A system and method for locating stacking faults of unmanned forklifts based on machine learning
By integrating multiple sensors and machine learning algorithms, the unmanned forklift stacking fault location system automatically detects and locates faults, solving the problem of low efficiency in manual diagnosis in existing technologies and achieving efficient and accurate fault identification and location.
Patent Information
- Application Number
- CN202410788630.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-19
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-06-19
AI Technical Summary
Existing fault diagnosis methods for unmanned forklifts rely on human experience, which is inefficient and prone to misjudgment, making it difficult to meet the high-efficiency operation requirements of modern logistics and warehousing industries.
An unmanned forklift stacking fault location system based on machine learning is adopted. It integrates multiple sensors to collect data in real time, and combines sensitivity recognition algorithm and similarity comparison method to realize automatic fault detection and location. It uses ant colony optimization algorithm to screen key data points and uses local area network historical data to assist in fault identification.
It improves the accuracy and efficiency of fault diagnosis, reduces manual intervention, can quickly identify minor abnormal data changes, and promptly detect potential faults, thus meeting the high-efficiency operation requirements of modern logistics and warehousing industries.
Smart Images

Figure CN118364411B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of logistics warehousing technology, and in particular to an unmanned forklift stacking fault location system and method based on machine learning. Background Art
[0002] Nowadays, in the modern logistics and warehousing industries, the widespread use of unmanned forklifts has greatly improved the efficiency of cargo handling and stacking operations, while reducing labor costs. Through automated operations, unmanned forklifts not only reduce the possibility of human errors, but also improve the efficiency and accuracy of the entire logistics chain. With the development of technology, unmanned forklifts are increasingly widely used in the fields of logistics and warehousing, becoming an indispensable part of the modern logistics system.
[0003] However, unmanned forklifts may experience various faults during long-term operation. These faults may include motor failures, such as motor overheating, damage, or unstable speed; sensor failures, such as damage or failure of sensors such as lidar and encoders; and transmission system failures, such as worn transmission belts and damaged gears. These faults will not only affect the normal operation of unmanned forklifts, but may also cause interruptions to the entire warehouse operation, thereby affecting the efficiency and stability of the entire logistics system.
[0004] At present, traditional fault diagnosis methods mainly rely on manual experience and on-site inspections by technicians. This method is often inefficient, takes a long time to diagnose faults, and may lead to misjudgment. In addition, manual diagnosis methods require technicians to have certain professional knowledge and experience, and have high requirements on the quality of technicians. Under the requirements of efficient operation in the logistics and warehousing industries, traditional fault diagnosis methods have been unable to meet the needs of modern logistics systems. Therefore, the present invention proposes a system and method for locating unmanned forklift stacking faults based on machine learning. Summary of the invention
[0005] In response to the problems in the related technology, the present invention proposes an unmanned forklift stacking fault location system and method based on machine learning to overcome the above-mentioned technical problems existing in the existing related technology.
[0006] To this end, the specific technical solution adopted by the present invention is as follows:
[0007] According to one aspect of the present invention, there is provided an unmanned forklift stacking fault location system based on machine learning, comprising a data acquisition module, a stacking control module, a fault location module and a visualization module;
[0008] The data acquisition module is used to collect real-time data of the unmanned forklift during operation through the sensor component;
[0009] The stacking control module is used to identify the position and posture of the shelf and control the unmanned forklift to perform stacking operations based on the identification results;
[0010] The fault location module is used to detect and locate stacking faults based on the collected data through a sensitivity-based recognition algorithm combined with a similarity comparison method;
[0011] The visualization module is used to display the operating status, sensor data and fault diagnosis results of the unmanned forklift in real time, automatically generate a fault report, and issue an alarm when a fault is detected;
[0012] Wherein, the fault location module includes a real-time data acquisition module, a fault data identification module and a stacking fault location module;
[0013] The real-time data acquisition module is used to acquire real-time data collected by the sensor component;
[0014] The fault data identification module is used to identify abnormal data in real-time data based on a sensitivity identification algorithm;
[0015] The stacking fault locating module is used to identify and locate stacking faults based on a similarity comparison method combined with the recognition result of abnormal data.
[0016] Preferably, the sensor assembly includes a positioning laser radar, an upper and lower laser radar, a draw wire encoder, a motor encoder, a temperature sensor, a vibration sensor, a current sensor and a weighing sensor;
[0017] The positioning laser radar is used to obtain the position data of the unmanned forklift in real time, as well as environmental scanning and obstacle detection;
[0018] The upper and lower laser radars are used to obtain point cloud data of the fork upper shelf and the ground shelf respectively;
[0019] The pull rope encoder is used to obtain the fork height data of the unmanned forklift in real time;
[0020] The motor encoder is used to obtain real-time speed and displacement data of each motor in the unmanned forklift;
[0021] The temperature sensor is used to obtain the real-time temperature of the motor and battery in the unmanned forklift;
[0022] The vibration sensor is used to obtain the real-time mechanical vibration conditions in the unmanned forklift;
[0023] The current sensor is used to obtain real-time current data of each electrical component in the unmanned forklift;
[0024] The weighing sensor is used to obtain real-time load data of the unmanned forklift during the handling process.
[0025] Preferably, the stacking control module includes a path planning module, a shelf identification module and a drive control module;
[0026] The path planning module is used to set the moving path of the unmanned forklift by combining the path planning algorithm with the operation requirements;
[0027] The shelf recognition module is used to recognize the position and posture of the shelves on the fork tine and the shelves on the ground by analyzing the point cloud data, and feed back the recognition results to the drive control module;
[0028] The drive control module is used to control the unmanned forklift to perform stacking operations according to the recognition result and the moving path.
[0029] Preferably, the fault data identification module includes:
[0030] Based on the structure and function of the unmanned forklift, an equivalent undirected graph is constructed, where the nodes in the equivalent undirected graph represent key components and sensors, and the edges in the equivalent undirected graph represent data flows or influence relationships;
[0031] Conduct sensitivity analysis on the acquired real-time data to determine the key data that affects the operating status of the unmanned forklift; and screen out the key data points for monitoring the status of the unmanned forklift based on the ant colony optimization algorithm;
[0032] Monitor the selected key data points in real time and perform standardization processing to calculate the standardized residuals of the key data points;
[0033] The analysis compares the standardized residuals of key data points with the preset standardized residual thresholds, and marks the key data points that exceed the thresholds as abnormal data.
[0034] Preferably, the sensitivity analysis of the acquired real-time data to determine the key data affecting the operating state of the unmanned forklift includes:
[0035] According to the physical characteristics and control logic of the unmanned forklift, a state equation is established, the state of the unmanned forklift is associated with the control instruction, and the established state equation is linearized to obtain a linearized state space model;
[0036] Calculating a Jacobian matrix of the state equation, wherein each column of the Jacobian matrix represents a partial derivative of a corresponding state variable with respect to each control instruction; analyzing elements in the Jacobian matrix to determine state variables whose sensitivity to the control instruction is higher than a preset threshold;
[0037] According to the analysis results of the Jacobian matrix, the state variables are sorted by sensitivity, and the key data that affect the operating status of the unmanned forklift are screened out according to the results of the sensitivity sorting.
[0038] Preferably, the key data points for monitoring the status of unmanned forklifts screened out based on the ant colony optimization algorithm include:
[0039] Initialize the ant population and initialize pheromones on each edge of the equivalent undirected graph; define the path length and transition probability;
[0040] Based on the transition probability, the ant selects the next path and updates the pheromone according to the quality of the selected path after completing a single cycle. The pheromone update includes local update and global update.
[0041] Determine whether the maximum number of iterations or the global optimal path has been reached. If not, continue iterating. If so, define the data points in the selected optimal path as key data points.
[0042] Preferably, the calculation formula of the transition probability is:
[0043]
[0044] The formula for pheromone update is:
[0045]
[0046] In the formula, t ij ( t )express t Time from node i To Node j pheromone concentration; or ij ( t )express t Time from node i To Node j Heuristic information; s Indicates the point to be transferred; T k Represents ants k Taboo table; α and β Represent the importance coefficients of pheromone and heuristic information respectively; r represents the pheromone evaporation rate; Dt ij Represents a slave node i To Node j pheromone increase.
[0047] Preferably, the stacking fault location module includes:
[0048] Obtain abnormal data of the current unmanned forklift, connect with other unmanned forklifts in the warehouse through the local area network, and identify unmanned forklifts of the same model;
[0049] Determine whether there is historical data similar to the abnormal data of the current unmanned forklift in the historical database of the same model of unmanned forklift. If not, use the pre-built fault detection model combined with the abnormal data of the current unmanned forklift to identify and locate the stacking fault. If yes, execute the next step;
[0050] The unmanned forklift is screened based on the judgment results of similar data. Based on the working environment, load type, shelf type and shelf layout of the unmanned forklift, the similarity between the operating conditions of the current forklift and the screened unmanned forklift is calculated using the similarity comparison method combined with the preset weight factor;
[0051] The unmanned forklift that is most similar to the current unmanned forklift state is determined based on the sorting results of the operating condition similarity, and the fault type and fault location of the current unmanned forklift are determined based on similar historical data in the unmanned forklift history database.
[0052] Preferably, the calculation formula of the operating condition similarity is:
[0053]
[0054] In the formula, S ( a , b ) indicates unmanned forklift a and unmanned forklifts b Similarity of working conditions; They represent the weight factors of working environment, load type, shelf type and shelf layout respectively; S e , S l , S h , S b They represent the similarity of working environment, load type, shelf type and shelf layout respectively.
[0055] According to another aspect of the present invention, a method for locating stacking faults of unmanned forklifts based on machine learning is provided. The method for locating stacking faults of unmanned forklifts based on machine learning comprises the following steps:
[0056] S1. Collect real-time data of the unmanned forklift during operation through sensor components;
[0057] S2, identify the position and posture of the shelf, and control the unmanned forklift to perform stacking operations based on the recognition results;
[0058] S3. Detection and location of stacking faults are achieved through a sensitivity-based recognition algorithm combined with a similarity comparison method based on the collected data;
[0059] S4, real-time display of the unmanned forklift's operating status, sensor data and fault diagnosis results, automatic generation of fault reports, and alarm when a fault is detected.
[0060] Compared with the prior art, the present invention provides an unmanned forklift stacking fault location system and method based on machine learning, which has the following beneficial effects:
[0061] (1) The present invention integrates multiple sensors to collect forklift operation data in real time, and combines advanced machine learning algorithms to achieve automatic detection and location of faults. This can not only improve the accuracy of fault diagnosis, but also greatly improve the diagnostic efficiency and reduce the need for manual intervention, thereby better meeting the efficient operation requirements of the modern logistics and warehousing industries.
[0062] (2) The present invention can not only use the sensitivity-based recognition algorithm to identify tiny abnormal data changes and improve the sensitivity of fault detection, but also use the ant colony optimization algorithm to screen out key data points for unmanned forklift status monitoring and improve the pertinence of fault diagnosis, so that abnormal data in the stacking process can be detected quickly and accurately, thereby discovering potential faults in a timely manner.
[0063] (3) The present invention can use the historical data of the same model of unmanned forklifts in the local area network as a reference, and determine the subsequent fault identification and location strategy by comparing the similarity between the abnormal data of the current unmanned forklift and the historical data. When there is no similar data in the historical data of the same model of unmanned forklift, the pre-built fault detection model can be combined with the abnormal data of the current unmanned forklift to realize the identification and location of the stacking fault, so that new or unprecedented fault conditions can be effectively handled. When there is similar data in the historical data of the same model of unmanned forklifts, the most similar unmanned forklift can be determined based on the sorting results of the operating condition similarity, and the fault category and fault location can be determined based on the similar historical data, so that the fault category and location can be quickly determined, and the information in the historical data can be fully utilized to improve the accuracy and efficiency of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0065] Figure 1 It is a structural block diagram of an unmanned forklift stacking fault location system based on machine learning according to an embodiment of the present invention.
[0066] In the figure:
[0067] 1. Data acquisition module; 2. Stacking control module; 21. Path planning module; 22. Shelf identification module; 23. Drive control module; 3. Fault location module; 31. Real-time data acquisition module; 32. Fault data identification module; 33. Stacking fault location module; 4. Visualization module. DETAILED DESCRIPTION
[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0069] According to an embodiment of the present invention, a system and method for locating stacking faults of unmanned forklifts based on machine learning are provided.
[0070] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, there is provided an unmanned forklift stacking fault location system based on machine learning, comprising a data acquisition module 1, a stacking control module 2, a fault location module 3 and a visualization module 4;
[0071] The data acquisition module 1 is used to collect real-time data of the unmanned forklift during operation through the sensor component;
[0072] Specifically, the sensor assembly includes a positioning laser radar, an upper and lower laser radar, a draw wire encoder, a motor encoder, a temperature sensor, a vibration sensor, a current sensor, and a weighing sensor;
[0073] The positioning laser radar is used to obtain the position data of the unmanned forklift in real time, as well as environmental scanning and obstacle detection;
[0074] The upper and lower laser radars are used to obtain point cloud data of the fork upper shelf and the ground shelf respectively;
[0075] The pull rope encoder is used to obtain the fork height data of the unmanned forklift in real time;
[0076] The motor encoder is used to obtain real-time speed and displacement data of each motor in the unmanned forklift;
[0077] The temperature sensor is used to obtain the real-time temperature of the motor and battery in the unmanned forklift;
[0078] The vibration sensor is used to obtain the real-time mechanical vibration conditions in the unmanned forklift;
[0079] The current sensor is used to obtain real-time current data of each electrical component in the unmanned forklift;
[0080] The weighing sensor is used to obtain real-time load data of the unmanned forklift during the handling process.
[0081] The stacking control module 2 is used to identify the position and posture of the shelf, and control the unmanned forklift to perform stacking operations based on the identification results;
[0082] Specifically, the stacking control module 2 includes a path planning module 21, a shelf identification module 22 and a drive control module 23;
[0083] The path planning module 21 is used to set the moving path of the unmanned forklift by combining the path planning algorithm with the operation requirements;
[0084] The shelf recognition module 22 is used to recognize the position and posture of the shelf on the fork and the shelf on the ground by analyzing the point cloud data, and feed back the recognition result to the drive control module; specifically, it includes:
[0085] Shelf recognition: By analyzing point cloud data, the position and posture of the shelves on the forks and the shelves on the ground are identified. The shelves are mainly located by finding the point cloud clusters of the four legs of the shelves.
[0086] Relative posture calculation: During the stacking operation, the posture of the shelf on the fork and the shelf on the ground are obtained at the same time, and the relative posture is fed back to the control system to ensure accurate stacking.
[0087] The driving control module 23 is used to control the unmanned forklift to perform stacking operations according to the recognition result and the moving path, specifically including:
[0088] Travel motor control: controls the forward, reverse and speed of the forklift.
[0089] Steering motor control: Controls the forklift's steering to ensure accurate positioning of the forklift.
[0090] Fork motor control: Control the lifting and tilting of the forks, and adjust the position of the unmanned forklift in real time according to the results of the shelf recognition algorithm.
[0091] In addition, the stacking control module may also include a stacking device, which mainly includes a chassis moving mechanism and a shelf lifting mechanism. When the shelf is lifted or lowered, it will be lifted or lowered with the sensor system. It is responsible for inserting a shelf from one place and then stacking the shelf on another shelf in another place.
[0092] Specifically, the stacking control module 2 identifies the position and posture of the shelf and controls the unmanned forklift to perform the stacking operation based on the identification result, including:
[0093] First, the unmanned forklift is switched to manual mode, and the forklift is manually controlled to move in the working area. Slam mapping is performed based on the positioning radar data. At this time, slam positioning based on the grid map and positioning radar can obtain the accurate position of the forklift in real time.
[0094] Then, the forklift is manually controlled to collect the positions of the first shelf and the second shelf, and the path to the first shelf and the path to the second shelf for stacking are planned. At this time, the forklift can move autonomously on the planned path based on the forklift's real-time position and control system.
[0095] Next, when the forklift is given a stacking task, it will automatically go to the first shelf point. When the forklift reaches the front of the shelf, the shelf recognition algorithm will be executed, and then the first shelf will be inserted according to the obtained shelf posture. The shelf recognition algorithm first obtains the point cloud cluster through point cloud clustering, and then analyzes the combination of all 4 point cloud clusters. The shelf legs are obtained according to the rectangularity, rectangular length, rectangular width and aspect ratio of the point cloud cluster. By adjusting the rectangularity, rectangular length and width and aspect ratio, multiple shelves can be recognized at the same time, and shelves on the fork teeth and shelves on the ground can be recognized at the same time during stacking.
[0096] Finally, the forklift automatically goes to the second shelf point. When it reaches the front of the shelf, it performs shelf recognition algorithms on both the shelf on the fork tine and the shelf on the ground. According to the relative position of the two shelves, the control system adjusts the forklift to complete the high-precision stacking task.
[0097] The fault location module 3 is used to detect and locate stacking faults based on the collected data through a sensitivity-based recognition algorithm combined with a similarity comparison method;
[0098] The fault location module 3 includes a real-time data acquisition module 31, a fault data identification module 32 and a stacking fault location module 33;
[0099] The real-time data acquisition module 31 is used to acquire real-time data collected by the sensor component;
[0100] The fault data identification module 32 is used to identify abnormal data in real-time data based on a sensitivity identification algorithm; when the fault data identification module 32 identifies abnormal data in real-time data based on a sensitivity identification algorithm, it includes:
[0101] 1) Based on the structure and function of the unmanned forklift, an equivalent undirected graph is constructed, where the nodes in the equivalent undirected graph represent key components and sensors (specifically, each node represents a key component or sensor in the unmanned forklift system. For example, motors, batteries, sensors, controllers, etc. can all be nodes, and each node has an important impact on the overall performance of the system), and the edges in the equivalent undirected graph represent data flows or influence relationships (specifically, the edges in the graph represent data flows or influence relationships between nodes. For example, the temperature of the motor will affect its performance, so there will be an edge between the motor node and the temperature sensor node to represent this relationship. This relationship can be a direct physical connection, or it can be a functional dependency or data interaction);
[0102] By constructing such an equivalent undirected graph, complex system structures can be made more intuitive, making it easier to understand the components of the system and their relationships. By analyzing the paths and connections in the graph, we can understand the interactions between different components or sensors and how they affect the behavior of the entire system. During fault diagnosis, potential fault points can be identified by checking the nodes and edges in the graph. For example, if a sensor node detects an anomaly, the possible source of the fault can be traced by checking the edges connected to it. The sensitivity of different components or sensors to system performance can be evaluated by analyzing the weights of the edges or the degrees of the nodes, thereby determining which parts are key to fault diagnosis.
[0103] 2) Conduct sensitivity analysis on the acquired real-time data to determine the key data that affects the operating status of the unmanned forklift, including:
[0104] According to the physical characteristics and control logic of the unmanned forklift, the state equation is established, the state of the unmanned forklift (such as position, speed, acceleration, etc.) is associated with the control instructions (such as motor current, steering angle, etc.), and the established state equation is linearized to obtain a linearized state space model. The linearization process helps to simplify the calculation and can clearly express the relationship between the state variables and the input variables;
[0105] Calculate the Jacobian matrix of the state equation, which describes the sensitivity of the state variable to the control input. Each column of the Jacobian matrix represents the partial derivative of the corresponding state variable to each control command.
[0106] Analyze the elements in the Jacobian matrix to determine which state variables are more sensitive to the control input. These state variables with higher sensitivity are usually the key data that affect the operating status of the unmanned forklift.
[0107] The state variables are sorted by sensitivity according to the analysis results of the Jacobian matrix. Sensitivity sorting helps to identify the key data that has the greatest impact on the operating status of the unmanned forklift, and screen out the key data that affects the operating status of the unmanned forklift according to the results of sensitivity sorting.
[0108] 3) Based on the ant colony optimization algorithm, the key data points for monitoring the status of unmanned forklifts are selected; specifically, they include:
[0109] Initialize the ant population, where each ant represents a possible solution, and initialize pheromone on each edge of the equivalent undirected graph, usually set to a small constant;
[0110] Define path length as the residual of measurement data or other indicators, such as signal quality, sensor stability, etc. Define the probability of an ant moving from one node to another (i.e., transition probability), based on pheromone concentration and heuristic information (e.g., the inverse of path length);
[0111] Based on the transition probability, the ant selects the next path and updates the pheromone according to the quality of the selected path after completing a single cycle (the pheromone of the path with better quality increases, and the pheromone of the path with poor quality decreases). The pheromone update includes local update (that is, each ant updates the local pheromone of the path after completing the path) and global update (that is, all ants update the pheromone of the global optimal path after completing one iteration);
[0112] The calculation formula of the transition probability is:
[0113]
[0114] The formula for pheromone update is:
[0115]
[0116] In the formula, t ij ( t )express t Time from node i To Node j pheromone concentration; or ij ( t )express t Time from node i To Node j Heuristic information; s Indicates the point to be transferred;T k Represents ants k Taboo table; α and β Represent the importance coefficients of pheromone and heuristic information respectively; r represents the pheromone evaporation rate; Dt ij Represents a slave node i To Node j pheromone increase.
[0117] Determine whether the maximum number of iterations or the global optimal path has been reached. If not, continue iterating. If so, define the data points in the selected optimal path as key data points.
[0118] 4) Real-time monitoring of the selected key data points and standardization processing, calculate the standardized residuals of the key data points, the calculation formula is: , where r represents the raw residual, which is the difference between the observed value and the value predicted by the model, represents the mean of the original residuals, s r represents the standard deviation of the original residual;
[0119] 5) Analyze and compare the standardized residuals of key data points with the preset standardized residual threshold, and mark the key data points exceeding the threshold as abnormal data.
[0120] The stacking fault locating module 33 is used to identify and locate the stacking fault based on the similarity comparison method combined with the recognition result of the abnormal data.
[0121] Specifically, the stacking fault location module 33 includes:
[0122] 1) Obtain abnormal data of the current unmanned forklift, connect with other unmanned forklifts in the warehouse through the local area network, and identify unmanned forklifts of the same model;
[0123] 2) Determine whether there is historical data similar to the abnormal data of the current unmanned forklift in the historical database of the same model of unmanned forklift. If not, use the pre-built fault detection model combined with the abnormal data of the current unmanned forklift to identify and locate the stacking fault. If yes, execute the next step;
[0124] To determine whether there is historical data similar to the abnormal data of the current unmanned forklift in the historical database of the same model of unmanned forklift, and to identify and locate the stacking fault based on this, the following steps can be followed:
[0125] Data collection: First, collect abnormal data of the current unmanned forklift. This data includes readings of sensors such as motor current, temperature, speed, acceleration, and any other indicators related to forklift performance.
[0126] Historical database query: Compare the abnormal data of the current unmanned forklift with the data in the historical database. The historical database should contain a large amount of normal and abnormal data of the same model of unmanned forklift under different working conditions.
[0127] Similarity judgment: Use similarity measurement algorithms, such as Euclidean distance, cosine similarity, or Mahalanobis distance, to judge the similarity between the current abnormal data and the historical data. If there is a high degree of similarity between the current data and the historical data, it can be considered that the fault has been recorded in the historical data.
[0128] Fault detection model application: If no similar data is found in the historical database, a pre-built fault detection model can be used to identify and locate the fault. These models are based on machine learning algorithms such as support vector machines (SVM), neural networks (NN) or decision trees, which have been trained on historical data.
[0129] Fault identification and location: The fault detection model is used to analyze the abnormal data of the current unmanned forklift to determine the type and possible location of the fault. The model will output a fault diagnosis result indicating the type of fault and the component or subsystem that may have failed.
[0130] In this way, the system can effectively use historical data to assist current fault diagnosis, while also being able to rely on pre-built fault detection models to identify and locate faults when encountering new or unknown fault conditions. This approach combines the experience of historical data with the predictive power of machine learning to improve the accuracy and efficiency of fault diagnosis.
[0131] 3) The unmanned forklift is screened according to the judgment results of similar data. Based on the working environment, load type, shelf type and shelf layout of the unmanned forklift, the similarity of the operating conditions between the current forklift and the screened unmanned forklift is calculated by using the similarity comparison method combined with the preset weight factor; wherein the calculation formula of the similarity of the operating conditions is:
[0132]
[0133] In the formula, S ( a , b ) indicates unmanned forklift a and unmanned forklifts b Similarity of working conditions; They represent the weight factors of working environment, load type, shelf type and shelf layout respectively; S e , S l , S h , S b They represent the similarity of working environment, load type, shelf type and shelf layout respectively.
[0134] 4) Based on the sorting results of the operating condition similarity, the unmanned forklift that is most similar to the current unmanned forklift status is determined, and the fault type and fault location of the current unmanned forklift are determined based on similar historical data in the unmanned forklift history database.
[0135] The steps of determining the unmanned forklift that is most similar to the current unmanned forklift status based on the sorting results of the operating condition similarity, and determining the fault type and fault location of the current unmanned forklift based on similar historical data in the unmanned forklift history database are as follows:
[0136] Calculation of similarity of operating conditions: First, the similarity of the operating conditions of the current forklift and all the same type of unmanned forklifts in the historical database is calculated based on the working environment, load type, shelf type and shelf layout of the current unmanned forklift. This is achieved by designing a similarity scoring system, where each factor is assigned a weight according to its importance, and then a weighted similarity score is calculated.
[0137] Sorting similarity results: The similarity scores of all unmanned forklifts are sorted, and the forklift with the highest score is considered to be the most similar to the current forklift status.
[0138] Find similar historical data: Search the historical data related to the most similar unmanned forklift in the historical database, especially the data similar to the current abnormal data.
[0139] Fault category and location determination: Analyze the historical data of the most similar unmanned forklifts to determine the fault category and location recorded in them. This information will be used as a reference to infer the possible fault category and location of the current unmanned forklift.
[0140] Output fault diagnosis results: Based on the above analysis, output the fault diagnosis results of the current unmanned forklift, including the fault type and possible occurrence location.
[0141] Update database: Update the abnormal data and fault diagnosis results of the current unmanned forklift to the historical database to facilitate future fault diagnosis and model optimization.
[0142] Through this method, the system can use the fault information under similar operating conditions in the historical data to assist in diagnosing the fault of the current unmanned forklift, improving the accuracy and efficiency of fault diagnosis. This method is particularly suitable for scenarios with rich historical data and little change in operating conditions.
[0143] The visualization module 4 is used to display the operating status, sensor data and fault diagnosis results of the unmanned forklift in real time, automatically generate a fault report, and issue an alarm when a fault is detected.
[0144] According to another embodiment of the present invention, a method for locating stacking faults of unmanned forklifts based on machine learning is provided. The method for locating stacking faults of unmanned forklifts based on machine learning comprises the following steps:
[0145] S1. Collect real-time data of the unmanned forklift during operation through sensor components;
[0146] S2, identify the position and posture of the shelf, and control the unmanned forklift to perform stacking operations based on the recognition results;
[0147] S3. Detection and location of stacking faults are achieved through a sensitivity-based recognition algorithm combined with a similarity comparison method based on the collected data;
[0148] S4, real-time display of the unmanned forklift's operating status, sensor data and fault diagnosis results, automatic generation of fault reports, and alarm when a fault is detected.
[0149] To sum up, with the help of the above-mentioned technical scheme of the present invention, the present invention integrates multiple sensors to collect forklift operation data in real time, and combines advanced machine learning algorithms to realize automatic detection and positioning of faults, which can not only improve the accuracy of fault diagnosis, but also greatly improve the diagnosis efficiency and reduce the need for manual intervention, thereby better meeting the efficient operation needs of modern logistics and warehousing industries.
[0150] In addition, the present invention can not only use the sensitivity-based recognition algorithm to identify tiny abnormal data changes and improve the sensitivity of fault detection, but also screen out key data points for unmanned forklift status monitoring through the ant colony optimization algorithm to improve the pertinence of fault diagnosis, so that abnormal data in the stacking process can be detected quickly and accurately, thereby discovering potential faults in a timely manner.
[0151] In addition, the present invention can use the historical data of the same model of unmanned forklifts in the local area network as a reference, and determine the subsequent fault identification and positioning strategy by comparing the similarity between the abnormal data of the current unmanned forklift and the historical data. When there is no similar data in the historical data of the same model of unmanned forklift, the pre-built fault detection model can be combined with the abnormal data of the current unmanned forklift to realize the identification and positioning of the stacking fault, so that new or unprecedented fault conditions can be effectively handled. When there is similar data in the historical data of the same model of unmanned forklifts, the most similar unmanned forklift can be determined based on the sorting results of the similarity of the operating conditions, and the fault category and fault location can be determined based on the similar historical data, so that the fault category and location can be quickly determined, and the information in the historical data can be fully utilized to improve the accuracy and efficiency of fault diagnosis.
[0152] The technical features of the above-mentioned embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. A person of ordinary skill in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program. The program can be stored in a computer-readable storage medium. When the program is executed, it includes the steps described in the above method, and the storage medium, such as ROM / RAM, a disk, an optical disk, etc.
[0153] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.
Claims
1. An unmanned forklift stacking fault location system based on machine learning, characterized in that: It comprises a data acquisition module (1), a stacking control module (2), a fault location module (3) and a visualization module (4); The data acquisition module (1) is used to collect real-time data of the unmanned forklift during operation through a sensor component; The stacking control module (2) is used to identify the position and posture of the shelf, and control the unmanned forklift to perform stacking operations based on the identification results; The fault location module (3) is used to detect and locate stacking faults based on the collected data by using a sensitivity-based recognition algorithm combined with a similarity comparison method; The visualization module (4) is used to display the operating status, sensor data and fault diagnosis results of the unmanned forklift in real time, automatically generate a fault report, and issue an alarm when a fault is detected; Wherein, the fault location module (3) comprises a real-time data acquisition module (31), a fault data identification module (32) and a stacking fault location module (33); The real-time data acquisition module (31) is used to acquire real-time data collected by the sensor component; The fault data identification module (32) is used to identify abnormal data in real-time data based on a sensitivity identification algorithm; The stacking fault locating module (33) is used to identify and locate the stacking fault based on a similarity comparison method combined with the recognition result of abnormal data; The fault data identification module (32) comprises, when realizing the identification of abnormal data in real-time data based on a sensitivity identification algorithm: Based on the structure and function of the unmanned forklift, an equivalent undirected graph is constructed, where the nodes in the equivalent undirected graph represent key components and sensors, and the edges in the equivalent undirected graph represent data flows or influence relationships; Conduct sensitivity analysis on the acquired real-time data to determine the key data that affects the operating status of the unmanned forklift; and screen out the key data points for monitoring the status of the unmanned forklift based on the ant colony optimization algorithm; Monitor the selected key data points in real time and perform standardization processing to calculate the standardized residuals of the key data points; Analyze and compare the standardized residuals of key data points with the preset standardized residual thresholds, and mark the key data points exceeding the thresholds as abnormal data; The stacking fault locating module (33) comprises: Obtain abnormal data of the current unmanned forklift, connect with other unmanned forklifts in the warehouse through the local area network, and identify unmanned forklifts of the same model; Determine whether there is historical data similar to the abnormal data of the current unmanned forklift in the historical database of the same model of unmanned forklift. If not, use the pre-built fault detection model combined with the abnormal data of the current unmanned forklift to identify and locate the stacking fault. If yes, execute the next step; The unmanned forklift is screened based on the judgment results of similar data. Based on the working environment, load type, shelf type and shelf layout of the unmanned forklift, the similarity between the operating conditions of the current forklift and the screened unmanned forklift is calculated using the similarity comparison method combined with the preset weight factor; The unmanned forklift that is most similar to the current unmanned forklift state is determined based on the sorting results of the operating condition similarity, and the fault type and fault location of the current unmanned forklift are determined based on similar historical data in the unmanned forklift history database.
2. The unmanned forklift stacking fault location system based on machine learning according to claim 1 is characterized in that: The sensor assembly includes a positioning laser radar, an upper and lower laser radar, a draw wire encoder, a motor encoder, a temperature sensor, a vibration sensor, a current sensor and a weighing sensor; The positioning laser radar is used to obtain the position data of the unmanned forklift in real time, as well as environmental scanning and obstacle detection; The upper and lower laser radars are used to obtain point cloud data of the fork upper shelf and the ground shelf respectively; The pull rope encoder is used to obtain the fork height data of the unmanned forklift in real time; The motor encoder is used to obtain real-time speed and displacement data of each motor in the unmanned forklift; The temperature sensor is used to obtain the real-time temperature of the motor and battery in the unmanned forklift; The vibration sensor is used to obtain the real-time mechanical vibration conditions in the unmanned forklift; The current sensor is used to obtain real-time current data of each electrical component in the unmanned forklift; The weighing sensor is used to obtain real-time load data of the unmanned forklift during the handling process.
3. The unmanned forklift stacking fault location system based on machine learning according to claim 1 is characterized in that: The stacking control module (2) comprises a path planning module (21), a shelf identification module (22) and a drive control module (23); The path planning module (21) is used to set the moving path of the unmanned forklift by combining the path planning algorithm with the operation requirements; The shelf recognition module (22) is used to recognize the position and posture of the shelf on the fork tine and the shelf on the ground by analyzing the point cloud data, and feed back the recognition result to the drive control module; The driving control module (23) is used to control the unmanned forklift to perform stacking operations based on the recognition result and the moving path.
4. The unmanned forklift stacking fault location system based on machine learning according to claim 1 is characterized in that: The sensitivity analysis of the acquired real-time data determines that the key data affecting the operating status of the unmanned forklift includes: According to the physical characteristics and control logic of the unmanned forklift, a state equation is established, the state of the unmanned forklift is associated with the control instruction, and the established state equation is linearized to obtain a linearized state space model; Calculating a Jacobian matrix of the state equation, wherein each column of the Jacobian matrix represents a partial derivative of a corresponding state variable with respect to each control instruction; analyzing elements in the Jacobian matrix to determine state variables whose sensitivity to the control instruction is higher than a preset threshold; According to the analysis results of the Jacobian matrix, the state variables are sorted by sensitivity, and the key data that affect the operating status of the unmanned forklift are screened out according to the results of the sensitivity sorting.
5. The unmanned forklift stacking fault location system based on machine learning according to claim 1 is characterized in that: The key data points for monitoring the status of unmanned forklifts screened out based on the ant colony optimization algorithm include: Initialize the ant population and initialize pheromones on each edge of the equivalent undirected graph; define the path length and transition probability; Based on the transition probability, the ant selects the next path and updates the pheromone according to the quality of the selected path after completing a single cycle. The pheromone update includes local update and global update. Determine whether the maximum number of iterations or the global optimal path has been reached. If not, continue iterating. If so, define the data points in the selected optimal path as key data points.
6. The unmanned forklift stacking fault location system based on machine learning according to claim 5 is characterized in that: The calculation formula of the transition probability is: The formula for pheromone update is: In the formula, τ ij ( t )express t Time from node i To Node j pheromone concentration; η ij ( t )express t Time from node i To Node j Heuristic information; s Indicates the point to be transferred; T k Represents ants k Taboo table; α and β Represent the importance coefficients of pheromone and heuristic information respectively; ρ represents the pheromone evaporation rate; Δτ ij Represents a slave node i To Node j pheromone increase.
7. The unmanned forklift stacking fault location system based on machine learning according to claim 1 is characterized in that: The calculation formula of the operating condition similarity is: In the formula, S ( a , b ) indicates unmanned forklift a and unmanned forklifts b Similarity of working conditions; They represent the weight factors of the working environment, load type, shelf type and shelf layout respectively; S e , S l , S h , S b They represent the similarity of working environment, load type, shelf type and shelf layout respectively.
8. A method for locating stacking faults of unmanned forklifts based on machine learning, used to implement the steps of the unmanned forklift stacking fault locating system based on machine learning as described in any one of claims 1 to 7, characterized in that: The unmanned forklift stacking fault location method based on machine learning includes the following steps: S1. Collect real-time data of the unmanned forklift during operation through sensor components; S2, identify the position and posture of the shelf, and control the unmanned forklift to perform stacking operations based on the recognition results; S3. Detection and location of stacking faults are achieved through a sensitivity-based recognition algorithm combined with a similarity comparison method based on the collected data; S4, real-time display of the unmanned forklift's operating status, sensor data and fault diagnosis results, automatic generation of fault reports, and alarm when a fault is detected.
Citation Information
Cited By
Unmanned clamping vehicle reinforcement learning reward function design method considering error accumulation
CN120029057A
