Fault detection method and system for mechanical arm of coffee beverage robot

Abstract

The invention provides a fault detection method and system for a mechanical arm of a coffee beverage robot, and relates to the technical field of fault detection of mechanical arms. And the single-mechanical-arm detection parameters and the mechanical-arm cooperative detection parameters are randomly called from the detection rule base to carry out single-mechanical-arm special detection and three-mechanical-arm cooperative detection, track deviation analysis is carried out on single-mechanical-arm track characteristics and cooperative-mechanical-arm tracks obtained through detection according to reference mechanical-arm track characteristics, and a mechanical-arm fault alarm is output. The technical problems that in the prior art, mechanical arm fault detection of a coffee beverage robot is carried out based on a fixed period, fault detection is separated from a coffee making process, and fault missing detection and false alarm are likely to happen are solved. The technical effects that the false alarm and missing alarm phenomena of mechanical arm faults are reduced, the maintenance efficiency of the mechanical arm of the coffee beverage robot is optimized, and the high precision and stability of the automatic coffee making process are guaranteed are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical arm fault detection, and in particular to a method and system for detecting mechanical arm faults of a coffee beverage robot. Background Art

[0002] With the popularization of automated coffee making technology, coffee beverage robots have become an important equipment in catering services. Their core relies on high-precision robotic arms to complete a series of actions such as cup removal, water injection, and sealing.

[0003] Existing robotic arm fault detection often relies on fixed-cycle shutdown detection or a single sensor threshold alarm mechanism, with the detection process being independent of the daily production process. These methods have significant drawbacks: fixed-cycle detection cannot cover the robotic arm's full motion path. Specific robotic arm motion patterns corresponding to low-frequency beverage production (such as lateral movement to add ingredients) are easily overlooked due to their low frequency of use, leading to missed detection of potential mechanical wear. Single-sensor threshold monitoring (such as motor current exceeding the limit) can only identify sudden hardware failures and has difficulty catching gradual performance degradation (such as slight trajectory deviation or acceleration decay).

[0004] In addition, existing technologies lack the monitoring of the spatiotemporal matching of the collaborative operation of the three robotic arms, and are unable to diagnose systemic failures caused by timing misalignment or path interference. These problems lead to low fault detection rates, high false alarm rates, and frequent maintenance shutdowns, seriously affecting equipment reliability and service efficiency. Summary of the Invention

[0005] The present invention provides a method and system for detecting faults in the robotic arm of a coffee beverage robot, which is used to solve the technical problem that the prior art performs fault detection of the robotic arm of a coffee beverage robot based on a fixed cycle, and the fault detection is separated from the coffee making process, resulting in easy occurrence of missed fault detections and false alarms.

[0006] In view of the above problems, the present invention provides a method and system for detecting faults in a coffee beverage robot arm.

[0007] A first aspect of the present invention provides a method for detecting a fault in a robotic arm of a coffee beverage robot, the method comprising:

[0008] A baseline robotic arm trajectory feature is constructed by performing no-load idling monitoring on the calibrated coffee beverage robot; a periodic detection window is preset, and a fault detection instruction is generated when the operating time limit of the coffee beverage robot meets the periodic detection window; the coffee beverage robot receives and randomly calls single robotic arm detection parameters and robotic arm collaborative detection parameters from the detection rule library according to the fault detection instruction; the single robotic arm detection parameters are used to control the coffee beverage robot to perform single robotic arm special detection, and output single robotic arm trajectory features; the robotic arm collaborative detection parameters are used to control the coffee beverage robot to perform three-robotic arm collaborative detection, and output collaborative robotic arm trajectory features; trajectory deviation analysis is performed on the single robotic arm trajectory features and collaborative robotic arm trajectory based on the baseline robotic arm trajectory features, and a robotic arm fault alarm is output.

[0009] In one implementation, a baseline robot arm trajectory feature is constructed by performing no-load running monitoring on a calibrated coffee beverage robot, and the following processing is also performed:

[0010] A first multidimensional sensor group, a second multidimensional sensor group and a third multidimensional sensor group are respectively installed on the first equipment robot arm, the second equipment robot arm and the third equipment robot arm; after calibrating the first equipment robot arm, the second equipment robot arm and the third equipment robot arm, no-load idling monitoring is performed on the first equipment robot arm, the second equipment robot arm and the third equipment robot arm to drive the first equipment robot arm, the second equipment robot arm and the third equipment robot arm to simulate the standard process of coffee making; during the process of the first equipment robot arm, the second equipment robot arm and the third equipment robot arm performing K rounds of no-load idling, operation data is synchronously collected through the first multidimensional sensor group, the second multidimensional sensor group and the third multidimensional sensor group to obtain K first space-time trajectory sequences, K second space-time trajectory sequences and K third space-time trajectory sequences, where K≥20; trajectory association features are collected for the K first space-time trajectory sequences, the K second space-time trajectory sequences and the K third space-time trajectory sequences to obtain the reference robot arm trajectory features.

[0011] In one implementation, trajectory association features are collected for the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences, and the K third spatiotemporal trajectory sequences to obtain the reference robot arm trajectory features, and the following processing is further performed:

[0012] After spatially aligning the K first spatiotemporal trajectory sequences, the first single-arm benchmark trajectory feature is output by performing trajectory spatiotemporal deviation analysis; similarly, the second single-arm benchmark trajectory feature is output by performing trajectory spatiotemporal deviation analysis on the K second spatiotemporal trajectory sequences; similarly, the third single-arm benchmark trajectory feature is output by performing trajectory spatiotemporal deviation analysis on the K third spatiotemporal trajectory sequences; after aligning the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences and the K third spatiotemporal trajectory sequences by spatiotemporal mapping, the mean of the manipulator joint motion timing series is extracted and the benchmark collaborative space feature is output; wherein, the first single-arm benchmark trajectory feature, the second single-arm benchmark trajectory feature, the third single-arm benchmark trajectory feature and the benchmark collaborative space feature constitute the benchmark manipulator trajectory feature.

[0013] In one implementation, after spatially aligning the K first spatiotemporal trajectory sequences, a spatiotemporal trajectory deviation analysis is performed to output a first single-arm reference trajectory feature, and the following processing is further performed:

[0014] By locating the trajectory deviation of the K first spatiotemporal trajectory sequences, the trajectory deviation boundary is extracted as the first robotic arm motion boundary; by calculating the multiple acceleration averages of the K first spatiotemporal trajectory sequences at multiple displacement trajectory key points as the first motion feature; wherein, the first robotic arm motion boundary and the first motion feature constitute the first single-arm reference trajectory feature.

[0015] In one implementation, the coffee beverage robot receives and, based on the fault detection instruction, randomly calls single-arm detection parameters and robotic arm collaborative detection parameters from a detection rule library, and further performs the following processing:

[0016] Historical coffee production records are backtracked based on the periodic detection window to obtain multiple production frequency characteristics of various coffee beverages; sample coffee beverages are reversely screened based on the multiple production frequency characteristics to locate a set of alternative test beverages; the set of alternative test beverages is used as a random call constraint, and the single robotic arm detection parameters and robotic arm collaborative detection parameters corresponding to the random test beverage production control information are called from the detection rule library, wherein the single robotic arm detection parameters are composed of a first robotic arm detection parameter, a second robotic arm detection parameter, and a third robotic arm detection parameter.

[0017] In one implementation, the single robotic arm detection parameters are used to control the coffee beverage robot to perform a single robotic arm special detection, output a single robotic arm trajectory feature, and further perform the following processing:

[0018] In the process of using the first robotic arm detection parameter to drive the single-arm operation of the first device robotic arm, the first multi-dimensional sensor group is used to synchronously collect operation data to obtain a first real-time trajectory sequence; multiple first real-time accelerations of the first real-time trajectory sequence at the multiple displacement trajectory key points are calculated; similarly, a second real-time trajectory sequence is collected and obtained, and multiple second real-time accelerations are calculated and output based on the second real-time trajectory sequence; similarly, a third real-time trajectory sequence is collected and obtained, and multiple third real-time accelerations are calculated and output based on the third real-time trajectory sequence; the first real-time trajectory sequence, the second real-time trajectory sequence, the third real-time trajectory sequence, the multiple first real-time accelerations, the multiple second real-time accelerations and the multiple third real-time accelerations are structured and stored, and the single robotic arm trajectory feature is output.

[0019] In one implementation, the robotic arm collaborative detection parameters are used to control the coffee beverage robot to perform three-robotic arm collaborative detection, output collaborative robotic arm trajectory features, and further perform the following processing:

[0020] The first robotic arm detection parameters, the second robotic arm detection parameters, and the third robotic arm detection parameters in the synchronous execution state are used as the robotic arm collaborative detection parameters; in the process of collaboratively driving the first device robotic arm, the second device robotic arm, and the third device robotic arm using the robotic arm collaborative detection parameters, the first multidimensional sensor group, the second multidimensional sensor group, and the third multidimensional sensor group are used to synchronously collect operation data to obtain a third real-time trajectory sequence, a fourth real-time trajectory sequence, and a fifth real-time trajectory sequence; the robotic arm joint motion time series data is extracted from the third real-time trajectory sequence, the fourth real-time trajectory sequence, and the fifth real-time trajectory sequence, and the collaborative robotic arm trajectory characteristics are output.

[0021] In one implementation, trajectory deviation analysis is performed on the single robot arm trajectory characteristics and the collaborative robot arm trajectory based on the reference robot arm trajectory characteristics, a robot arm fault alarm is output, and the following processing is further performed:

[0022] If the Euclidean distance between the multiple first real-time accelerations and the mean of the multiple accelerations is greater than the preset acceleration deviation scale, and / or the first real-time trajectory sequence does not completely fall within the motion boundary of the first robotic arm, a first robotic arm alarm is output; similarly, a fault judgment is made on the second device robotic arm based on the second real-time trajectory sequence and the multiple second real-time accelerations, and a second robotic arm alarm is output; similarly, a fault judgment is made on the third device robotic arm based on the third real-time trajectory sequence and the multiple third real-time accelerations, and a third robotic arm alarm is output; based on the spatial deviation between the collaborative robotic arm trajectory feature and the benchmark collaborative space feature, a fourth robotic arm alarm is quantified and output; the first robotic arm alarm, the second robotic arm alarm, the third robotic arm alarm and the fourth robotic arm alarm are associated, and the robotic arm fault alarm is output.

[0023] The second aspect of the present invention provides a coffee beverage robot robotic arm fault detection system, the system comprising: a trajectory construction unit, for constructing a baseline robotic arm trajectory feature by performing no-load idling monitoring on a calibrated coffee beverage robot; an instruction output unit, for presetting a periodic detection window, and generating a fault detection instruction when the operating time limit of the coffee beverage robot meets the periodic detection window; a parameter calling unit, for the coffee beverage robot to receive and randomly call single robotic arm detection parameters and robotic arm collaborative detection parameters from a detection rule library according to the fault detection instruction; a single-arm detection unit, for using the single robotic arm detection parameter to control the coffee beverage robot to perform single robotic arm special detection, and output single robotic arm trajectory features; a collaborative detection unit, for using the robotic arm collaborative detection parameter to control the coffee beverage robot to perform three-robotic arm collaborative detection, and output collaborative robotic arm trajectory features; a fault alarm unit, for performing trajectory deviation analysis on the single robotic arm trajectory features and the collaborative robotic arm trajectory based on the baseline robotic arm trajectory features, and outputting a robotic arm fault alarm.

[0024] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0025] The method provided by an embodiment of the present invention constructs a baseline robotic arm trajectory signature by performing no-load idling monitoring on a calibrated coffee beverage robot. A periodic detection window is preset, and when the coffee beverage robot's operating time limit meets the periodic detection window, a fault detection instruction is generated. The coffee beverage robot receives and, based on the fault detection instruction, randomly calls single robotic arm detection parameters and robotic arm collaborative detection parameters from a detection rule library. The coffee beverage robot uses the single robotic arm detection parameters to control the coffee beverage robot to perform single robotic arm-specific detection, outputting a single robotic arm trajectory signature. The robotic arm collaborative detection parameters are used to control the coffee beverage robot to perform three-robotic arm collaborative detection, outputting a collaborative robotic arm trajectory signature. Trajectory deviation analysis is performed on the single robotic arm trajectory signature and the collaborative robotic arm trajectory based on the baseline robotic arm trajectory signature, and a robotic arm fault alarm is output. This method achieves the technical advantages of effectively distinguishing between single component failures and robotic arm collaborative function anomalies, shortening fault location time, reducing false alarms and missed alarms for robotic arm faults, optimizing the coffee beverage robot's robotic arm maintenance efficiency, and ensuring high precision and stability in the automated coffee production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic flow chart of a method for detecting a fault in a coffee beverage robot arm is shown in the present invention;

[0027] Figure 2 The schematic diagram of the structure of the coffee beverage robot mechanical arm fault detection system provided by the present invention is shown.

[0028] Explanation of the accompanying symbols: trajectory construction unit 1, instruction output unit 2, parameter calling unit 3, single-arm detection unit 4, collaborative detection unit 5, fault alarm unit 6. DETAILED DESCRIPTION

[0029] The present invention provides a method and system for detecting faults in the robotic arm of a coffee beverage robot, which is used to solve the technical problem that the prior art performs fault detection of the robotic arm of a coffee beverage robot based on a fixed cycle, and the fault detection is separated from the coffee making process, resulting in easy occurrence of missed fault detections and false alarms.

[0030] Below, the technical solutions of the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention, rather than all, are shown in the accompanying drawings.

[0031] Example 1, a flowchart of a method for detecting a coffee beverage robot arm fault according to an embodiment of the present invention, see Figure 1 , the method comprising:

[0032] S100: Build a baseline robot arm trajectory feature by performing no-load running monitoring on the calibrated coffee beverage robot;

[0033] In one implementation, by performing no-load idling monitoring on the calibrated coffee beverage robot to construct a baseline robot arm trajectory feature, the method step S100 provided by the present invention includes:

[0034] S110: Installing a first multi-dimensional sensor group, a second multi-dimensional sensor group, and a third multi-dimensional sensor group on the first device robotic arm, the second device robotic arm, and the third device robotic arm of the coffee beverage robot, respectively;

[0035] S120: After calibrating the first, second, and third equipment robotic arms, performing no-load idling monitoring on the first, second, and third equipment robotic arms to drive the first, second, and third equipment robotic arms to simulate a standard coffee making process;

[0036] S130: While the first device robotic arm, the second device robotic arm, and the third device robotic arm perform K rounds of no-load idling, synchronously collecting operation data through the first multidimensional sensor group, the second multidimensional sensor group, and the third multidimensional sensor group to obtain K first spatiotemporal trajectory sequences, K second spatiotemporal trajectory sequences, and K third spatiotemporal trajectory sequences, where K ≥ 20;

[0037] S140: Collect trajectory association features on the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences, and the K third spatiotemporal trajectory sequences to obtain the reference robotic arm trajectory features.

[0038] Specifically, in this embodiment, the coffee beverage robot includes three equipment robotic arms, and through the coordinated operation of the three robotic arms, programmed automatic production of coffee beverage orders can be achieved.

[0039] Among them, the functional designs of the first equipment robot arm, the second equipment robot arm and the third equipment robot arm are not limited in this embodiment. For example, the first robot arm is responsible for cup management, grabs empty cups and transfers them to the raw material addition area (such as the syrup / ice cube placement area), and accurately places the cups at the primary handover position; the second robot arm then takes over the raw material processing, completes the auxiliary material addition operation (such as jam injection or ice amount calibration) at the primary handover position, and then transfers the cups to the secondary handover position; the third robot arm leads the liquid operation, takes over the cups from the secondary handover position to perform coffee liquid injection, milk foam fusion and cup lid pressing and sealing, forming a functional segmented handover chain of cup retrieval → processing → liquid injection.

[0040] The entire process ensures collaborative safety and efficiency through layered spatiotemporal constraints: the second robotic arm must start grasping the cup within 0.4 seconds after the first robotic arm releases it to prevent the cup from tipping over, and the third robotic arm must take over within 0.3 seconds after the second robotic arm releases it to maintain the temperature of the drink; spatially, the primary and secondary handover points maintain a distance of ≥25cm to avoid cross-contamination of raw materials. When the three arms work collaboratively, the end effector forms a dynamic mutually exclusive area in a spherical space with a diameter of 40cm (only a single arm is allowed to enter), effectively avoiding the risk of collision.

[0041] In this embodiment, a first multi-dimensional sensor group, a second multi-dimensional sensor group, and a third multi-dimensional sensor group are installed on the first, second, and third equipment robotic arms of the coffee beverage robot, respectively. It should be understood that the sensors in the multi-dimensional sensor groups are not of a single type, but rather integrate complex functions such as angle detection, displacement tracking, and acceleration measurement to ensure that the motion details of each joint and end of the robotic arm can be fully captured.

[0042] After completing the initial calibration of the first, second, and third equipment robotic arms (such as joint zero correction and sensor zeroing), the three robotic arms are driven to perform dry runs according to the standardized process of coffee making. For example, the first robotic arm accurately grabs the cup from the cup holder and moves it along the optimized path to the raw material addition area (such as the ice outlet area), placing the cup stably in the primary handover position; the second robotic arm then takes over the cup within 0.4 seconds, completes the raw material addition operation (such as quantitative ice or syrup injection), and then moves it to the secondary handover position; the third robotic arm takes over from this position within 0.3 seconds, performs liquid operations such as high-pressure coffee liquid injection and milk foam layering and fusion, and finally moves to the cup lid to complete adaptive pressure sealing, forming a full-process closed-loop action chain of cup removal-processing-liquid injection-sealing. This process completely reproduces the real operation process, but eliminates the actual material operation.

[0043] By continuously executing the above-mentioned idle run process for at least 20 times (K≥20), the motion data of the three robotic arms during each run are synchronously collected by using the first multi-dimensional sensor group, the second multi-dimensional sensor group and the third multi-dimensional sensor group.

[0044] Each run generates three sets of data: a time-space trajectory sequence for the first robotic arm (recording its path, velocity, and acceleration changes), a time-space trajectory sequence for the second robotic arm (recording its path, velocity, and acceleration changes), and a time-space trajectory sequence for the third robotic arm (recording its path, velocity, and acceleration changes). The minimum number of runs is 20 to accumulate sufficient data, eliminate random errors, and ensure statistical significance.

[0045] Trajectory association features are collected for the K first, K second, and K third spatiotemporal trajectory sequences. Specifically, an algorithm is used to extract two types of features: first single-arm baseline trajectory features, second single-arm baseline trajectory features, third single-arm baseline trajectory features, and baseline collaborative space features. These features are finally integrated into baseline robotic arm trajectory features, which serve as a reference standard for subsequent fault detection.

[0046] This embodiment will describe in detail the method for analyzing and determining the reference robot arm trajectory characteristics in the subsequent description.

[0047] This embodiment constructs the baseline robot arm trajectory feature through the baseline data accumulated by high-frequency idle running monitoring, thereby achieving the technical effect of providing a reliable reference for subsequent diagnosis of whether the robot arm handles a fault state.

[0048] S200: Preset a periodic detection window, and generate a fault detection instruction when the operating time limit of the coffee beverage robot meets the periodic detection window;

[0049] Specifically, the preset periodic detection window is to set regular self-inspection trigger conditions for the coffee beverage robot to ensure that the equipment performs regular health checks during long-term operation.

[0050] This window is usually set based on the cumulative working time of the equipment or the number of times the equipment is made. For example, the detection process is automatically activated after every 200 cups of production or 8 hours of continuous operation.

[0051] This periodic design balances equipment availability and maintenance needs, avoiding service interruptions during peak usage periods. When the actual operating time or workload of the coffee beverage robot reaches the periodic detection window, a fault detection instruction is generated to trigger the subsequent robotic arm fault diagnosis process.

[0052] S300: The coffee beverage robot receives and randomly calls single robotic arm detection parameters and robotic arm collaborative detection parameters from a detection rule library according to the fault detection instruction;

[0053] In one implementation, the coffee beverage robot receives and randomly calls single-arm detection parameters and robotic arm collaborative detection parameters from a detection rule library according to the fault detection instruction. Step S300 of the method provided by the present invention includes:

[0054] S310: Backtracking historical coffee production records based on the periodic detection window to obtain multiple production frequency characteristics of multiple coffee beverages;

[0055] S320: Perform reverse screening of sample coffee beverages based on the multiple production frequency features to locate a set of candidate test beverages;

[0056] S330: Use the alternative test beverage set as a random call constraint, and call the single robotic arm detection parameters and robotic arm collaborative detection parameters corresponding to the random test beverage production control information from the detection rule library, wherein the single robotic arm detection parameters are composed of the first robotic arm detection parameters, the second robotic arm detection parameters and the third robotic arm detection parameters.

[0057] Specifically, this embodiment systematically traces back all production records within the time period of the periodic detection window and analyzes the production frequency of different beverages. For example, statistics show that 70% of the production tasks in the past 8 hours were American coffee, 25% were latte, and 5% were mocha. By extracting these frequency features, the impact of the action mode of high-frequency beverages on the specific load of the robotic arm is identified (such as the multiple water injection actions of American coffee may accelerate the wear of the second robotic arm's vertical downward mechanism), providing a data basis for subsequent detection parameter screening.

[0058] We reverse-screened test samples based on historical frequency data, prioritizing the production processes of low-frequency beverages. For example, if mocha coffee only accounts for 5% of production volume, the corresponding robotic arm movements (such as lateral movement during the chocolate sauce addition phase) are rarely used in daily operations, and the related transmission components may accumulate potential problems due to lack of regular activity.

[0059] By incorporating the production process of low-frequency beverages into the set of alternative test beverages, we ensure that the detection covers the entire range of motion of the robotic arm, avoiding conventional detection that only verifies high-frequency paths and misses edge scenarios.

[0060] A random test beverage is randomly selected from the candidate test set, and the production control parameters corresponding to the random test beverage are extracted from the detection rule library. Specifically, the detection rule library stores the production process control information of different coffee beverages (such as the water injection height of American coffee, the milk foam stirring speed of latte, etc.). Each beverage corresponds to a specific set of single robot arm detection parameters (for example, including the movement path parameters of the first robot arm, the pressure control parameters of the second robot arm, and the sealing control pressure parameters of the third robot arm) and robot arm collaborative detection parameters.

[0061] Based on the drink type in the candidate set (e.g., mocha, cold brew), the control parameters of one of the drinks are randomly selected and converted into detection action instructions for the robot arm. For example, if mocha coffee is selected, the single robot arm detection parameters pre-set during its production are called. The single robot arm detection parameters are composed of the first robot arm detection parameters, the second robot arm detection parameters, and the third robot arm detection parameters.

[0062] The detection parameters of the first robotic arm correspond to the speed curve of its lateral movement to the raw material addition area (such as the chocolate sauce area), the detection parameters of the second robotic arm define the auxiliary material addition action (such as the jam injection pressure threshold), and the detection parameters of the third robotic arm are related to the liquid injection height and sealing pressure value in the liquid operation stage; when these parameters are called to collaboratively drive the three robotic arms to operate, if the robotic arms are not faulty, their motion trajectory will strictly conform to the baseline robotic arm trajectory characteristics.

[0063] This random calling mechanism based on actual beverage production parameters ensures that the detection actions cover all working conditions of the robotic arm, rather than just testing fixed paths.

[0064] This embodiment covers all working conditions of the robotic arm by randomly calling diverse detection parameters, thereby achieving the technical effects of avoiding detection blind spots, improving the probability of fault mode discovery, and enhancing the robustness of fault detection.

[0065] S400: Using the single robotic arm detection parameters to control the coffee beverage robot to perform a single robotic arm special detection, and outputting a single robotic arm trajectory feature;

[0066] In one implementation, the single robotic arm detection parameters are used to control the coffee beverage robot to perform a single robotic arm special detection and output a single robotic arm trajectory feature. Step S400 of the method provided by the present invention includes:

[0067] S410: In a process of driving the single-arm operation of the first device robotic arm using the first robotic arm detection parameter, synchronously collecting operation data through the first multi-dimensional sensor group to obtain a first real-time trajectory sequence;

[0068] S420: Calculating a plurality of first real-time accelerations of the first real-time trajectory sequence at the plurality of displacement trajectory key points;

[0069] S430: Similarly, a second real-time trajectory sequence is acquired, and a plurality of second real-time accelerations are calculated and output based on the second real-time trajectory sequence;

[0070] S440: Similarly, a third real-time trajectory sequence is acquired, and a plurality of third real-time accelerations are calculated and output based on the third real-time trajectory sequence;

[0071] S450: Structurally store the first real-time trajectory sequence, the second real-time trajectory sequence, the third real-time trajectory sequence, multiple first real-time accelerations, multiple second real-time accelerations, and multiple third real-time accelerations, and output the single robotic arm trajectory feature.

[0072] This embodiment drives each robotic arm individually to perform specific detection actions, obtains its real-time motion data and extracts key features to provide input for subsequent fault judgment. During detection, only the target robotic arm (such as the first robotic arm) is activated, and the other two robotic arms remain stationary to ensure that the individual performance evaluation is not subject to collaborative interference.

[0073] Specifically, during the process of using the first robotic arm detection parameters (such as the lateral movement speed curve) to drive the first device robotic arm to operate independently, the motion data of the first device robotic arm is continuously collected through the installed first multi-dimensional sensor group.

[0074] For example, when the detection parameters require the first robotic arm to perform a cup-picking and transferring action, the sensor records its complete path coordinates, joint angle changes, and acceleration fluctuations from the cup holder to the ice discharge area, forming the first real-time trajectory sequence containing a timestamp. The first real-time trajectory sequence fully reflects the actual motion state of the robotic arm under the detection parameters.

[0075] The same method as that for calculating the multiple accelerations of the K first spatiotemporal trajectory sequences at the multiple displacement trajectory key points in the detailed explanation of step S1412 is used to calculate the multiple first real-time accelerations of the first real-time trajectory sequence at the multiple displacement trajectory key points.

[0076] Similarly, a second real-time trajectory sequence is acquired, and a plurality of second real-time accelerations are calculated and output based on the second real-time trajectory sequence. Similarly, a third real-time trajectory sequence is acquired, and a plurality of third real-time accelerations are calculated and output based on the third real-time trajectory sequence.

[0077] The first real-time trajectory sequence, the second real-time trajectory sequence, the third real-time trajectory sequence, a plurality of first real-time accelerations, a plurality of second real-time accelerations, and a plurality of third real-time accelerations are structured and stored, and the single robotic arm trajectory feature is output.

[0078] This embodiment achieves the technical effect of providing input for subsequent fault judgment by individually driving each robotic arm to perform specific detection actions, obtaining its real-time motion data and extracting key features.

[0079] S500: Using the robotic arm collaborative detection parameters to control the coffee beverage robot to perform three-robotic arm collaborative detection, and output collaborative robotic arm trajectory features;

[0080] In one implementation, the robotic arm collaborative detection parameters are used to control the coffee beverage robot to perform three-robotic arm collaborative detection and output collaborative robotic arm trajectory features. Step S500 of the method provided by the present invention includes:

[0081] S510: Using the first robotic arm detection parameter, the second robotic arm detection parameter, and the third robotic arm detection parameter in the synchronous execution state as the robotic arm collaborative detection parameter;

[0082] S520: During the process of collaboratively driving the first device robotic arm, the second device robotic arm, and the third device robotic arm to operate using the robotic arm collaborative detection parameters, synchronously collecting operation data through the first multidimensional sensor group, the second multidimensional sensor group, and the third multidimensional sensor group to obtain a third real-time trajectory sequence, a fourth real-time trajectory sequence, and a fifth real-time trajectory sequence;

[0083] S530: Extracting robot arm joint motion time series data from the third real-time trajectory sequence, the fourth real-time trajectory sequence, and the fifth real-time trajectory sequence, and outputting the collaborative robot arm trajectory features.

[0084] Specifically, this embodiment combines the independent detection parameters of the first, second, and third robotic arms into collaborative operation instructions, serving as the robotic arm collaborative detection parameters. For example, in the mocha production process, the first robotic arm's speed curve (0.4 m / s) for moving to the raw material area must be synchronized with the second robotic arm's jam injection pressure threshold (5 N) and the third robotic arm's milk foam fusion height parameter (10 cm) according to a preset timing sequence to ensure a seamless transition from cup removal to processing to liquid injection.

[0085] During the process of collaboratively driving the first device robotic arm, the second device robotic arm, and the third device robotic arm to operate using the robotic arm collaborative detection parameters, the first multidimensional sensor group, the second multidimensional sensor group, and the third multidimensional sensor group synchronously collect operation data to obtain a third real-time trajectory sequence, a fourth real-time trajectory sequence, and a fifth real-time trajectory sequence.

[0086] The robot arm joint motion time series data is extracted from the third real-time trajectory sequence, the fourth real-time trajectory sequence, and the fifth real-time trajectory sequence, and the collaborative robot arm trajectory feature is output. There is a mapping relationship between the collaborative robot arm trajectory feature and the data structure of the reference collaborative space feature. Specifically:

[0087] The collaborative robot arm trajectory features are composed of the spatiotemporal interaction data of the three robots when they collaborate, including timing synchronization parameters, spatial overlap spacing, path distances between the three robots, and measured joint linkage timing.

[0088] The collaborative detection performed in this embodiment effectively supplements the single-arm detection and provides the technical effect of providing effective detection data for the subsequent all-round identification of single-arm to three-arm faults.

[0089] S600: Perform trajectory deviation analysis on the single robot arm trajectory characteristics and the collaborative robot arm trajectory based on the reference robot arm trajectory characteristics, and output a robot arm fault alarm.

[0090] In one implementation, a trajectory deviation analysis is performed on the single robot arm trajectory characteristics and the collaborative robot arm trajectory based on the reference robot arm trajectory characteristics, and a robot arm fault alarm is output. Step S600 of the method provided by the present invention includes:

[0091] S610: If the Euclidean distance between the multiple first real-time accelerations and the multiple acceleration means is greater than a preset acceleration deviation scale, and / or the first real-time trajectory sequence does not completely fall within the first robotic arm motion boundary, output a first robotic arm alarm;

[0092] S620: Similarly, a fault diagnosis is performed on the second device robotic arm based on the second real-time trajectory sequence and the plurality of second real-time accelerations, and a second robotic arm alarm is output.

[0093] S630: Similarly, a fault diagnosis is performed on the third device robotic arm based on the third real-time trajectory sequence and the plurality of third real-time accelerations, and a third robotic arm alarm is output;

[0094] S640: Quantify and output a fourth robotic arm alarm based on the spatial deviation between the collaborative robotic arm trajectory feature and the reference collaborative spatial feature;

[0095] S650: Associate the first robotic arm alarm, the second robotic arm alarm, the third robotic arm alarm and the fourth robotic arm alarm, and output the robotic arm failure alarm.

[0096] Specifically, if the Euclidean distance between the multiple first real-time accelerations and the multiple acceleration means is greater than a preset acceleration deviation scale, and / or the first real-time trajectory sequence does not completely fall within the motion boundary of the first robotic arm, it indicates that the first device robotic arm has gear wear or loose joints, and a first robotic arm alarm is output.

[0097] Similarly, based on the second real-time trajectory sequence and multiple second real-time accelerations, the fault of the second equipment robotic arm is judged, and a second robotic arm alarm is output. Similarly, based on the third real-time trajectory sequence and multiple third real-time accelerations, the fault of the third equipment robotic arm is judged, and a third robotic arm alarm is output.

[0098] The spatial deviation index is calculated by comparing the trajectory characteristics of the collaborative robot arms (such as the handover time difference and the end-to-end distance) with the baseline collaborative spatial characteristics (such as the time difference threshold of 0.3 seconds and the distance threshold of 5mm). For example, if the handover time difference of the cups reaches 0.35 seconds (exceeding the 0.3 second threshold) or the end-to-end distance increases to 6mm (exceeding the 5mm threshold), a collaborative fault alarm (third robot arm alarm) is generated, indicating that the program timing is misaligned or the positioning accuracy has decreased.

[0099] Correlation analysis is performed on both single-arm alarms and collaborative alarms. For example, if the first arm issues a separate alarm and a collaborative alarm occurs simultaneously, it may indicate a path deviation that caused a collaborative anomaly. If only the collaborative alarm is triggered and the single arm detects normally, it may be due to communication delays or incorrect collaborative parameter configuration. The final fault alarm output will be marked with the specific fault type (such as "first arm gear wear" or "collaborative timing limit exceeded"), providing precise guidance for maintenance.

[0100] This embodiment constructs the single-arm motion boundary and collaborative space-time rules through dry run modeling, combines random parameter calls to cover all working condition paths, uses trajectory deviation analysis and dynamic threshold comparison to accurately identify mechanical wear, motor anomalies and timing misalignment, and uses a layered alarm mechanism to distinguish between single and system collaborative failures, thereby improving detection sensitivity and positioning accuracy, reducing false alarms and missed alarms, and ensuring the technical effect of reliable operation of the equipment.

[0101] This embodiment achieves the technical effect of effectively distinguishing between single component failures and abnormalities in the coordinated functions of the robotic arm, shortening the fault location time, reducing false alarms and missed alarms of robotic arm failures, optimizing the maintenance efficiency of the coffee beverage robot robotic arm, and ensuring the high precision and stability of the automatic coffee making process.

[0102] In one implementation, trajectory correlation features are collected for the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences, and the K third spatiotemporal trajectory sequences to obtain the reference robot arm trajectory features. Step S140 of the method provided by the present invention includes:

[0103] S141: After spatially aligning the K first spatiotemporal trajectory sequences, outputting a first single-arm reference trajectory feature by performing spatiotemporal deviation analysis of the trajectory;

[0104] S142: Similarly, by performing trajectory spatiotemporal deviation analysis on the K second spatiotemporal trajectory sequences, outputting a second single-arm reference trajectory feature;

[0105] S143: Similarly, by performing trajectory spatiotemporal deviation analysis on the K third spatiotemporal trajectory sequences, outputting a third single-arm reference trajectory feature;

[0106] S144: After aligning the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences, and the K third spatiotemporal trajectory sequences through spatiotemporal mapping, extracting the mean of the manipulator joint motion time series and outputting a reference collaborative space feature;

[0107] Among them, the first single-arm reference trajectory feature, the second single-arm reference trajectory feature, the third single-arm reference trajectory feature and the reference collaborative space feature constitute the reference robotic arm trajectory feature.

[0108] In one implementation, after spatially aligning the K first spatiotemporal trajectory sequences, a spatiotemporal deviation analysis of the trajectories is performed to output a first single-arm reference trajectory feature. Step S141 of the method provided by the present invention includes:

[0109] S1411: performing trajectory deviation positioning on the K first spatiotemporal trajectory sequences, and extracting a trajectory deviation boundary as a first robotic arm motion boundary;

[0110] S1412: Calculating multiple acceleration means of the K first spatiotemporal trajectory sequences at multiple displacement trajectory key points as first motion features;

[0111] Among them, the first robotic arm motion boundary and the first motion feature constitute the first single-arm reference trajectory feature.

[0112] Specifically, the K first spatiotemporal trajectory sequences recorded by the first device robot arm in K empty running tests are spatially calibrated to ensure that all data are analyzed in the same coordinate system.

[0113] For example, each time the first device's robotic arm performs a cup-removal action, it theoretically should start from the cup holder's fixed position (set as the coordinate origin). However, due to mechanical assembly errors or sensor drift, the actual starting point may be distributed within a small area around the origin. Mathematical transformations align the starting points of all trajectories to the theoretical origin, eliminating the impact of device-specific errors on data comparability.

[0114] After the alignment is completed, the fluctuations of these trajectories in terms of movement path, speed curve, acceleration change, etc. are analyzed, and the allowable deviation range under normal conditions is calculated as the first single-arm reference trajectory feature.

[0115] The specific implementation method of trajectory spatiotemporal deviation analysis is as follows:

[0116] By analyzing the K first spatiotemporal trajectory sequences of K dry run data, the maximum allowable deviation of each position point of the first device's robotic arm during the movement process is determined. For example, when the robotic arm bypasses the corner of the coffee machine body, due to joint flexibility or control delay, the actual path may present a slight arc rather than an ideal straight line. The lateral and longitudinal positions of each trajectory point in the 20 tests are counted, and the mean plus or minus three times the standard deviation is taken as the boundary (such as -1.5mm to +1.8mm horizontally, ±2mm vertically), forming a spatial envelope channel as the first robotic arm motion boundary. The first robotic arm motion boundary represents the motion corridor that the robotic arm should strictly follow in a fault-free state. Once the real-time trajectory exceeds this range, it indicates that there may be loose mechanical structure or abnormal program parameters.

[0117] The plurality of displacement trajectory key points are obtained based on equal time division, and instantaneous accelerations are calculated between nodes in the K first spatiotemporal trajectory sequences based on the plurality of displacement trajectory key points to obtain multiple sets of instantaneous accelerations. Mean values ​​of the multiple sets of instantaneous accelerations are calculated to obtain multiple acceleration means corresponding to the plurality of displacement trajectory key points as the first motion features. The first robotic arm motion boundary and the first motion feature constitute the first single-arm reference trajectory feature.

[0118] The same processing method is used for the second device robotic arm. By performing trajectory spatiotemporal deviation analysis on the K second spatiotemporal trajectory sequences, the second single-arm reference trajectory feature is output. The same processing method is used for the third device robotic arm. By performing trajectory spatiotemporal deviation analysis on the K third spatiotemporal trajectory sequences, the third single-arm reference trajectory feature is output.

[0119] The motion data of the first device robot arm, the second device robot arm and the third device robot arm are correlated and analyzed in time and space, and the spatiotemporal coordination rule parameters when the three robot arms collaborate are established as the benchmark collaborative space features.

[0120] Exemplarily, the benchmark collaborative space features include: timing synchronization thresholds, such as the second robot arm must grasp the cup within 0.3 seconds after the first robot arm releases it; spatial overlap rules, the end effector spacing must be less than 5mm during handover; path interference constraints, the minimum safe distance of the three robot arm motion envelope (such as ≥50mm); joint linkage timing, such as the phase synchronization requirement for the rotation of the second robot arm's wrist and the extension of the first robot arm's elbow.

[0121] For example, when the first robot arm places the cup on the handover platform, the second robot arm needs to reach the grasping position within 0.25 seconds, and the spatial distance between the end effectors of the two must not exceed 3 mm. By counting the time difference between the two actions in 20 tests (such as a mean of 0.18 seconds and a standard deviation of 0.03 seconds) and the spatial overlap (such as 95% of the trajectory points are within 1 mm), the time and space tolerance range of collaborative operation is defined. At the same time, the joint linkage timing is analyzed. For example, the wrist rotation movement of the second robot arm must be strictly synchronized with the elbow retraction movement of the first robot arm. If the time phase difference between the two exceeds 0.1 seconds, it is considered an abnormality.

[0122] The first single-arm reference trajectory feature, the second single-arm reference trajectory feature, the third single-arm reference trajectory feature and the reference collaborative space feature constitute the reference robotic arm trajectory feature.

[0123] This embodiment achieves the technical effect of providing an effective reference benchmark for the hierarchical detection of robot arm faults described in subsequent steps S200-S600 by separating the single robot arm features and the collaborative features.

[0124] Example 2, based on the same inventive concept as the coffee beverage robot arm fault detection method in the previous embodiment, Figure 2 As shown, the present invention provides a coffee beverage robot arm fault detection system, wherein the system includes:

[0125] The trajectory construction unit 1 is used to construct a baseline robot arm trajectory feature by performing no-load running monitoring on the calibrated coffee beverage robot;

[0126] The instruction output unit 2 is used to preset a periodic detection window and generate a fault detection instruction when the operating time limit of the coffee beverage robot meets the periodic detection window;

[0127] A parameter calling unit 3 is used for the coffee beverage robot to receive and randomly call single robot arm detection parameters and robot arm collaborative detection parameters from a detection rule library according to the fault detection instruction;

[0128] A single-arm detection unit 4 is configured to control the coffee beverage robot to perform a single-arm special detection using the single-arm detection parameters and output a single-arm trajectory feature;

[0129] A collaborative detection unit 5 is used to control the coffee beverage robot to perform three-manipulator collaborative detection using the robotic arm collaborative detection parameters, and output collaborative robotic arm trajectory features;

[0130] The fault alarm unit 6 is used to perform trajectory deviation analysis on the single robot arm trajectory characteristics and the collaborative robot arm trajectory according to the reference robot arm trajectory characteristics, and output a robot arm fault alarm.

[0131] In one implementation, the trajectory construction unit 1 is further configured to:

[0132] A first multidimensional sensor group, a second multidimensional sensor group, and a third multidimensional sensor group are installed on the first equipment robot arm, the second equipment robot arm, and the third equipment robot arm, respectively; after calibrating the first equipment robot arm, the second equipment robot arm, and the third equipment robot arm, no-load idling monitoring is performed on the first equipment robot arm, the second equipment robot arm, and the third equipment robot arm to drive the first equipment robot arm, the second equipment robot arm, and the third equipment robot arm to simulate a standard coffee making process; during the process of the first equipment robot arm, the second equipment robot arm, and the third equipment robot arm performing K rounds of no-load idling, operation data is synchronously collected by the first multidimensional sensor group, the second multidimensional sensor group, and the third multidimensional sensor group to obtain K first spatiotemporal trajectory sequences, K second spatiotemporal trajectory sequences, and K third spatiotemporal trajectory sequences, where K ≥ 20; trajectory association features are collected for the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences, and the K third spatiotemporal trajectory sequences to obtain the reference robot arm trajectory features. In one implementation, the trajectory construction unit 1 is further used to:

[0133] After spatially aligning the K first spatiotemporal trajectory sequences, the first single-arm benchmark trajectory feature is output by performing trajectory spatiotemporal deviation analysis; similarly, the second single-arm benchmark trajectory feature is output by performing trajectory spatiotemporal deviation analysis on the K second spatiotemporal trajectory sequences; similarly, the third single-arm benchmark trajectory feature is output by performing trajectory spatiotemporal deviation analysis on the K third spatiotemporal trajectory sequences; after aligning the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences and the K third spatiotemporal trajectory sequences by spatiotemporal mapping, the mean of the manipulator joint motion timing series is extracted and the benchmark collaborative space feature is output; wherein, the first single-arm benchmark trajectory feature, the second single-arm benchmark trajectory feature, the third single-arm benchmark trajectory feature and the benchmark collaborative space feature constitute the benchmark manipulator trajectory feature.

[0134] In one implementation, the trajectory construction unit 1 is further configured to:

[0135] By locating the trajectory deviation of the K first spatiotemporal trajectory sequences, the trajectory deviation boundary is extracted as the first robotic arm motion boundary; by calculating the multiple acceleration averages of the K first spatiotemporal trajectory sequences at multiple displacement trajectory key points as the first motion feature; wherein, the first robotic arm motion boundary and the first motion feature constitute the first single-arm reference trajectory feature.

[0136] In one implementation, the parameter calling unit 3 is further configured to:

[0137] Historical coffee production records are backtracked based on the periodic detection window to obtain multiple production frequency characteristics of various coffee beverages; sample coffee beverages are reversely screened based on the multiple production frequency characteristics to locate a set of alternative test beverages; the set of alternative test beverages is used as a random call constraint, and the single robotic arm detection parameters and robotic arm collaborative detection parameters corresponding to the random test beverage production control information are called from the detection rule library, wherein the single robotic arm detection parameters are composed of a first robotic arm detection parameter, a second robotic arm detection parameter, and a third robotic arm detection parameter.

[0138] In one implementation, the single-arm detection unit 4 is further configured to:

[0139] In the process of using the first robotic arm detection parameter to drive the single-arm operation of the first device robotic arm, the first multi-dimensional sensor group is used to synchronously collect operation data to obtain a first real-time trajectory sequence; multiple first real-time accelerations of the first real-time trajectory sequence at the multiple displacement trajectory key points are calculated; similarly, a second real-time trajectory sequence is collected and obtained, and multiple second real-time accelerations are calculated and output based on the second real-time trajectory sequence; similarly, a third real-time trajectory sequence is collected and obtained, and multiple third real-time accelerations are calculated and output based on the third real-time trajectory sequence; the first real-time trajectory sequence, the second real-time trajectory sequence, the third real-time trajectory sequence, the multiple first real-time accelerations, the multiple second real-time accelerations and the multiple third real-time accelerations are structured and stored, and the single robotic arm trajectory feature is output.

[0140] In one implementation, the collaborative detection unit 5 is further configured to:

[0141] The first robotic arm detection parameters, the second robotic arm detection parameters, and the third robotic arm detection parameters in the synchronous execution state are used as the robotic arm collaborative detection parameters; in the process of collaboratively driving the first device robotic arm, the second device robotic arm, and the third device robotic arm using the robotic arm collaborative detection parameters, the first multidimensional sensor group, the second multidimensional sensor group, and the third multidimensional sensor group are used to synchronously collect operation data to obtain a third real-time trajectory sequence, a fourth real-time trajectory sequence, and a fifth real-time trajectory sequence; the robotic arm joint motion time series data is extracted from the third real-time trajectory sequence, the fourth real-time trajectory sequence, and the fifth real-time trajectory sequence, and the collaborative robotic arm trajectory characteristics are output.

[0142] In one implementation, the fault alarm unit 6 is further configured to:

[0143] If the Euclidean distance between the multiple first real-time accelerations and the mean of the multiple accelerations is greater than the preset acceleration deviation scale, and / or the first real-time trajectory sequence does not completely fall within the motion boundary of the first robotic arm, a first robotic arm alarm is output; similarly, a fault judgment is made on the second device robotic arm based on the second real-time trajectory sequence and the multiple second real-time accelerations, and a second robotic arm alarm is output; similarly, a fault judgment is made on the third device robotic arm based on the third real-time trajectory sequence and the multiple third real-time accelerations, and a third robotic arm alarm is output; based on the spatial deviation between the collaborative robotic arm trajectory feature and the benchmark collaborative space feature, a fourth robotic arm alarm is quantified and output; the first robotic arm alarm, the second robotic arm alarm, the third robotic arm alarm and the fourth robotic arm alarm are associated, and the robotic arm fault alarm is output.

[0144] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A coffee beverage robot arm fault detection method, characterized in that: include: The baseline robot arm trajectory characteristics were constructed by performing no-load running monitoring on the calibrated coffee beverage robot. A periodic detection window is preset, and a fault detection instruction is generated when the operating time limit of the coffee beverage robot meets the periodic detection window; The coffee beverage robot receives and randomly calls single robotic arm detection parameters and robotic arm collaborative detection parameters from a detection rule library according to the fault detection instruction; Using the single robotic arm detection parameters to control the coffee beverage robot to perform a single robotic arm special detection, and outputting a single robotic arm trajectory feature; Using the robotic arm collaborative detection parameters to control the coffee beverage robot to perform three-robotic arm collaborative detection, and output collaborative robotic arm trajectory features; A trajectory deviation analysis is performed on the single robot arm trajectory characteristics and the collaborative robot arm trajectory according to the reference robot arm trajectory characteristics, and a robot arm fault alarm is output.

2. The coffee beverage robot arm fault detection method according to claim 1, characterized in that: The baseline robot arm trajectory features were constructed by performing no-load running monitoring on the calibrated coffee beverage robot, including: A first multi-dimensional sensor group, a second multi-dimensional sensor group, and a third multi-dimensional sensor group are respectively installed on the first device robotic arm, the second device robotic arm, and the third device robotic arm of the coffee beverage robot; After calibrating the first, second, and third equipment robotic arms, performing no-load idling monitoring on the first, second, and third equipment robotic arms to drive the first, second, and third equipment robotic arms to simulate a standard coffee making process; During the process of the first device robotic arm, the second device robotic arm, and the third device robotic arm performing K rounds of no-load idling, the first multidimensional sensor group, the second multidimensional sensor group, and the third multidimensional sensor group synchronously collect operation data to obtain K first spatiotemporal trajectory sequences, K second spatiotemporal trajectory sequences, and K third spatiotemporal trajectory sequences, where K ≥ 20; Trajectory association features are collected for the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences, and the K third spatiotemporal trajectory sequences to obtain the reference manipulator trajectory features.

3. The coffee beverage robot arm fault detection method according to claim 2, characterized in that: Performing trajectory association feature collection on the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences, and the K third spatiotemporal trajectory sequences to obtain the reference manipulator trajectory feature includes: After spatially aligning the K first spatiotemporal trajectory sequences, outputting a first single-arm reference trajectory feature by performing spatiotemporal deviation analysis of the trajectory; Similarly, by performing trajectory spatiotemporal deviation analysis on the K second spatiotemporal trajectory sequences, a second single-arm reference trajectory feature is output; Similarly, by performing trajectory spatiotemporal deviation analysis on the K third spatiotemporal trajectory sequences, a third single-arm reference trajectory feature is output; After aligning the K first spatiotemporal trajectory sequences, the K second spatiotemporal trajectory sequences, and the K third spatiotemporal trajectory sequences by spatiotemporal mapping, performing a temporal mean extraction of the joint motion of the manipulator and outputting a benchmark collaborative space feature; Among them, the first single-arm reference trajectory feature, the second single-arm reference trajectory feature, the third single-arm reference trajectory feature and the reference collaborative space feature constitute the reference robotic arm trajectory feature.

4. The coffee beverage robot arm fault detection method according to claim 3, characterized in that: After spatially aligning the K first spatiotemporal trajectory sequences, a first single-arm reference trajectory feature is output by performing spatiotemporal trajectory deviation analysis, including: By performing trajectory deviation positioning on the K first spatiotemporal trajectory sequences, a trajectory deviation boundary is extracted as a first robotic arm motion boundary; Calculating multiple acceleration means of the K first spatiotemporal trajectory sequences at multiple displacement trajectory key points as first motion features; Among them, the first robotic arm motion boundary and the first motion feature constitute the first single-arm reference trajectory feature.

5. The coffee beverage robot arm fault detection method according to claim 4, characterized in that: The coffee beverage robot receives and randomly calls single robotic arm detection parameters and robotic arm collaborative detection parameters from a detection rule library according to the fault detection instruction, including: Backtracking historical coffee production records based on the periodic detection window to obtain multiple production frequency characteristics of multiple coffee beverages; Perform reverse screening of sample coffee beverages based on the multiple production frequency features to locate a set of candidate test beverages; The alternative test beverage set is used as a random call constraint, and the single robotic arm detection parameters and robotic arm collaborative detection parameters corresponding to the random test beverage production control information are called from the detection rule library, wherein the single robotic arm detection parameters are composed of the first robotic arm detection parameters, the second robotic arm detection parameters and the third robotic arm detection parameters.

6. The coffee beverage robot arm fault detection method according to claim 5, characterized in that: The single robotic arm detection parameters are used to control the coffee beverage robot to perform a single robotic arm special detection, and output a single robotic arm trajectory feature, including: In the process of driving the single-arm operation of the first device mechanical arm by using the first mechanical arm detection parameter, synchronously collecting operation data through the first multi-dimensional sensor group to obtain a first real-time trajectory sequence; Calculate a plurality of first real-time accelerations of the first real-time trajectory sequence at the plurality of displacement trajectory key points; Similarly, a second real-time trajectory sequence is acquired, and a plurality of second real-time accelerations are calculated and output based on the second real-time trajectory sequence; Similarly, a third real-time trajectory sequence is acquired, and a plurality of third real-time accelerations are calculated and output based on the third real-time trajectory sequence; The first real-time trajectory sequence, the second real-time trajectory sequence, the third real-time trajectory sequence, a plurality of first real-time accelerations, a plurality of second real-time accelerations, and a plurality of third real-time accelerations are structured and stored, and the single robotic arm trajectory feature is output.

7. The coffee beverage robot arm fault detection method according to claim 6, characterized in that: The robot arm collaborative detection parameters are used to control the coffee beverage robot to perform three-robot collaborative detection, and output collaborative robot arm trajectory features, including: Using the first robotic arm detection parameter, the second robotic arm detection parameter, and the third robotic arm detection parameter in a synchronous execution state as the robotic arm collaborative detection parameters; During the process of collaboratively driving the first device robotic arm, the second device robotic arm, and the third device robotic arm to operate using the robotic arm collaborative detection parameters, synchronously collecting operation data through the first multidimensional sensor group, the second multidimensional sensor group, and the third multidimensional sensor group to obtain a third real-time trajectory sequence, a fourth real-time trajectory sequence, and a fifth real-time trajectory sequence; The robot arm joint motion time series data is extracted from the third real-time trajectory sequence, the fourth real-time trajectory sequence, and the fifth real-time trajectory sequence, and the collaborative robot arm trajectory feature is output.

8. The coffee beverage robot arm fault detection method according to claim 7, characterized in that: Performing trajectory deviation analysis on the single robot arm trajectory characteristics and the collaborative robot arm trajectory based on the reference robot arm trajectory characteristics, and outputting a robot arm fault alarm, including: If the Euclidean distance between the multiple first real-time accelerations and the multiple acceleration means is greater than a preset acceleration deviation scale, and / or the first real-time trajectory sequence does not completely fall within the first robotic arm motion boundary, output a first robotic arm alarm; Similarly, a fault diagnosis is performed on the second device robotic arm based on the second real-time trajectory sequence and the plurality of second real-time accelerations, and a second robotic arm alarm is output; Similarly, a fault judgment is performed on the third device robotic arm based on the third real-time trajectory sequence and the plurality of third real-time accelerations, and a third robotic arm alarm is output; quantify and output a fourth robotic arm alarm based on a spatial deviation between the collaborative robotic arm trajectory feature and a reference collaborative spatial feature; The first robotic arm alarm, the second robotic arm alarm, the third robotic arm alarm and the fourth robotic arm alarm are associated, and the robotic arm failure alarm is output.

9. Coffee beverage robot arm fault detection system, characterized in that, The steps for implementing the method according to any one of claims 1 to 8 include: A trajectory construction unit is used to construct a baseline robot arm trajectory feature by performing no-load running monitoring on the calibrated coffee beverage robot; An instruction output unit, configured to preset a periodic detection window and generate a fault detection instruction when the operating time limit of the coffee beverage robot meets the periodic detection window; A parameter calling unit, configured for the coffee beverage robot to receive and randomly call single robotic arm detection parameters and robotic arm collaborative detection parameters from a detection rule library according to the fault detection instruction; a single-arm detection unit, configured to control the coffee beverage robot to perform a single-arm special detection using the single-arm detection parameters and output a single-arm trajectory feature; a collaborative detection unit, configured to control the coffee beverage robot to perform three-manipulator collaborative detection using the robotic arm collaborative detection parameters, and output collaborative robotic arm trajectory features; A fault alarm unit is used to perform trajectory deviation analysis on the single robot arm trajectory characteristics and the collaborative robot arm trajectory based on the reference robot arm trajectory characteristics, and output a robot arm fault alarm.

Images