Material handling robot control system and method
By working collaboratively across the device, environment, behavior, and cloud, and combining multi-source data analysis, the problems of delayed response and data silos in traditional operation and maintenance models have been solved, enabling intelligent and refined operation and maintenance of handling robots, and improving production efficiency and equipment stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING SHIJIAN INTELLIGENT TECHNOLOGY CO LTD
- Filing Date
- 2025-12-02
- Publication Date
- 2026-06-02
AI Technical Summary
Traditional operation and maintenance of handling robots relies on manual inspection and single data analysis, resulting in delayed response, high operation and maintenance costs, and serious data silos. This makes it impossible to achieve accurate early warning and efficient fault location, affecting production efficiency and equipment stability.
By employing collaborative work across device, environment, behavior, and edge terminals, risk coefficients are generated through data collection and weighted self-learning models. Combined with cloud-based machine learning models, operation and maintenance time is predicted, enabling the fusion analysis of multi-source data and accurate prediction of operation and maintenance time.
It enables accurate prediction and proactive management of robot maintenance time, reduces unplanned downtime, lowers maintenance costs, improves maintenance efficiency and equipment stability, and achieves intelligent and refined management.
Smart Images
Figure CN121468529B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of operation and maintenance technology for material handling robots, specifically a control system and method for a material handling robot. Background Technology
[0002] Material handling robots are playing an increasingly important role on production lines. However, traditional operation and maintenance models for material handling robots mainly rely on regular inspections and troubleshooting by on-site engineers. In recent years, although cloud computing technology has been introduced into the industrial field to achieve remote monitoring and data analysis, it currently relies solely on data processing and analysis, resulting in limited utilization methods.
[0003] Patent application CN114647645A discloses an industrial mobile robot operation and maintenance system, comprising: an embedded data acquisition gateway for collecting data information from the industrial mobile robot; an operation and maintenance platform for collecting and storing the data information from the industrial mobile robot, analyzing and issuing early warnings based on the data information, generating corresponding fault information, and providing corresponding fault handling solutions based on the fault information; and at least one client for receiving fault handling notifications, accessing the data information from the industrial mobile robot, querying fault handling history, and accessing operation and maintenance knowledge. This invention enables rapid fault diagnosis, shortens fault repair time, standardizes the entire fault handling process, and improves the quality of industrial mobile robot operation and maintenance services.
[0004] As mentioned in the application, the operation and maintenance of handling robots currently generally adopts two traditional methods: one is to passively judge based solely on operational status data, and the other is to rely on engineers for regular inspections and troubleshooting. Both methods have significant shortcomings in practical applications. On the one hand, the response lag is prominent—from the occurrence of a fault to its resolution, the entire process relies on manual intervention, which not only fails to achieve early warning but also takes a long time for fault location and spare parts allocation, leading to extended unplanned downtime. On the other hand, the operation and maintenance costs remain high, not only reflected in the manpower investment required for dedicated inspections and expert dispatch but also in the lack of precision in maintenance strategies: over-maintenance leads to resource waste, while under-maintenance may cause secondary damage, further increasing repair costs. A deeper problem lies in data silos; data is difficult to exchange between the device, behavior, environment, and edge devices, and production, operation and maintenance, and management systems are not interconnected, making data integration and analysis impossible. A large amount of status information and experiential knowledge is stored offline, making it difficult to systematically support decision-making and limiting further improvements in operation and maintenance efficiency. Summary of the Invention
[0005] To address the above problems, the present invention provides a material handling robot control system and method.
[0006] This invention adopts the following technical solution: a material handling robot control system, comprising:
[0007] On the device side, the device side integrates a first data acquisition module and a first risk analysis module;
[0008] The first data acquisition module is used to collect the operating status data of the handling robot;
[0009] The first risk analysis module is used to acquire the operating status data of the handling robot, assign weights to the operating status data of the handling robot based on a weighted self-learning model, and determine the operating risk coefficient of the handling robot.
[0010] On the environmental end, the environmental end integrates a second data acquisition module and a second risk analysis module;
[0011] The second data acquisition module is used to collect environmental status data at the production site of the handling robot.
[0012] The second risk analysis module is used to acquire environmental status data of the production site of the handling robot, assign weights to the environmental status data of the production site of the handling robot based on a weighted self-learning model, and determine the environmental risk coefficient of the handling robot.
[0013] The behavior terminal includes a load feature acquisition module and a task type acquisition module;
[0014] The load characteristic acquisition module is used to acquire the load characteristic parameters of the current operation, including the weight of the load and the rate of change of the load.
[0015] The task type acquisition module is used to acquire the current task type of the current handling robot. The task type includes handling, palletizing and transportation, and is encoded using one-hot encoding.
[0016] The edge terminal includes a data acquisition module and a fault type prediction module;
[0017] The data acquisition module is used to acquire the operational risk coefficient output by the first risk analysis module, the environmental risk coefficient output by the second risk analysis module, and the load characteristic parameters.
[0018] The fault type prediction module is used to predict the type of fault the handling robot will experience next.
[0019] The cloud service center integrates an operation and maintenance time prediction module and an operation and maintenance time optimization module;
[0020] The operation and maintenance time prediction module takes the acquired operation risk coefficient, environmental risk coefficient, load characteristic parameters, and next failure type, and inputs them into the trained machine learning model to predict the operation and maintenance time.
[0021] As a further description of the above technical solution: the operating status data includes cumulative operating time, robot joint movement frequency, motor operating temperature, and joint limit movement amplitude frequency;
[0022] The method for obtaining the operational risk coefficient includes:
[0023] Set corresponding weight coefficients for the running status data;
[0024] The operational status data are weighted and summed according to their respective weights to obtain the operational risk coefficient, which comprehensively represents the overall operational risk level of the robot.
[0025] As a further description of the above technical solution: the method for obtaining the joint motion frequency includes: recording the angle change of the joint during operation through a joint angle sensor, obtaining the joint motion frequency of each joint by counting the number of angle changes per unit time, establishing a set, and obtaining the average value.
[0026] As a further description of the above technical solution: the method for obtaining the frequency of the joint's limit motion amplitude includes: setting a threshold for joint angle change, recording the angle change of the joint during operation through a joint angle sensor, and triggering a counter to record once when the joint angle change exceeds the set threshold, and statistically analyzing the counting results per unit time.
[0027] As a further description of the above technical solution: the environmental status data includes workshop noise data, workshop voltage fluctuation data, and vibration velocity data of the handling robot;
[0028] The method for obtaining the environmental risk coefficient includes:
[0029] Set corresponding weight coefficients for the environmental status data;
[0030] The environmental risk coefficient, which comprehensively represents the environmental status of the production site, is obtained by weighting and summing the above environmental status data according to their respective weights.
[0031] As a further description of the above technical solution: the training method of the machine learning model for predicting operation and maintenance time includes:
[0032] The collected operational risk coefficient, environmental risk coefficient, load characteristic parameters, and next fault type are converted into a corresponding set of feature vectors;
[0033] Each set of feature vectors is used as input to the machine learning model. The machine learning model outputs the operational risk coefficient, environmental risk coefficient, load characteristic parameters, and the maintenance time corresponding to the next fault type for each set of operational risk coefficients, environmental risk coefficients, load characteristic parameters, and the actual maintenance time corresponding to the next fault type for each set of operational risk coefficients, environmental risk coefficients, load characteristic parameters, and the next fault type. The training objective is to minimize the loss function value of the machine learning model. Training stops when the loss function value of the machine learning model is less than or equal to the preset target loss value.
[0034] The machine learning model is one of the following: support vector machine regression, random forest regression, or neural network regression model.
[0035] As a further description of the above technical solution: the operation and maintenance time optimization module includes:
[0036] The maintenance record acquisition module retrieves the occurrence time of historical repairable faults of the handling robot, the interval T between two consecutive faults, and establishes a set M, M={T1, T2, ..., T...}. n};
[0037] The average value calculation module obtains the average value over the interval time.
[0038] The analysis and control module compares the predicted maintenance time with the average of the interval time to obtain the time decay value. It then compares and analyzes the time decay value with the preset safe time decay threshold to generate maintenance time compensation instructions.
[0039] As a further description of the above technical solution:
[0040] The expression for the time decay value is:
[0041] ;
[0042] In the formula, This is the time decay value. For the predicted maintenance time, This represents the average value over the intervals.
[0043] The maintenance time compensation instructions include a first shortening instruction, a second shortening instruction, and a third shortening instruction, and the shortening time of the first shortening instruction, the second shortening instruction, and the third shortening instruction increases sequentially;
[0044] Methods for generating maintenance time compensation instructions include:
[0045] Preset time decay thresholds SB1, SB2, and SB3, where 0 < SB1 < SB2 < SB3;
[0046] when If ≤SB1, then no maintenance time compensation instruction will be generated;
[0047] When SB1 < If ≤SB2, then the first shortening instruction is generated;
[0048] When SB2 < If ≤SB3, then a second shortening instruction is generated;
[0049] when When >SB3, a third shortening instruction is generated.
[0050] As a further description of the above technical solution:
[0051] The method for predicting the next failure type of the handling robot includes:
[0052] The collected operational risk coefficient, environmental risk coefficient, and load characteristic parameters are input into a pre-built fault type prediction model, which outputs the next fault type of the handling robot.
[0053] The fault types include motor faults, sensor malfunctions, control system errors, driver overloads, communication malfunctions, mechanical structure wear, wheel set faults, and load offset malfunctions.
[0054] The training method for the fault type prediction model includes:
[0055] Q sets of training data are collected in advance, where Q is a positive integer greater than 0. The training data includes the operating risk coefficient, the environmental risk coefficient, and the load characteristic parameters, as well as the fault type labels corresponding to the operating risk coefficient, the environmental risk coefficient, and the load characteristic parameters.
[0056] A deep learning model was selected as the fault type prediction model. The model was trained using training data. The operational risk coefficient, environmental risk coefficient, and load characteristic parameters were used as inputs to the model, and the fault type label was used as the output. The stochastic gradient descent method was used, and the weights and biases of the model were adjusted through backpropagation to minimize the error between the prediction results and the actual results. A loss function, which is the mean squared error, was set. When the loss function value converged, the training of the fault type prediction model was stopped, and the fault type prediction model corresponding to the convergence of the loss function value was taken as the trained fault type prediction model.
[0057] A method for controlling a material handling robot, comprising:
[0058] Collect operational status data of the handling robot;
[0059] The operation status data of the handling robot is obtained, and the operation status data of the handling robot is weighted and assigned based on the weight self-learning model to determine the operation risk coefficient of the handling robot.
[0060] Collect environmental status data from the production site of the handling robot;
[0061] The environmental status data of the production site of the handling robot is obtained, and the environmental status data of the production site of the handling robot is weighted and assigned based on a weighted self-learning model to determine the environmental risk coefficient of the handling robot.
[0062] Collect the load characteristic parameters of the current operation, including the weight of the load and the rate of change of the load;
[0063] Obtain the current job type of the current handling robot, which includes handling, palletizing and transportation, and encode it using one-hot encoding;
[0064] Predict the type of failure that will occur next for the handling robot;
[0065] The acquired operational risk coefficient, environmental risk coefficient, load characteristic parameters, and next failure type are input into the trained machine learning model for predicting operation and maintenance time to predict the operation and maintenance time.
[0066] Obtain the occurrence time of repairable faults in the history of the handling robot, obtain the time interval T between two adjacent faults, establish a set, and obtain the average value of the time interval;
[0067] The predicted maintenance time is compared with the average of the interval time to obtain the time decay value. The time decay value is then compared and analyzed with the preset safe time decay threshold to generate maintenance time compensation instructions.
[0068] Beneficial effects:
[0069] This invention provides a material handling robot control system that achieves accurate prediction and proactive management of robot maintenance time through collaborative work of the equipment end, environment end, behavior end, edge end, and cloud end. The system collects operating status data at the equipment end and generates an operating risk coefficient based on a weighted self-learning model; it collects production site environmental data at the edge end and generates an environmental risk coefficient; and it collects load characteristics and operation types at the behavior end and predicts the next possible fault type. The above multi-source data is input into a maintenance time prediction model trained in the cloud to achieve accurate prediction of the robot's next maintenance time. At the same time, it calculates the time decay value by combining historical fault intervals and compares it with a safety threshold to generate maintenance compensation instructions, thereby achieving dynamic adjustment of the maintenance cycle.
[0070] Through the above methods, this invention overcomes the shortcomings of traditional operation and maintenance, such as delayed response, data silos, and inaccurate maintenance strategies. It realizes intelligent operation and maintenance management based on big data and machine learning. Relying on the "cloud-edge-device" collaborative architecture, the system can integrate historical and real-time data, scientifically predict operation and maintenance time, and formulate preventive maintenance plans. This improves the stability of robot operation, reduces unplanned downtime, lowers operation and maintenance costs, and realizes intelligent, refined, and interpretable management of industrial robot operation and maintenance. Attached Figure Description
[0071] The present invention will be further explained below with reference to the accompanying drawings and embodiments:
[0072] Figure 1 This is a module connection diagram of a material handling robot control system provided in Embodiment 1 of the present invention;
[0073] Figure 2 This is a module connection diagram of a material handling robot control system provided in Embodiment 2 of the present invention;
[0074] Figure 3 This is a flowchart of a method for obtaining the frequency of the joint's limit motion amplitude provided in Embodiment 1 of the present invention;
[0075] Figure 4 This is a flowchart of a material handling robot control method provided in Embodiment 3 of the present invention. Detailed Implementation
[0076] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.
[0077] Example 1
[0078] Please see Figure 1 and Figure 3 This invention provides a technical solution: a material handling robot control system, comprising:
[0079] On the device side, the device side integrates a first data acquisition module and a first risk analysis module;
[0080] The first data acquisition module is used to collect the operating status data of the handling robot;
[0081] The operational status data includes cumulative operating time, robot joint movement frequency, motor operating temperature, and joint limit movement amplitude frequency.
[0082] It should be noted that the higher the frequency of joint movement, that is, the more frequent the movement of each joint, the more likely it is to wear and tear, and thus require a shorter maintenance cycle.
[0083] Methods for obtaining joint motion frequency include: recording the angle changes of each joint during operation using joint angle sensors, and obtaining the frequency by counting the number of angle changes per unit time;
[0084] Methods for obtaining joint motion frequency also include: most handling robot controllers have built-in joint motion counting functions, which can directly read joint motion frequency data. This data is usually automatically recorded and counted by the controller when executing motion commands.
[0085] The longer the cumulative operating time, the shorter the maintenance cycle is required;
[0086] The higher the motor's operating temperature, the faster the motor and related electronic components will age. Therefore, the higher the motor's operating temperature, the shorter the maintenance cycle is required.
[0087] The method for obtaining the joint limit motion amplitude frequency includes: setting a threshold for joint angle change; when the joint angle change exceeds the set threshold, triggering a counter to record once; and then statistically analyzing the counting results per unit time. The higher the joint limit motion amplitude frequency, the more wear will increase, requiring more frequent inspections and maintenance.
[0088] The first risk analysis module is used to acquire the operating status data of the handling robot, assign weights to the operating status data of the handling robot based on a weighted self-learning model, and determine the operating risk coefficient of the handling robot.
[0089] It should be noted that the weight self-learning model is an adaptive neural network model, which is an existing technology. Its self-learning process includes:
[0090] Data input phase: Input the collected multidimensional operational status data into the model;
[0091] Model training phase: The model automatically calculates and updates the weights of each parameter based on the correlation between different parameters and operational risks in historical data samples;
[0092] Weighting Output Stage: The final output is a set of weight values that reflect the importance of each operating parameter, which are used for subsequent risk assessment or health prediction calculations.
[0093] The method for obtaining the operational risk coefficient includes:
[0094] Set corresponding weight coefficients for the running status data;
[0095] The above operational status data are weighted and summed according to their weights to obtain the operational risk coefficient, which comprehensively represents the overall operational risk level of the robot.
[0096] Optionally, the expression for the operational risk coefficient is:
[0097] ;
[0098] In the formula, To account for the operational risk factor, For cumulative running time, This refers to the frequency of joint movements in the robot. The operating temperature of the motor. The frequency of the joint's limit range of motion. , , and These are the weighting coefficients. , , and Greater than zero.
[0099] It should be noted that the weighting coefficient is a specific value obtained by quantifying each data point to facilitate subsequent comparisons.
[0100] On the environmental end, the environmental end integrates a second data acquisition module and a second risk analysis module;
[0101] The second data acquisition module is used to collect environmental status data at the production site of the handling robot.
[0102] The environmental status data includes workshop noise data, workshop voltage fluctuation data, and vibration velocity data of the handling robot;
[0103] The second risk analysis module is used to acquire environmental status data of the production site of the handling robot, assign weights to the environmental status data of the production site of the handling robot based on a weighted self-learning model, and determine the environmental risk coefficient of the handling robot.
[0104] The method for obtaining the environmental risk coefficient includes:
[0105] Set corresponding weight coefficients for the environmental status data;
[0106] The environmental risk coefficient, which comprehensively represents the environmental status of the production site, is obtained by weighting and summing the above environmental status data according to their respective weights.
[0107] Optionally, the expression for the environmental risk coefficient is:
[0108] ;
[0109] In the formula, For environmental risk coefficient, For workshop noise data, For workshop voltage fluctuation data, For the vibration velocity data of the handling robot, , and These are the weighting coefficients. , and Greater than zero.
[0110] It should be noted that the weighting coefficient is a specific value obtained by quantifying each data point to facilitate subsequent comparisons.
[0111] The behavior terminal includes a load feature acquisition module and a task type acquisition module;
[0112] The load characteristic acquisition module is used to acquire the load characteristic parameters of the current operation, including the weight of the load and the rate of change of the load.
[0113] The load weight is obtained by collecting data from a weighing sensor installed on the robot's carrying platform or at the end of the robotic arm. The sensor measures the weight of the material being transported by the robot in real time.
[0114] The load change rate can be obtained by calculating the weight measurement values at continuous time intervals; the weight change value is obtained by dividing the change time.
[0115] The task type acquisition module is used to acquire the current task type of the current handling robot. The task type includes handling, palletizing and transportation, and is encoded using one-hot encoding.
[0116] The method for obtaining the current job type is as follows: call the job plan scheduling information of the robot control system (PLC or host computer system), read the current job type, and perform numerical processing on the obtained job type using one-hot encoding, where the handling code is [1,0,0], the palletizing code is [0,1,0], and the transportation code is [0,0,1], so as to facilitate subsequent processing by the machine learning model.
[0117] The edge terminal includes a data acquisition module and a fault type prediction module;
[0118] The data acquisition module is used to acquire the operational risk coefficient output by the first risk analysis module, the environmental risk coefficient output by the second risk analysis module, and the load characteristic parameters.
[0119] The fault type prediction module is used to predict the type of fault the handling robot will experience next.
[0120] The method for predicting the next failure type of the handling robot includes:
[0121] The collected operational risk coefficient, environmental risk coefficient, and load characteristic parameters are input into a pre-built fault type prediction model, which outputs the next fault type of the handling robot.
[0122] The fault types include motor faults, sensor malfunctions, control system errors, driver overloads, communication malfunctions, mechanical structure wear, wheel set faults, and load offset malfunctions.
[0123] The training method for the fault type prediction model includes:
[0124] Q sets of training data are collected in advance, where Q is a positive integer greater than 0. The training data includes the operating risk coefficient, the environmental risk coefficient, and the load characteristic parameters, as well as the fault type labels corresponding to the operating risk coefficient, the environmental risk coefficient, and the load characteristic parameters.
[0125] It should be noted that the following can be tagged at once: motor failure, sensor malfunction, control system error, drive overload, communication malfunction, mechanical structure wear, wheel set failure, and load offset malfunction, respectively set to 01, 02, 03, 04, 05, 06, 07, and 08.
[0126] A deep learning model was selected as the fault type prediction model. The model was trained using training data. The operational risk coefficient, environmental risk coefficient, and load characteristic parameters were used as inputs to the model, and the fault type label was used as the output. The stochastic gradient descent method was used, and the weights and biases of the model were adjusted through backpropagation to minimize the error between the prediction results and the actual results. A loss function, which is the mean squared error, was set. When the loss function value converged, the training of the fault type prediction model was stopped, and the fault type prediction model corresponding to the convergence of the loss function value was taken as the trained fault type prediction model.
[0127] The cloud service center integrates an operation and maintenance time prediction module and an operation and maintenance time optimization module;
[0128] The operation and maintenance time prediction module takes the acquired operation risk coefficient, environmental risk coefficient, load characteristic parameters, and next failure type, and inputs them into the trained machine learning model to predict the operation and maintenance time.
[0129] The training method for the machine learning model that predicts maintenance time includes:
[0130] The collected operational risk coefficient, environmental risk coefficient, load characteristic parameters, and next fault type are converted into a corresponding set of feature vectors;
[0131] Each set of feature vectors is used as input to the machine learning model. The machine learning model outputs the operational risk coefficient, environmental risk coefficient, load characteristic parameters, and the maintenance time corresponding to the next fault type for each set of operational risk coefficients, environmental risk coefficients, load characteristic parameters, and the actual maintenance time corresponding to the next fault type for each set of operational risk coefficients, environmental risk coefficients, load characteristic parameters, and the next fault type. The training objective is to minimize the loss function value of the machine learning model. Training stops when the loss function value of the machine learning model is less than or equal to the preset target loss value.
[0132] The machine learning model is one of the following: support vector machine regression, random forest regression, or neural network regression model.
[0133] The loss function value of the machine learning model is the mean squared error.
[0134] By using the loss function The model is trained with minimization as the objective, which enables the machine learning model to better fit the data, thereby improving the model's performance and accuracy.
[0135] In the loss function Here, s represents the loss function value of the machine learning model, q represents the feature vector group number, and s represents the number of feature vector groups. For the first Maintenance time predicted by group feature vectors This represents the actual maintenance time corresponding to the s-th group of feature vectors.
[0136] Other model parameters of the machine learning model, such as the target loss value, optimization algorithm, ratio of training set to test set to validation set, and optimization of the loss function, are all obtained through actual engineering implementation and continuous experimental tuning. In this embodiment, firstly, at the device end, the system collects the operating status data of the handling robot, and assigns weights to the operating status data based on the weighted self-learning model to determine the operating risk coefficient of the handling robot, thereby quantifying the operating health status of the equipment;
[0137] Secondly, at the edge, the system collects environmental status data of the production site of the handling robot, and also assigns weights based on the weight self-learning model to generate an environmental risk coefficient, which is used to reflect the potential impact of the external environment on the safety of robot operation.
[0138] Furthermore, the behavior end collects the operational characteristics of the handling robot, including load weight, load change rate, and current operation type (handling, palletizing, transportation), and realizes the numerical representation of operation type through one-hot encoding; at the same time, the edge end predicts the next possible fault type through the fault data collection and prediction module.
[0139] Then, the operational risk coefficient of the equipment, the environmental risk coefficient of the edge, the load characteristics and operation type of the behavior, and the predicted fault type are taken as input and sent to the maintenance time prediction model trained in the cloud service center to achieve accurate prediction of the next maintenance time of the handling robot.
[0140] In addition, the system also combines the occurrence time of historical repairable faults to calculate the average interval between two adjacent faults, and generates a time decay value by comparing it with the predicted maintenance time. Then, it compares it with the preset safe time decay threshold to generate maintenance time compensation instructions, thereby realizing dynamic adjustment of maintenance cycle and safe redundancy control.
[0141] Through the collection, analysis, and prediction of the aforementioned multi-source data, this invention realizes an operation and maintenance management strategy based on big data and machine learning, overcoming problems such as delayed response, data silos, and inaccurate maintenance strategies in traditional operation and maintenance models. At the same time, relying on the "cloud-edge-device" collaborative architecture, the system can combine historical and real-time data to scientifically predict operation and maintenance time and formulate preventive maintenance plans, thereby improving the operational stability and equipment utilization of handling robots, reducing unplanned downtime, lowering operation and maintenance costs, and realizing intelligent and refined management of industrial robot operation and maintenance.
[0142] Example 2
[0143] Please see Figures 1-3 This embodiment adds an operation and maintenance time optimization module based on the above embodiments. That is, based on the above embodiments, the cloud service center predicts the operation and maintenance time of the handling robot according to historical data and real-time data, and then optimizes the operation and maintenance time by combining the historical repairable failure time of the handling robot, thereby reducing the occurrence of repairable failures and improving the stability of the process robot operation.
[0144] The operation and maintenance time optimization module includes:
[0145] The maintenance record acquisition module retrieves the occurrence time of historical repairable faults of the handling robot, the interval T between two consecutive faults, and establishes a set M, M={T1, T2, ..., T...}. n};
[0146] It should be noted that a repairable fault is one that can be fixed, meaning that the probability of such a fault occurring can be avoided, reduced, or lowered through technical maintenance.
[0147] The average value calculation module obtains the average value over the interval time. , ;
[0148] The analysis and control module compares the predicted maintenance time with the average of the interval time to obtain the time decay value. It then compares and analyzes the time decay value with the preset safe time decay threshold to generate maintenance time compensation instructions.
[0149] The expression for the time decay value is:
[0150] ;
[0151] In the formula, This is the time decay value. For the predicted maintenance time, This represents the average value over the interval.
[0152] The maintenance time compensation instructions include a first shortening instruction, a second shortening instruction, and a third shortening instruction, and the shortening time of the first shortening instruction, the second shortening instruction, and the third shortening instruction increases sequentially;
[0153] Methods for generating maintenance time compensation instructions include:
[0154] Preset time decay thresholds SB1, SB2, and SB3, where 0 < SB1 < SB2 < SB3;
[0155] when If ≤SB1, then no maintenance time compensation instruction will be generated;
[0156] When SB1 < If ≤SB2, then the first shortening instruction is generated;
[0157] When SB2 < If ≤SB3, then a second shortening instruction is generated;
[0158] when When >SB3, a third shortening instruction is generated.
[0159] In this embodiment, based on the prediction of the operation and maintenance time of the handling robot by the cloud service center according to historical and real-time data, and combined with the occurrence time of historical repairable faults of the handling robot, the interval between two adjacent faults is taken to optimize the operation and maintenance time, thereby reducing the occurrence of repairable faults and improving the stability of the process robot operation.
[0160] Example 3
[0161] Please see Figure 4 This invention provides a control method for a material handling robot, comprising:
[0162] Collect operational status data of the handling robot;
[0163] The operation status data of the handling robot is obtained, and the operation status data of the handling robot is weighted and assigned based on the weight self-learning model to determine the operation risk coefficient of the handling robot.
[0164] Collect environmental status data from the production site of the handling robot;
[0165] The environmental status data of the production site of the handling robot is obtained, and the environmental status data of the production site of the handling robot is weighted and assigned based on a weighted self-learning model to determine the environmental risk coefficient of the handling robot.
[0166] Collect the load characteristic parameters of the current operation, including the weight of the load and the rate of change of the load;
[0167] Obtain the current job type of the current handling robot, which includes handling, palletizing and transportation, and encode it using one-hot encoding;
[0168] Predict the type of failure that will occur next for the handling robot;
[0169] The acquired operational risk coefficient, environmental risk coefficient, load characteristic parameters, and next failure type are input into the trained machine learning model for predicting operation and maintenance time to predict the operation and maintenance time.
[0170] Obtain the occurrence time of repairable faults in the history of the handling robot, obtain the time interval T between two adjacent faults, establish a set, and obtain the average value of the time interval;
[0171] The predicted maintenance time is compared with the average of the interval time to obtain the time decay value. The time decay value is then compared and analyzed with the preset safe time decay threshold to generate maintenance time compensation instructions.
[0172] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.
Claims
1. A material handling robot control system, characterized in that, include: On the device side, the device side integrates a first data acquisition module and a first risk analysis module; The first data acquisition module is used to collect the operating status data of the handling robot; The first risk analysis module is used to acquire the operating status data of the handling robot, assign weights to the operating status data of the handling robot based on a weighted self-learning model, and determine the operating risk coefficient of the handling robot. On the environmental end, the environmental end integrates a second data acquisition module and a second risk analysis module; The second data acquisition module is used to collect environmental status data at the production site of the handling robot. The second risk analysis module is used to acquire environmental status data of the production site of the handling robot, assign weights to the environmental status data of the production site of the handling robot based on a weighted self-learning model, and determine the environmental risk coefficient of the handling robot. The behavior terminal includes a load feature acquisition module and a task type acquisition module; The load characteristic acquisition module is used to acquire the load characteristic parameters of the current operation, including the weight of the load and the rate of change of the load. The task type acquisition module is used to acquire the current task type of the current handling robot. The task type includes handling, palletizing and transportation, and is encoded using one-hot encoding. The edge terminal includes a data acquisition module and a fault type prediction module; The data acquisition module is used to acquire the operational risk coefficient output by the first risk analysis module, the environmental risk coefficient output by the second risk analysis module, and the load characteristic parameters. The fault type prediction module is used to predict the type of fault the handling robot will experience next. The cloud service center integrates an operation and maintenance time prediction module and an operation and maintenance time optimization module; The operation and maintenance time prediction module takes the acquired operation risk coefficient, environmental risk coefficient, load characteristic parameters, and next failure type, and inputs them into the trained machine learning model to predict the operation and maintenance time.
2. The material handling robot control system according to claim 1, characterized in that, The operational status data includes cumulative operating time, robot joint movement frequency, motor operating temperature, and joint limit movement amplitude frequency. The method for obtaining the operational risk coefficient includes: Set corresponding weight coefficients for the running status data; The operational status data are weighted and summed according to their respective weights to obtain the operational risk coefficient, which comprehensively represents the overall operational risk level of the robot.
3. The material handling robot control system according to claim 2, characterized in that, The method for obtaining the joint motion frequency includes: recording the angle change of the joint during operation using a joint angle sensor, obtaining the frequency by counting the number of angle changes per unit time, sequentially obtaining the joint motion frequency of each joint, establishing a set, and obtaining the average value.
4. The material handling robot control system according to claim 2, characterized in that, The method for obtaining the frequency of the joint's limit motion amplitude includes: Preset threshold for joint angle changes; The angle changes of the joint during operation are recorded using a joint angle sensor; When the change in joint angle exceeds the set threshold, the counter is triggered to record once; Statistical counting results within a unit of time.
5. The material handling robot control system according to claim 1, characterized in that, The environmental status data includes workshop noise data, workshop voltage fluctuation data, and vibration velocity data of the handling robot; The method for obtaining the environmental risk coefficient includes: Set corresponding weight coefficients for the environmental status data; The environmental risk coefficient, which comprehensively represents the environmental status of the production site, is obtained by weighting and summing the above environmental status data according to their respective weights.
6. The material handling robot control system according to claim 1, characterized in that, The training method for the machine learning model that predicts maintenance time includes: The collected operational risk coefficient, environmental risk coefficient, load characteristic parameters, and next fault type are converted into a corresponding set of feature vectors; Each set of feature vectors is used as input to the machine learning model. The machine learning model outputs the operational risk coefficient, environmental risk coefficient, load characteristic parameters, and the maintenance time corresponding to the next fault type for each set of operational risk coefficients, environmental risk coefficients, load characteristic parameters, and the actual maintenance time corresponding to the next fault type for each set of operational risk coefficients, environmental risk coefficients, load characteristic parameters, and the next fault type. The training objective is to minimize the loss function value of the machine learning model. Training stops when the loss function value of the machine learning model is less than or equal to the preset target loss value. The machine learning model is one of the following: support vector machine regression, random forest regression, or neural network regression model.
7. The material handling robot control system according to claim 1, characterized in that, The maintenance time optimization module includes: The maintenance record acquisition module retrieves the occurrence time of historical repairable faults of the handling robot, the interval T between two consecutive faults, and establishes a set M, M={T1, T2, ..., T...}. n }; The average value calculation module obtains the average value over the interval time. ; The analysis and control module compares the predicted maintenance time with the average of the interval time to obtain the time decay value. It then compares and analyzes the time decay value with the preset safe time decay threshold to generate maintenance time compensation instructions.
8. The material handling robot control system according to claim 7, characterized in that, The expression for the time decay value is: ； In the formula, This is the time decay value. For the predicted maintenance time, This represents the average value over the intervals. The maintenance time compensation instructions include a first shortening instruction, a second shortening instruction, and a third shortening instruction, and the shortening time of the first shortening instruction, the second shortening instruction, and the third shortening instruction increases sequentially; Methods for generating maintenance time compensation instructions include: Preset time decay thresholds SB1, SB2, and SB3, where 0 < SB1 < SB2 < SB3; when If ≤SB1, then no maintenance time compensation instruction will be generated; When SB1 < If ≤SB2, then the first shortening instruction is generated; When SB2 < If ≤SB3, then a second shortening instruction is generated; when When >SB3, a third shortening instruction is generated.
9. A material handling robot control system according to claim 1, characterized in that, The method for predicting the next failure type of the handling robot includes: The collected operational risk coefficient, environmental risk coefficient, and load characteristic parameters are input into a pre-built fault type prediction model, which outputs the next fault type of the handling robot. The fault types include motor faults, sensor malfunctions, control system errors, drive overloads, communication malfunctions, mechanical structure wear, wheel set faults, and load offset anomalies; the training method for the fault type prediction model includes: Q sets of training data are collected in advance, where Q is a positive integer greater than 0. The training data includes the operating risk coefficient, the environmental risk coefficient, and the load characteristic parameters, as well as the fault type labels corresponding to the operating risk coefficient, the environmental risk coefficient, and the load characteristic parameters. A deep learning model was selected as the fault type prediction model. The model was trained using training data. The operational risk coefficient, environmental risk coefficient, and load characteristic parameters were used as inputs to the model, and the fault type label was used as the output. The stochastic gradient descent method was used, and the weights and biases of the model were adjusted through backpropagation to minimize the error between the prediction results and the actual results. A loss function, which is the mean squared error, was set. When the loss function value converged, the training of the fault type prediction model was stopped, and the fault type prediction model corresponding to the convergence of the loss function value was taken as the trained fault type prediction model.
10. A control method for a material handling robot, characterized in that, include: Collect operational status data of the handling robot; The operation status data of the handling robot is obtained, and the operation status data of the handling robot is weighted and assigned based on the weight self-learning model to determine the operation risk coefficient of the handling robot. Collect environmental status data from the production site of the handling robot; This is used to acquire environmental status data of the production site of the handling robot, and to assign weights to the environmental status data of the production site of the handling robot based on a weighted self-learning model to determine the environmental risk coefficient of the handling robot. Collect the load characteristic parameters of the current operation, including the weight of the load and the rate of change of the load; Obtain the current job type of the handling robot, which includes handling, palletizing, and transporting, and encode it using one-hot encoding. Predict the type of failure that will occur next for the handling robot; The acquired operational risk coefficient, environmental risk coefficient, load characteristic parameters, and next failure type are input into the trained machine learning model for predicting operation and maintenance time to predict the operation and maintenance time. Obtain the occurrence time of repairable faults in the history of the handling robot, obtain the time interval T between two adjacent faults, establish a set, and obtain the average value of the time interval; The predicted maintenance time is compared with the average of the interval time to obtain the time decay value. The time decay value is then compared and analyzed with the preset safe time decay threshold to generate maintenance time compensation instructions.