisition period are input into the fault prediction model. The fault prediction model outputs the occurrence probability of each fault type for each component of the target robot. The occurrence probability of each fault type is compared with a preset probability threshold. If the occurrence probability of any fault type is greater than the probability threshold, a warning signal for the fault type is generated, and the target robot is marked as being in a fault type warning state. If the occurrence probability of each fault type is less than or equal to the probability threshold, the target robot is marked as being in a normal state. The process operation parameters, vibration data, and environmental interference data of the target robot during the acquisition period are marked as normal data and uploaded to the sample database.

[0009] Furthermore, the process of performing load rate analysis on a target robot in a normal state includes: When the target robot is in normal condition, obtain the tag information of the target robot and other robots within the factory area, as well as the production plan of the corresponding process. Based on the production plan, obtain the load rate of the target robot and the task tag and material tag of the corresponding process.

[0010] The load rate range is divided into sub-ranges with different load levels by selecting a threshold point. The load rate of the target robot is determined to be within the sub-range to obtain the load level of the target robot. If the load level of the target robot is high load, a distributed collaborative scheduling operation is performed.

[0011] Furthermore, the process of cross-plant distributed collaborative scheduling of high-load target robots based on tag information, load rate, task tags, and material tags includes: Extract the tag information, load rate, and corresponding process task and material tags of each other robot within the factory area. Match the tag information of each other robot with the corresponding process task tags based on the target robot's tag information and the corresponding process task tags to filter out other robots that meet the preset requirements. Based on the tag information, load rate, and corresponding task and material tags of the target robot and other robots that meet the preset requirements, several optimization functions are constructed. Several task handover schemes are randomly generated. Chromosome encoding and population initialization are performed on the several task handover schemes to generate an initial population. The dynamic optimization weights of each optimization function are obtained. Based on the dynamic optimization weights and several optimization functions, an objective function is constructed. Based on the initial population and the objective function, a multi-objective genetic algorithm is used to obtain the target task handover scheme. The task handover process is executed according to the target task handover scheme.

[0012] Furthermore, the process of obtaining the dynamic optimization weights of each optimization function includes: The order priority and logistics channel congestion coefficient are obtained based on the task tags and material tags of the corresponding process of the target robot. The historical failure rate of the target robot is obtained based on the failure data of the target robot in several historical collection periods in the sample database. The safety coefficient of the target robot is obtained based on the tag information and historical failure rate of the target robot. Construct a weight quantization table for several optimization functions. Obtain the weight coefficients corresponding to several optimization functions based on order priority, logistics channel congestion coefficient, safety coefficient, and the weight quantization table. Normalize the weight coefficients corresponding to several optimization functions to generate dynamic optimization weights for several optimization functions.

[0013] Furthermore, the process of analyzing the maintenance time of target robots in fault type warning states and determining whether to execute cross-plant distributed collaborative scheduling based on the analysis results includes: When the target robot is marked as a fault type warning state, the maintenance duration of the fault type is obtained, the production plan of the target robot and the corresponding process of each other robot on the production line is obtained, the production interruption window duration of the target robot is obtained based on the production plan, and it is determined whether the maintenance duration of the fault type is less than the production interruption window duration. If it is less, the maintenance operation is performed on the target robot. If it is not less, cross-plant distributed collaborative scheduling is performed on the target robot and the maintenance operation is performed on the target robot.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. Real-time monitoring of product quality inspection data and process operation parameters through lightweight modules enables dynamic tracking of product defect rates; relying on parameter-state-quality comparison tables and multi-objective genetic algorithms, process operation parameters are accurately adapted to different environmental interference conditions, avoiding quality problems caused by parameter mismatch from the source; at the same time, parameter standards are continuously calibrated through a closed-loop optimization mechanism to ensure the stability and consistency of product quality and reduce production losses caused by quality fluctuations.

[0015] 2. The fault early warning module uses multi-dimensional feature extraction and professional fault prediction models to accurately assess robots in critical predicted states, identify potential fault risks in advance and clarify fault types, breaking through the limitations of the lag in traditional fault detection. Combined with the continuous accumulation of historical fault data and model iteration, it further improves the accuracy of fault identification, provides a scientific basis for operation and maintenance decisions, effectively reduces the impact of sudden faults on production, and ensures the continuity of production processes.

[0016] 3. The collaborative scheduling module achieves global visual management of resource status based on the tag information, load rate, and task and material tags of robots throughout the plant. For robots with high load or fault status, it optimizes task allocation through cross-plant distributed collaborative scheduling, making full use of idle resources and avoiding local overcapacity and resource waste. At the same time, it combines the analysis of production interruption window duration to balance maintenance needs and production plans, ensuring the flexibility and rationality of scheduling decisions and improving overall production and operation efficiency. Attached Figure Description

[0017] Figure 1 This is a flowchart of a cloud-based robot remote monitoring and data management system, which is an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] like Figure 1 As shown, the cloud-based robot remote monitoring and data management system includes a cloud monitoring platform, a data acquisition terminal, a lightweight module, a sample database, an adaptive operating parameter module, a fault early warning module, and a collaborative scheduling module. The lightweight module is used to monitor the product quality inspection data of the corresponding process of the target robot in real time, obtain the product defect rate, and monitor the process operation parameters of the target robot in real time. Based on the monitoring results, the target robot is marked as a fault state or a critical prediction state. The adaptive operation parameter module is used to build a parameter-state-quality lookup table. When the product defect rate does not meet the preset defect rate threshold, the adaptive operation parameter of the target robot is performed based on the target robot's process operation parameters, environmental interference data, and parameter-state-quality lookup table. The fault early warning module is used to build a fault prediction model, extract multi-dimensional features from the vibration data of the target robot in the critical prediction state, obtain time domain features, frequency domain features and time-frequency domain features and input them into the fault prediction model, and mark the target robot as normal state or fault type early warning state according to the fault prediction model. The collaborative scheduling module is used to analyze the load rate of target robots in normal state, extract the tag information, load rate, and corresponding task tags and material tags of each robot within the factory area, perform cross-factory distributed collaborative scheduling of target robots in high load or fault state based on tag information, load rate, task tags and material tags, and analyze the maintenance time of target robots in fault type warning state, and determine whether to perform cross-factory distributed collaborative scheduling based on the analysis results.

[0020] It should be further explained that, in the specific implementation process, data acquisition terminals and lightweight modules are deployed on the target robot and the product production line. The cloud monitoring platform communicates with the data acquisition terminals and lightweight modules within the preset factory area (including several factory areas). The data acquisition terminals are used to transmit the collected data to the lightweight modules and the cloud monitoring platform. The cloud monitoring platform is equipped with a sample database, an adaptive operating parameter module, a fault early warning module, and a collaborative scheduling module. The sample database, the adaptive operating parameter module, the fault early warning module, and the collaborative scheduling module communicate with each other.

[0021] It should be further explained that, in the specific implementation process, the process of real-time monitoring of product quality inspection data for the corresponding process of the target robot to obtain the product defect rate includes: Acquire the following process operation parameters for the target robot's current acquisition cycle (total time for a single product to complete all processes, such as the time for a welding robot to complete one body weld or the time for a cutting robot to process one part): welding current, cutting speed, assembly pressure, vibration data (sampling frequency 1000Hz, measurement range 0.1~10kHz, such as vibration acceleration signals of bearing outer ring and gearbox input end), environmental interference data (temperature and humidity, power grid voltage fluctuations), and corresponding process product quality inspection data (dimensional deviations of visual inspection, defect levels of non-destructive testing, and weight tolerances). Set a time window (the total time for 30 individual products to complete the entire process), statistically analyze the product quality inspection data of the target robot within the time window, and obtain the product defect rate (the ratio of the number of products identified as defective within the time window to the total number of products identified; when the dimensional deviation of visual inspection, the defect level of non-destructive testing, or the weight tolerance does not meet the preset requirements, it is identified as a defective product). If the product defect rate within the time window is greater than the preset defect rate threshold, then the operating parameters will be adaptively adjusted.

[0022] It should be further explained that, in the specific implementation process, the process of constructing a parameter-state-quality comparison table and, when the product defect rate does not meet the preset defect rate threshold, executing adaptive operation of the target robot's operating parameters based on the target robot's process operating parameters, environmental interference data, and the parameter-state-quality comparison table includes: The sample database extracts the target robot's real-time operating parameters, environmental interference data, and corresponding product quality inspection data for several historical collection periods as sample data. Based on the sample data, a parameter-state-quality comparison table is constructed using a Bayesian network. The parameter-state-quality comparison table includes the product defect rate corresponding to different process operating parameters under different environmental interference data conditions. A state space is constructed based on the process operating parameters and environmental interference data of the current acquisition cycle and the parameter-state-quality comparison table. Several target operating parameters are randomly generated. A fitness function is constructed based on the state space and the target operating parameters. Chromosome encoding is performed on the several target operating parameters and the population is initialized to generate an initial population. Constraints are introduced (including equipment safety thresholds and production plans, requiring that the target operating parameters do not exceed the equipment safety thresholds and comply with the production plan, such as the production plan requiring "cutting cycle time ≤ 12 seconds / piece"). Based on the initial population, fitness function, and constraints, the optimal process operating parameters are obtained through a multi-objective genetic algorithm. The optimal process operating parameters are marked as standard process operating parameters and fed back to the lightweight module.

[0023] It should be further explained that, in the specific implementation process, the process of constructing the fitness function based on the state space and the target operating parameters includes: Based on the target operating parameters, environmental interference data, and parameter-state-quality comparison table, obtain the product defect rate corresponding to the target operating parameters. At the same time, based on the target operating parameters and the process operating parameters of the current collection period, obtain the parameter adjustment range (the absolute value of the target operating parameter minus the process operating parameter divided by the process operating parameter). Fitness function: R = α × (1 - product defect rate) + β × (1 - parameter adjustment range), where α and β are preset weight coefficients, which are set by those skilled in the art according to the actual situation; The process of obtaining optimal process operating parameters using a multi-objective genetic algorithm includes: The target operating parameters are encoded as chromosomes, for example, using a string of numbers to represent process operating parameters. A certain number of chromosomes are randomly generated to form an initial population, with each chromosome representing a possible target operating parameter. The fitness function is obtained, and a tournament selection method is used to select chromosomes with higher fitness from the current population as parents. The parent chromosomes are crossbred, exchanging some genes to generate new offspring chromosomes, simulating biological gene exchange to generate new target operating parameters. The offspring chromosomes are mutated, randomly changing some genes to increase population diversity and avoid getting trapped in local optima. The above steps are repeated iteratively until the termination condition is met, such as reaching the maximum number of iterations or the fitness function value no longer significantly improving, at which point the optimal process operating parameters are output.

[0024] It should be further explained that, in the specific implementation process, the process of real-time monitoring of the target robot's process operating parameters and marking the target robot as in a fault state or a critical predicted state based on the monitoring results includes: When the lightweight module receives the standard process operating parameters, it removes the original standard process operating parameters and adjusts the process operating parameters of the target robot to the standard process operating parameters. Then, it compares the process operating parameters of the target robot with the standard process operating parameters in real time to obtain the parameter deviation time sequence. It presets the error range (based on the evaluation standard deviation of the process operating parameters when the target robot is running in normal state), compares the parameter deviation time sequence with the error range, and obtains the cumulative time when the parameter deviation is not within the error range. It presets the upper limit of the cumulative time (three percent of the collection cycle). When the cumulative time is greater than the upper limit of the cumulative time, it marks the target robot as a fault state and obtains the fault type (obtained through manual detection). It marks the process operating parameters, vibration data and environmental interference data of the target robot in the collection cycle as fault data, associates the fault data with the fault type and uploads them to the sample database for storage, and performs cross-plant distributed collaborative scheduling for the target robot. When the cumulative time is less than or equal to the upper limit of the cumulative time, the target robot is marked as a critical prediction state.

[0025] It should be further explained that, in the specific implementation process, the process of constructing a fault prediction model, extracting multi-dimensional features from the vibration data of the target robot in the critical prediction state, obtaining time-domain features, frequency-domain features, and time-frequency-domain features, and inputting them into the fault prediction model, and marking the target robot as a normal state or a fault type warning state according to the fault prediction model includes: Multi-dimensional feature extraction is performed on the vibration data of each component of the target robot in the critical prediction state during the current acquisition cycle to obtain time-domain features, frequency-domain features, and time-frequency-domain features. Simultaneously, auxiliary features are obtained based on the target robot's process operation parameters and environmental interference data during the current acquisition cycle. The multi-dimensional feature extraction includes time-domain features: peak value, kurtosis, and impulse factor (ratio of peak value to RMS value) for each vibration signal segment; a Fast Fourier Transform (FFT) is performed on each time-domain signal segment to convert it into a frequency sequence (frequency resolution 1 Hz), and the frequency-domain features of the frequency sequence are obtained: fundamental frequency amplitude, sideband amplitude, and spectral centroid; time-frequency-domain features are obtained by performing a three-level decomposition of the time-domain signal using wavelet packet transform (WPT) to obtain wavelet packet coefficients for eight frequency bands; auxiliary features include temperature, runtime, and load data (motor torque). A fault prediction model is constructed by extracting time-domain features, frequency-domain features, time-frequency-domain features, and auxiliary features of fault data and normal data from the sample database as training data. The fault prediction model is trained using the training data to obtain the completed fault prediction model. The time-domain features, frequency-domain features, time-frequency-domain features, and auxiliary features of the target robot during the current acquisition period are input into the fault prediction model. The fault prediction model outputs the occurrence probability of each fault type of each component of the target robot. The occurrence probability of each fault type is compared with a preset probability threshold (95%). If the occurrence probability of any fault type is greater than the probability threshold, a warning signal for the fault type is generated, and the target robot is marked as being in a fault type warning state. If the occurrence probability of each fault type is less than or equal to the probability threshold, the target robot is marked as being in a normal state. The process operation parameters, vibration data, and environmental interference data of the target robot during the acquisition period are marked as normal data and uploaded to the sample database for storage.

[0026] It should be further explained that, in the specific implementation process, the fault prediction model is designed with the following structure: Input → Feature Attention → LSTM → Temporal Attention → Output. Input layer: Multimodal data fusion input, with dimensions (time step, feature dimension) = (60, 17): Time step 60: corresponds to 60 consecutive sampling segments (200ms each, 12 seconds in total), covering signal changes during the fault initiation period; Feature dimension 17: includes 14 core vibration spectrum features (peak value, kurtosis, impulse factor, fundamental frequency amplitude, sideband amplitude, spectral centroid, wavelet packet coefficients of 8 frequency bands) + 3 auxiliary features (temperature, runtime, and load data). Feature Attention Layer: The input features are mapped to attention scores through a fully connected layer (normalized to 0-1 using the Sigmoid activation function), and then multiplied element-wise with the original features to highlight the contribution of high-weight features. The specific process is as follows: ,in, The output is a feature weight matrix with a dimension of (60×17). For the input feature matrix, This is the weight matrix. For bias terms, For the Sigmoid function; LSTM layer, capturing short-term and long-term dependencies in time series data (such as the gradual change in gearbox wear from slight to severe); Structure: 3 stacked LSTM layers, each with 128 hidden units, using Dropout (dropoutrate=0.2) to prevent overfitting; Output: The hidden state of the 3rd LSTM layer (dimension 128), containing fault-related information from the past 60 time steps; Temporal attention layer: Highlights key time steps for fault prediction (such as abnormal signal segments in the fault initiation stage); maps the hidden states output by the LSTM layer to temporal attention scores, normalizes them using Softmax, and then weights and sums them with the hidden states to obtain the final temporal feature vector. The specific process is as follows: ,in Let be the hidden state at time step t. Time attention score ( ), The one-dimensional feature vector is obtained by weighting and summing the temporal hidden states output by the LSTM layer through the temporal attention layer; Output layer: Fault type classification: Using the Softmax activation function, output the occurrence probability of 12 types of faults (gearbox wear, bearing pitting, reducer oil leakage, etc.) + 1 type of normal state (a total of 13 types).

[0027] It should be further explained that, in the specific implementation process, the load rate analysis of the target robot in normal state includes: When the target robot is in normal operation, acquire the tag information and corresponding production plans for the target robot and other robots within the factory area. Tag information includes: proficient processes (welding / cutting / assembly, etc.), load capacity (e.g., maximum load 50kg), energy consumption level (low / medium / high), rated operating efficiency, factory area number, and workshop workstation coordinates. Based on the production plan, acquire the target robot's load rate (quantified by the workload set in the production plan (number of processed parts / total workload) and runtime; e.g., completing 20% ​​of the rated workload within 1 hour results in a load rate ≈ 20%) and the corresponding process's task tags (order priority, process, load capacity, process requirements, batch size, time period; e.g., order priority = ...). Grade A, Load Capacity = 50kg, Process = Welding, Process Requirements = Weld Strength ≥ 350MPa, Time Period = 10:30-11:30, Batch Size = 200 pieces) and material labels (including material coordinates, unit distance transportation price (AGV = 0.3 yuan / piece·km), material weight (task-related material labels: such as "5 tons", "2kg / piece × 200 pieces = 400kg", weight billing coefficient (logistics label: billed by weight = 1.0, billed by volume = 0.5), logistics channel congestion coefficient (smooth = 0.05, slightly congested = 0.15, severely congested = 0.3) and material additional cost coefficient (material label: fragile items = 0.2, ordinary materials = 0.05, dangerous goods = 0.5)); The load rate range is divided into sub-ranges of different load levels by selecting threshold points (≤30% is low load, 30%~70% is medium load, and >70% is high load). The load rate of the target robot is determined to be within the sub-range to obtain the load level of the target robot. If the load level of the target robot is high load, a distributed collaborative scheduling operation is performed.

[0028] It should be further explained that, in the specific implementation process, the process of extracting the tag information, load rate, and corresponding task and material tags of each robot within the factory area, and then performing cross-factory distributed collaborative scheduling of high-load target robots based on the tag information, load rate, task tags, and material tags includes: Extract the tag information, load rate, and corresponding process task tags and material tags of each other robot within the factory area. Match the tag information of each other robot with the corresponding process task tags based on the target robot's tag information and the corresponding process task tags, and filter out other robots that meet the preset requirements. The preset requirements include that the other robots are good at the process of the target robot's corresponding process task tag, that the other robots' load capacity is greater than or equal to the load capacity of the target robot's corresponding process task tag, that the other robots' process operation parameters can achieve the process requirements of the target robot's corresponding process task tag, and that the other robots' load rate is low. Based on the tag information, load rate, and corresponding task and material tags of the target robot and other robots that meet the preset requirements, several optimization functions are constructed. Several task handover schemes are randomly generated. Chromosome encoding and population initialization are performed on the several task handover schemes to generate an initial population. The dynamic optimization weights of each optimization function are obtained. Based on the dynamic optimization weights and several optimization functions, an objective function is constructed. Based on the initial population and the objective function, a multi-objective genetic algorithm is used to obtain the target task handover scheme. The task handover process is executed according to the target task handover scheme.

[0029] The task handover process includes: The cloud monitoring platform generates cross-plant scheduling instructions based on the target task handover plan; the process operation parameters (e.g., welding current 200A, welding speed 5mm / s) and process requirements (e.g., weld strength ≥350MPa) of the target robot in Plant A are synchronized to the robot to be handed over in Plant B with a single click, with a synchronization delay ≤50ms; material transfer instructions are also generated (e.g., "from raw material warehouse in Plant A → workstation 2 in Plant B"), specifying the logistics channel and transportation vehicle, estimating the material arrival time, and synchronizing this information to the robot to be handed over in Plant B; task takeover: After receiving the material arrival signal, the robot to be handed over in Plant B automatically starts task execution, and the execution progress is uploaded to the cloud monitoring platform in real time; progress synchronization: The production scheduling center in Plant A views the task execution status in real time through the cloud monitoring platform, ensuring information transparency.

[0030] It should be further explained that, in the specific implementation process, the task handover plan includes the task allocation quantity (number of material allocation items) of the robots to be handed over, the execution time sequence (the start time of the task), and the logistics route segment (the transfer route of materials from the storage location to other robot workstations); the objective function is: ;in, For task completion time, V represents the total cost across plant areas, and V is the load balancing coefficient. , , To dynamically optimize weights, The value is the objective function value.

[0031] The specific process of constructing several optimization functions based on the tag information, load rate, and corresponding task tags of the target robot and other robots that meet the preset requirements is as follows: Where n is the total number of robots participating in the collaboration (target robot + other robots undertaking the same task). Let be the load rate of the i-th robot. To participate in the average robot load rate, The load rate is 70%. It should be further noted that if the target robot is in a faulty state, the load rate of the target robot is zero by default. ,in, For task batch (task label: e.g., "batch=200 pieces"), Pi is the rated operating efficiency of the i-th robot. This is the load correction factor. max() represents taking the maximum value; the higher the load factor L, the better. The smaller the value, the longer the task completion time (which aligns with the industrial reality that higher loads result in lower efficiency). ,in, The unit price for transportation per unit distance (AGV = 0.3 yuan / piece·km) Let represent the inter-factory transport distance of the i-th robot (robot tag information: actual transport distance between workshop workstation coordinates and material coordinates). For material weight, This is a weight-based billing factor. This represents the congestion coefficient of the logistics channel. This is the material additional cost coefficient. Let be the energy consumption level of the i-th robot.

[0032] The process of obtaining a target task handover scheme based on the initial population and objective function using a multi-objective genetic algorithm includes: The task handover scheme is encoded as chromosomes. For example, a string of numbers represents the task allocation, execution time segment, and logistics route segment information of the robot to be handed over. A certain number of chromosomes are randomly generated to form an initial population. Each chromosome represents a possible task handover scheme. The objective function is obtained. Using a tournament selection method, the chromosome with the higher objective function value is selected from the current population as the parent. The parent chromosomes are cross-crossed, exchanging some genes to generate new offspring chromosomes. This simulates the exchange of genes in biological genetics to generate new task handover schemes. The offspring chromosomes are mutated, randomly changing some genes to increase population diversity and avoid getting trapped in local optima. The above steps are repeated iteratively until the termination condition is met, such as reaching the maximum number of iterations or the objective function value no longer significantly improving. The target task handover scheme is then output.

[0033] It should be further explained that, in the specific implementation process, the process of obtaining the dynamic optimization weights of each optimization function includes: Based on the task tags and material tags of the corresponding process of the target robot, the order priority and logistics channel congestion coefficient are obtained. Based on the fault data of the target robot in several historical collection periods in the sample database, the historical failure rate of the target robot (the ratio of the number of historical collection periods to which the fault data belongs to the total number of historical collection periods) is obtained. Based on the tag information and historical failure rate of the target robot, the safety factor of the target robot is obtained. Construct a weight quantization table for several optimization functions. Obtain the weight coefficients corresponding to several optimization functions based on order priority, logistics channel congestion coefficient, safety coefficient, and the weight quantization table. Normalize the weight coefficients corresponding to several optimization functions to generate dynamic optimization weights for several optimization functions.

[0034] The weight quantization table is shown in Table 1 below: Table 1

[0035] It should be further explained that, in the specific implementation process, the maintenance time analysis of the target robot in the fault type warning state, and the determination of whether to execute cross-plant distributed collaborative scheduling based on the analysis results, includes: When a target robot is marked as being in a fault type warning state, the maintenance duration for the corresponding fault type (based on the historical maintenance duration of the fault type) is obtained. The production plans for the target robot and other robots on the production line (continuous production line) for the corresponding processes are obtained. Based on the production plans, the interruptible production window duration for the target robot is obtained (interrupted production window duration = time when the trigger condition of the next process is met - completion time of the current process; for example, after the cutting robot finishes processing a part, it must wait for the assembly robot to pick up the part before starting the next one. The part picking cycle is 5 minutes, so the interruptible production window duration = 5 minutes). It is determined whether the maintenance duration for the fault type is less than the interruptible production window duration. If it is less, maintenance operations are performed on the target robot, and relevant personnel are arranged to perform maintenance and correction. If it is not less, cross-plant distributed collaborative scheduling is performed on the target robot, and maintenance operations are performed on the target robot.

[0036] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A cloud-based robot remote monitoring and data management system, characterized in that, It includes a cloud monitoring platform, data acquisition terminal, sample database, lightweight module, adaptive operating parameter module, fault early warning module and collaborative scheduling module; The lightweight module is used to monitor the product quality inspection data of the corresponding process of the target robot in real time, obtain the product defect rate, and monitor the process operation parameters of the target robot in real time. Based on the monitoring results, the target robot is marked as a fault state or a critical prediction state. The adaptive operation parameter module is used to build a parameter-state-quality lookup table. When the product defect rate does not meet the preset defect rate threshold, the adaptive operation parameter of the target robot is performed based on the target robot's process operation parameters, environmental interference data, and parameter-state-quality lookup table. The fault early warning module is used to build a fault prediction model, extract multi-dimensional features from the vibration data of the target robot in the critical prediction state, obtain time domain features, frequency domain features and time-frequency domain features and input them into the fault prediction model, and mark the target robot as normal state or fault type early warning state according to the fault prediction model. The collaborative scheduling module is used to analyze the load rate of target robots in normal state, extract the tag information, load rate, and corresponding task tags and material tags of each robot within the factory area, perform cross-factory distributed collaborative scheduling of target robots in high load or fault state based on tag information, load rate, task tags and material tags, and analyze the maintenance time of target robots in fault type warning state, and determine whether to perform cross-factory distributed collaborative scheduling based on the analysis results.

2. The cloud-based robot remote monitoring and data management system according to claim 1, characterized in that, Data acquisition terminals and lightweight modules are deployed on the target robot and the product production line. The cloud monitoring platform communicates with the data acquisition terminals and lightweight modules within the pre-defined factory area. The data acquisition terminals are used to transmit the collected data to the lightweight modules and the cloud monitoring platform. The cloud monitoring platform is equipped with a sample database, an adaptive operating parameter module, a fault early warning module, and a collaborative scheduling module. The sample database, the adaptive operating parameter module, the fault early warning module, and the collaborative scheduling module communicate with each other.

3. The cloud-based robot remote monitoring and data management system according to claim 2, characterized in that, The process of obtaining product defect rates includes: Acquire the process operation parameters, vibration data, environmental interference data, and product quality inspection data of the corresponding process of the target robot in the current data collection cycle; Set a time window, perform statistical analysis on the product quality inspection data of the target robot within the time window, obtain the product defect rate. If the product defect rate within the time window is greater than the preset defect rate threshold, then perform adaptive operation of the running parameters.

4. The cloud-based robot remote monitoring and data management system according to claim 3, characterized in that, The process of performing adaptive operation on the target robot's operating parameters includes: The sample database is used to extract the real-time operating parameters, environmental interference data and corresponding product quality inspection data of the target robot in several historical collection periods as sample data. A parameter-state-quality comparison table is constructed based on the sample data. The parameter-state-quality comparison table includes the product defect rate corresponding to different process operating parameters under different environmental interference data conditions. A state space is constructed based on the process operating parameters and environmental interference data of the current acquisition cycle and the parameter-state-quality comparison table. Several target operating parameters are randomly generated. A fitness function is constructed based on the state space and the target operating parameters. Chromosome encoding is performed on the several target operating parameters and the population is initialized to generate an initial population. Constraints are introduced. Based on the initial population, fitness function and constraints, the optimal process operating parameters are obtained through a multi-objective genetic algorithm. The optimal process operating parameters are marked as standard process operating parameters and fed back to the lightweight module.

5. The cloud-based robot remote monitoring and data management system according to claim 4, characterized in that, The process of real-time monitoring of the target robot's process operating parameters includes: When the lightweight module receives the standard process operating parameters, it removes the original standard process operating parameters and adjusts the process operating parameters of the target robot to the standard process operating parameters. Then, it compares the process operating parameters of the target robot with the standard process operating parameters in real time to obtain the parameter deviation time series. It presets the error range, compares the parameter deviation time series with the error range to obtain the cumulative time when the parameter deviation is not within the error range, and presets the cumulative time upper limit. When the cumulative time is greater than the cumulative time upper limit, it marks the target robot as a fault state, obtains the fault type, marks the process operating parameters, vibration data and environmental interference data of the target robot in the collection period as fault data, associates the fault data with the fault type and uploads them to the sample database, and performs cross-plant distributed collaborative scheduling for the target robot. When the cumulative time is less than or equal to the upper limit of the cumulative time, the target robot is marked as a critical prediction state.

6. The cloud-based robot remote monitoring and data management system according to claim 5, characterized in that, The process of constructing a fault prediction model, extracting multi-dimensional features from the vibration data of the target robot in the critical prediction state, obtaining time-domain features, frequency-domain features, and time-frequency-domain features, and inputting them into the fault prediction model, and marking the target robot as a normal state or a fault type warning state according to the fault prediction model includes: Multi-dimensional feature extraction is performed on the vibration data of each component of the target robot in the current acquisition cycle to obtain time-domain features, frequency-domain features, and time-frequency-domain features. At the same time, auxiliary features are obtained based on the process operation parameters and environmental interference data of the target robot in the current acquisition cycle. A fault prediction model is constructed by extracting time-domain features, frequency-domain features, time-frequency-domain features, and auxiliary features of fault data and normal data from the sample database as training data. The fault prediction model is trained using the training data to obtain the completed fault prediction model. The time-domain features, frequency-domain features, time-frequency-domain features, and auxiliary features of the target robot during the current acquisition period are input into the fault prediction model. The fault prediction model outputs the occurrence probability of each fault type for each component of the target robot. The occurrence probability of each fault type is compared with a preset probability threshold. If the occurrence probability of any fault type is greater than the probability threshold, a warning signal for the fault type is generated, and the target robot is marked as being in a fault type warning state. If the occurrence probability of each fault type is less than or equal to the probability threshold, the target robot is marked as being in a normal state. The process operation parameters, vibration data, and environmental interference data of the target robot during the acquisition period are marked as normal data and uploaded to the sample database.

7. The cloud-based robot remote monitoring and data management system according to claim 6, characterized in that, The process of performing load rate analysis on a target robot in a normal state includes: When the target robot is in normal condition, obtain the tag information of the target robot and other robots within the factory area, as well as the production plan of the corresponding process. Based on the production plan, obtain the load rate of the target robot and the task tag and material tag of the corresponding process. The load rate range is divided into sub-ranges with different load levels by selecting a threshold point. The load rate of the target robot is determined to be within the sub-range to obtain the load level of the target robot. If the load level of the target robot is high load, a distributed collaborative scheduling operation is performed.

8. The cloud-based robot remote monitoring and data management system according to claim 7, characterized in that, The process of cross-plant distributed collaborative scheduling of high-load target robots based on tag information, load rate, task tags, and material tags includes: Extract the tag information, load rate, and corresponding process task and material tags of each other robot within the factory area. Match the tag information of each other robot with the corresponding process task tags based on the target robot's tag information and the corresponding process task tags to filter out other robots that meet the preset requirements. Based on the tag information, load rate, and corresponding task and material tags of the target robot and other robots that meet the preset requirements, several optimization functions are constructed. Several task handover schemes are randomly generated. Chromosome encoding and population initialization are performed on the several task handover schemes to generate an initial population. The dynamic optimization weights of each optimization function are obtained. Based on the dynamic optimization weights and several optimization functions, an objective function is constructed. Based on the initial population and the objective function, a multi-objective genetic algorithm is used to obtain the target task handover scheme. The task handover process is executed according to the target task handover scheme.

9. The cloud-based robot remote monitoring and data management system according to claim 8, characterized in that, The process of obtaining the dynamic optimization weights of each optimization function includes: The order priority and logistics channel congestion coefficient are obtained based on the task tags and material tags of the corresponding process of the target robot. The historical failure rate of the target robot is obtained based on the failure data of the target robot in several historical collection periods in the sample database. The safety coefficient of the target robot is obtained based on the tag information and historical failure rate of the target robot. Construct a weight quantization table for several optimization functions. Obtain the weight coefficients corresponding to several optimization functions based on order priority, logistics channel congestion coefficient, safety coefficient, and the weight quantization table. Normalize the weight coefficients corresponding to several optimization functions to generate dynamic optimization weights for several optimization functions.

10. The cloud-based robot remote monitoring and data management system according to claim 9, characterized in that, The process of analyzing maintenance time for target robots in fault type warning states and determining whether to execute cross-plant distributed collaborative scheduling based on the analysis results includes: When the target robot is marked as a fault type warning state, the maintenance duration of the fault type is obtained, the production plan of the target robot and the corresponding process of each other robot on the production line is obtained, the production interruption window duration of the target robot is obtained based on the production plan, and it is determined whether the maintenance duration of the fault type is less than the production interruption window duration. If it is less, the maintenance operation is performed on the target robot. If it is not less, cross-plant distributed collaborative scheduling is performed on the target robot and the maintenance operation is performed on the target robot.

Images