Power battery disassembling method and related equipment

Abstract

The invention discloses a power battery disassembling method and related equipment, and relates to the technical field of power batteries, and the method comprises the following steps: obtaining multi-mode sensing data of a to-be-disassembled battery pack; performing risk assessment on the multi-modal sensing data based on a preset safety threshold, and generating a risk level matrix; based on the multi-modal sensing data and the risk level matrix, analyzing and processing the battery pack topological relation of the to-be-disassembled battery pack through a preset adaptive decision model to obtain a disassembling path planning scheme; and based on the disassembling force parameter and the disassembling path priority, controlling the heterogeneous robot to collaboratively execute the disassembling operation on the to-be-disassembled battery pack by an execution unit. Through multi-mode sensing, risk assessment, self-adaptive path planning and heterogeneous robot collaborative operation, high safety, high efficiency and high precision in the power battery pack disassembling process are achieved, and the material recycling value is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power batteries, and more particularly, relates to a power battery disassembly method and related equipment. BACKGROUND

[0002] With the rapid development of the new energy automobile industry, power batteries, as the core energy unit, need to be safely and efficiently disassembled and recycled after the end of their service life to realize resource reuse and reduce environmental risks. The power battery contains various chemical components and precise structures, and when it is retired, it may have various potential risks such as residual electricity, gas leakage, and structural deformation. Therefore, safety, efficiency, and precision are particularly important during the disassembly process. Power battery disassembly not only relates to the economic benefits of recycling and utilization, but also directly affects the personal safety of operating personnel and the level of environmental protection.

[0003] In related technologies, power battery disassembly relies on manual operation or semi-automatic mechanical equipment. Manual disassembly is flexible, but when facing complex structures, high-voltage electrical environments, and hazardous gas leakage risks, it has problems such as high safety risks and low efficiency. Semi-automatic equipment can improve some efficiency, but lacks comprehensive perception and risk assessment of the battery pack state before disassembly, making it difficult to develop precise disassembly strategies for different structures and different risk levels. That is, the related technologies have the technical problems of high safety risk of power battery disassembly, single path planning, and insufficient disassembly precision and efficiency. SUMMARY

[0004] A series of simplified concepts are introduced in the summary part of the present application, which will be further described in detail in the specific embodiment part. The summary part of the present application does not mean to try to limit the key features and necessary technical features of the claimed technical solution, and even less means to determine the protection scope of the claimed technical solution.

[0005] The power battery disassembly method and related equipment provided by the present application can realize high safety, high efficiency, and high precision in the power battery pack disassembly process through multi-modal perception, risk assessment, adaptive path planning, and heterogeneous robot collaborative work, and improve the material recycling value.

[0006] In a first aspect, a method for disassembling a power battery is provided. The method includes: obtaining multi-modal perception data of a battery pack to be disassembled, wherein the multi-modal perception data includes 3D visual point cloud data, real-time voltage detection data, and gas sensing data; performing risk assessment on the multi-modal perception data based on a preset safety threshold to generate a risk level matrix; analyzing and processing a battery pack topology relationship of the battery pack to be disassembled based on the multi-modal perception data and the risk level matrix through a preset adaptive decision model to obtain a disassembly path planning scheme, wherein the disassembly path planning scheme includes a disassembly strength parameter and a disassembly path priority; and controlling a heterogeneous robot cooperative execution unit to perform a disassembly operation on the battery pack to be disassembled based on the disassembly strength parameter and the disassembly path priority, wherein the heterogeneous robot cooperative execution unit includes at least one collaborative mechanical arm and one precision cutting robot.

[0007] In some embodiments, the obtaining multi-modal perception data of a battery pack to be disassembled includes: performing scanning processing on the battery pack to be disassembled by a three-dimensional laser scanner to obtain the 3D visual point cloud data; performing voltage collection on the battery pack to be disassembled by a distributed voltage sensor array to obtain the real-time voltage detection data; and performing gas detection on the battery pack to be disassembled by a multi-channel gas mass spectrometer to obtain the gas sensing data, wherein a detection range of the multi-channel gas mass spectrometer includes electrolyte volatiles and thermal runaway products.

[0008] In some embodiments, the performing risk assessment on the multi-modal perception data based on a preset safety threshold to generate a risk level matrix includes: generating a voltage risk matrix based on a comparison result of the real-time voltage detection data and a voltage safety threshold in combination with the 3D visual point cloud data; generating a chemical risk matrix based on a comparison result of the gas sensing data and a gas concentration threshold in combination with the 3D visual point cloud data; identifying a structural weak area based on the 3D visual point cloud data to obtain a structural risk matrix; and fusing the voltage risk matrix, the chemical risk matrix, and the structural risk matrix to obtain the risk level matrix.

[0009] In some embodiments, the fusing the voltage risk matrix, the chemical risk matrix, and the structural risk matrix to obtain the risk level matrix includes: registering the voltage risk matrix, the chemical risk matrix, and the structural risk matrix based on a spatial position weight to obtain a voltage registration matrix, a chemical registration matrix, and a structural registration matrix, wherein a weight coefficient of a core area of the spatial position weight is higher than a weight coefficient of an edge area; and performing multi-modal risk coupling processing based on the voltage risk matrix, the chemical risk matrix, and the structural risk matrix to obtain the risk level matrix.

[0010] In some embodiments, the multi-modal risk coupling processing based on the voltage risk matrix, the chemical risk matrix and the structure risk matrix to obtain the risk level matrix comprises: based on the voltage risk matrix, the chemical risk matrix and the structure risk matrix, the multi-modal risk coupling processing is performed by the following formula to obtain the risk level matrix:

[0011] Y = a A x b B + g C2

[0012] In the formula, Y is the risk level matrix, A is the voltage risk matrix, B is the chemical risk matrix, C is the structure risk matrix, a is the first adjustment coefficient, b is the second adjustment coefficient, and g is the third adjustment coefficient.

[0013] In some embodiments, the analysis and processing of the battery pack topology relationship of the battery pack to be disassembled based on the multi-modal perception data and the risk level matrix by a preset adaptive decision model to obtain a disassembly path planning scheme comprises: based on the 3D visual point cloud data, a battery pack module connection topology graph is constructed, wherein the nodes of the battery pack module connection topology graph represent battery modules, and the edges represent the connection relationship strength between the battery modules; the risk level matrix is mapped to the battery pack module connection topology graph to obtain a risk weight topology graph; based on a preset risk propagation algorithm, a minimum risk propagation path from a starting disassembly point to a target disassembly point is determined in the risk weight topology graph, wherein the risk propagation algorithm simulates the propagation process of risk in the topology graph, dynamically updates the node risk value, and backtracks to generate a disassembly path; based on the connection relationship strength on the minimum risk propagation path, the disassembly force parameter is determined; based on the risk degree of each path in the risk weight topology graph, the disassembly path priority is determined.

[0014] In some embodiments, the control of the heterogeneous robot cooperative execution unit to perform disassembly operation on the battery pack to be disassembled based on the disassembly force parameter and the disassembly path priority comprises: according to the disassembly path priority, the disassembly operation sequence is decomposed into a grasping task and a cutting task, and the grasping task is assigned to the collaborative manipulator and the cutting task is assigned to the precision cutting robot; based on the disassembly force parameter, the grasping force of the collaborative manipulator and the cutting force of the precision cutting robot are set; the task execution order is determined according to the disassembly path priority, and the operation timing of the collaborative manipulator and the precision cutting robot is synchronized through a distributed cooperative controller; during the execution of the disassembly operation, the multi-modal perception data is monitored in real time, and when the monitoring data in the multi-modal perception data exceeds a preset safety threshold, a safety interruption mechanism is triggered.

[0015] In a second aspect, the application further provides a power battery disassembly device, comprising: a data acquisition unit configured to acquire multi-modal perception data of a battery pack to be disassembled, wherein the multi-modal perception data comprises 3D visual point cloud data, real-time voltage detection data and gas sensing data; a risk assessment unit configured to perform risk assessment on the multi-modal perception data based on a preset safety threshold to generate a risk level matrix; a disassembly planning unit configured to analyze and process a battery pack topology relationship of the battery pack to be disassembled based on the multi-modal perception data and the risk level matrix through a preset adaptive decision model to obtain a disassembly path planning scheme, wherein the disassembly path planning scheme comprises a disassembly strength parameter and a disassembly path priority; and a battery disassembly unit configured to control a heterogeneous robot cooperative execution unit to perform a disassembly operation on the battery pack to be disassembled based on the disassembly strength parameter and the disassembly path priority, wherein the heterogeneous robot cooperative execution unit comprises at least one cooperative mechanical arm and one precision cutting robot.

[0016] In a third aspect, the application further provides an electronic device, comprising a memory and a processor, wherein the processor is configured to implement the steps of the power battery disassembly method of the first aspect when executing a computer program stored in the memory.

[0017] In a fourth aspect, the application further provides a computer-readable storage medium storing a computer program, wherein the computer program is configured to implement the steps of the power battery disassembly method of the first aspect when executed by a processor.

[0018] In a fifth aspect, the application further provides a computer program product comprising a computer program or computer executable instructions, wherein the computer program or computer executable instructions are configured to implement the power battery disassembly method provided by the embodiments of the application when executed by a processor.

[0019] In summary, the application realizes all-round monitoring of the appearance structure, electrical state and internal gas environment of the battery pack to be disassembled by collecting multi-modal perception data including 3D vision point cloud data, real-time voltage detection data and gas sensing data, can discover potential risks such as deformation, short circuit and leakage in time before and during disassembly, and can improve the safety of disassembly operation from the source; risk assessment is performed on the multi-modal perception data based on a preset safety threshold to generate a risk level matrix, which can finely classify the safety of different parts of the battery pack, can avoid blind operation on high-risk areas in traditional disassembly, and thus effectively reduces the probability of safety accidents; the adaptive decision model is used to analyze the topological relationship of the battery pack by combining the multi-modal perception data and the risk level matrix, and generate a disassembly path planning scheme including disassembly force parameters and disassembly path priority, which realizes the intelligentization and individualization of the disassembly process and significantly improves the disassembly efficiency; through dynamic adjustment of the disassembly force parameters, appropriate force can be selected for parts with different risk levels and structural characteristics, which can not only ensure smooth disassembly, but also reduce secondary damage to the battery monomer and the surrounding structure, and can improve the recycling value of battery materials; the heterogeneous robot cooperative execution unit is composed of at least one collaborative manipulator and one precision cutting robot, which can realize cooperative operation of tasks such as grabbing and carrying, positioning and fixing, and accurate cutting, which not only can improve the work efficiency, but also can reduce the demand for manual operation in high-risk environment. In summary, the power battery disassembly method provided by the application realizes high safety, high efficiency and high precision in the disassembly process of the power battery pack through multi-modal perception, risk assessment, adaptive path planning and heterogeneous robot cooperative operation, and improves the material recycling value. BRIEF DESCRIPTION OF DRAWINGS

[0020] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of preferred embodiments, and are not meant to limit the present description. Moreover, the same reference numerals are used throughout the various drawings to designate identical elements. In the drawings:

[0021] Figure 1 A flowchart of a power battery disassembly method provided by an embodiment of the application;

[0022] Figure 2 A schematic diagram of the composition structure of a power battery disassembly device provided by an embodiment of the application;

[0023] Figure 3 A schematic diagram of the composition structure of an electronic device provided by an embodiment of the application. DETAILED DESCRIPTION

[0024] The terms “first”, “second”, “third”, “fourth” etc. (if any) in the specification, claims and drawings of the present application are used to distinguish similar objects, not to describe a particular sequential or chronological order. Therefore, it is understood that these terms can be used interchangeably, as appropriate, to describe the embodiments described, unless a specific requirement is otherwise required by the illustration or description. In addition, the terms “is” and “has” and any variants thereof in the present application are intended to cover non-exclusive inclusion of all possible constituent elements. For example, a process, method, system, product or device including several steps or units does not necessarily limit to only the steps or units explicitly listed, but can also include other steps or units not explicitly listed, or steps or units inherent to the process, method, product or device.

[0025] In the present application, “module” or “unit” refers to a computer program or a part of a computer program with a specific function, and works in cooperation with other related parts to achieve a predetermined target. These modules or units can be implemented by software, hardware (such as processing circuitry or memory) or a combination of both. One or more processors or memories can implement one or more modules or units. At the same time, each module or unit can also be part of a larger module or unit.

[0026] The technical solutions in the present application will be described in detail below in conjunction with the drawings in the embodiments. It should be noted that the described embodiments are only a part of the present application, not all embodiments. In the following description, “some embodiments” mentioned is only a subset of all possible embodiments, which can be the same or different subset, and different embodiments can be combined with each other without conflict.

[0027] Figure 1 is a flowchart of a power battery disassembly method provided by an embodiment of the present application. For example, referring to Figure 1 The power battery disassembly method provided by the embodiment of the present application can include the following steps 101 to 104:

[0028] Step 101, acquiring multi-modal perception data of a battery pack to be disassembled, wherein the multi-modal perception data can include 3D visual point cloud data, real-time voltage detection data and gas sensing data;

[0029] In some examples, the battery pack to be disassembled is a power battery pack reaching the retirement standard or requiring recycling, including but not limited to new energy vehicle retired battery pack, energy storage system scrapped battery pack, etc., and its typical structure includes a plurality of cell modules, a metal shell, a high-voltage connector, a cooling pipeline and other components, and there may be potential risk states such as cell bulging, electrolyte leakage, excessive residual capacity, etc. The multi-modal perception data is a heterogeneous data set collected by multiple perception means, reflecting the physical state and potential risks of the battery pack to be disassembled. Through multi-dimensional information fusion, a comprehensive evaluation of the battery pack can be achieved, which can overcome the limitations of a single data type in risk identification. The 3D vision point cloud data is a three-dimensional coordinate data set generated by scanning the outer surface and internal structure of the battery pack by a three-dimensional laser scanning device, which can accurately present the three-dimensional profile, spatial position and geometric size (such as screw hole position, module splicing seam, shell deformation, etc.) of the battery pack; a three-dimensional laser scanner with an accuracy of not less than 0.1 mm can be used to scan from the top, side and other three or more perspectives after the battery pack is fixed on the disassembly station, and the scanning range covers the whole battery pack and key connection parts (such as module locking points and high-voltage interfaces). After denoising and splicing processing, the scanning data generates a point cloud model, and the 3D vision point cloud data can clearly identify screw heads with a diameter of ≥3 mm and gap structures with a depth of ≥5 mm, providing a spatial coordinate reference for subsequent topological relationship analysis. The real-time voltage detection data is the real-time voltage value of the battery pack internal cell or module collected by the voltage sensor array, which is used to monitor the residual capacity and potential short circuit risk; distributed voltage sensors can be deployed at the positive and negative electrode interfaces of the battery pack and the end of each module. The sensors are connected to the electrode contacts through probes or pens, the sampling frequency is set to 1 kHz, and the data is transmitted to the data acquisition card through shielded cables, and the voltage value of each monitoring point is output in real time; for example, when the voltage of a certain cell drops to 0V, the internal short circuit risk can be identified instantaneously; when the overall voltage is > 36V, the high-voltage early warning mechanism is triggered. The gas sensing data is the volatile gas composition and concentration data around the battery pack collected by the gas detection device, mainly including electrolyte volatiles (such as lithium hexafluorophosphate, dimethyl carbonate) and thermal runaway characteristic gases (such as CO, H2, HF); more than three multi-channel gas mass spectrometers can be arranged around the battery pack, the detection probe is 30-50 cm away from the surface of the battery pack, the detection range covers the molecular weight interval of 1-500 amu, and the sampling interval is 1 second; for example, when the concentration of lithium hexafluorophosphate exceeds 5 ppm, it can be determined that there is a slight electrolyte leakage; when the CO concentration suddenly rises to 100 ppm, the thermal runaway early warning is triggered.

[0030] In the implementation process, after the battery pack to be disassembled is conveyed to the disassembly station by the conveying belt, the posture can be fixed by the mechanical positioning device; then the control system synchronously starts the three-dimensional laser scanner, the voltage sensor array and the gas mass spectrometer, wherein the laser scanner completes 360° scanning along the preset track, the voltage sensor continuously collects voltage data during scanning, and the gas mass spectrometer monitors and records the change of gas composition in real time; after all data are aligned by time stamp, they are packaged and transmitted to the edge computing unit to form a multi-modal data set containing spatial structure, electrical state and chemical risk, thereby providing original input for subsequent risk level matrix construction.

[0031] Through the implementation of step 101, multi-modal perception data including 3D visual point cloud data, real-time voltage detection data and gas sensing data are obtained, which can realize all-round and multi-dimensional real-time monitoring of the appearance structure, electrical state and internal gas environment of the battery pack to be disassembled, thereby identifying potential risks such as deformation, short circuit and gas leakage in time before and during disassembly, and providing accurate basic data support for subsequent safe and efficient disassembly operation.

[0032] In step 102, the multi-modal perception data is evaluated based on a preset safety threshold to generate a risk level matrix.

[0033] In some examples, the preset safety threshold is a multidimensional safety threshold value preset based on the power battery disassembly safety standard, historical failure data and material characteristics, and is used to determine whether the multi-modal perception data exceeds the safety range; the preset safety threshold can be determined after calibration by an expert system in combination with statistical analysis of 1000+ groups of retired battery disassembly failure cases; the preset safety threshold can specifically include: a voltage safety threshold (such as a single cell voltage ≤ 0V to determine a short circuit risk, and a total voltage > 36V to trigger a high voltage early warning), a gas concentration threshold (such as lithium hexafluorophosphate ≥ 5ppm for a leakage early warning, and CO ≥ 50ppm for a first-level thermal runaway early warning), and a structure parameter threshold (such as a shell deformation amount ≥ 5mm for a structure risk, and a screw loosening displacement ≥ 2mm for a connection failure risk). The process of risk assessment is a process of comparing multi-modal perception data with a preset safety threshold, identifying risk types and quantifying risk levels in dimensions; specifically, real-time voltage detection data can be determined according to the threshold value amplitude to determine the risk level (such as voltage 36-50V for a medium risk, and > 50V for a high risk); gas sensing data can be graded in combination with the concentration increase rate and the threshold value ratio (such as a concentration of 1-1.5 times the threshold value for a medium risk, and > 1.5 times the threshold value for a high risk); 3D visual point cloud data can be rated according to the structure parameter out-of-tolerance rate (such as a deformation amount of 80%-100% of the threshold value for a medium risk, and > 100% of the threshold value for a high risk). Example: when the voltage of a certain cell drops to -0.5V (below the 0V threshold) and the CO concentration in the corresponding area reaches 80ppm (above the 50ppm threshold), it is determined as a composite high-level risk of high-voltage short circuit and thermal runaway. The risk level matrix is a two-dimensional matrix with battery pack space partition as the horizontal axis and risk level as the vertical axis, which is used to intuitively present the risk distribution of different areas; each cell in the risk level matrix can contain a three-tuple composed of area coordinates, risk level and risk type, and the risk level is divided into four levels: low (1st level), medium (2nd level), high (3rd level) and extremely high (4th level); the generation logic of the risk level matrix is: the battery pack is divided into areas according to a 10cm×10cm grid, the voltage, gas and structure data of each area are matched based on the spatial coordinates of the 3D visual point cloud data, the comprehensive risk value (1-4) is calculated by a multidimensional risk coupling algorithm, and finally the matrix is formed.

[0034] In the implementation process, the edge computing unit can first perform timestamp alignment (error ≤ 10 ms) on the multi-modal data obtained in step 101, and then call the preset safety threshold database for regional comparison: for voltage data, the distributed sensor data is mapped to each grid area by an interpolation algorithm to generate a voltage risk sub-matrix; for gas data, a Gaussian diffusion model is used to calculate the gas concentration in each area to generate a chemical risk sub-matrix; for 3D visual data, the deformation amount in each area is calculated by a structure deviation detection algorithm to generate a structure risk sub-matrix; then, the three sub-matrices are combined into a global risk level matrix by a weighted fusion algorithm (voltage weight 0.4, chemical weight 0.35, and structure weight 0.25), the entire process takes ≤ 2 seconds, and the final output matrix can directly display key information such as "the upper right corner area (X: 80-100 cm, Y: 0-20 cm) of the battery pack is a 4-level high risk" to provide a risk coordinate benchmark for subsequent path planning.

[0035] Through the implementation of step 102, the risk assessment of multi-modal perception data based on the preset safety threshold and the generation of the risk level matrix can finely classify the safety risks of different parts of the battery pack, thereby realizing differentiated disassembly strategies for different risk levels, effectively avoiding blind operation in high-risk areas, reducing the probability of safety accidents, and improving the controllability of the operation.

[0036] In step 103, based on the multi-modal perception data and the risk level matrix, the battery pack topology relationship of the battery pack to be disassembled is analyzed and processed by a preset adaptive decision model to obtain a disassembly path planning scheme, wherein the disassembly path planning scheme can include a disassembly intensity parameter and a disassembly path priority.

[0037] In some examples, the preset adaptive decision model is an intelligent decision system constructed based on the fusion of deep learning and reinforcement learning, which is used to dynamically analyze the battery pack topology relationship and generate an optimal disassembly path; the preset adaptive decision model includes three core modules:

[0038] Topology analysis module: a graph neural network (GNN) architecture can be used, which contains 3 layers of hidden layers (64 neurons per layer), the input is the topology graph features generated from 3D visual point cloud data, and the output is a component connection strength matrix;

[0039] Path search module: based on improved A* algorithm and deep Q network (DQN), wherein the DQN contains an experience replay pool (capacity 100,000) and a target network (update period 500 steps), and the risk level matrix is used as the cost function weight;

[0040] Parameter optimization module: a random forest regression model (100 decision trees) can be used, the input is the connection strength and the risk level, and the output is the disassembly intensity parameter.

[0041] The training process combines supervised learning and reinforcement learning:

[0042] The training data includes 5000+ disassembly cases of different battery pack models, including 3D point cloud, risk matrix, and optimal path label.

[0043] In the supervised phase, the path deviation rate (Euclidean distance between actual path and labeled path) is used as the loss function, the learning rate is 0.001, and the iteration is 200 rounds.

[0044] In the reinforcement phase, the disassembly efficiency (completion time) and safety index (risk trigger frequency) are used as the reward function, the ε-greedy strategy (ε decays from 0.9 to 0.1), and the training is 100,000 steps.

[0045] For example, when the preset adaptive decision model analyzes a certain CTP battery pack, the topology analysis module can identify the series connection relationship of 28 modules within 0.5 seconds, and the path search module preferentially avoids risk areas above level 3.

[0046] The battery pack topology relationship is the spatial connection structure and logical association of each component inside the battery pack to be disassembled, including physical connection (such as screw locking, adhesive bonding) and electrical connection (such as Busbar welding, cable insertion); the battery pack topology relationship can be based on 3D visual point cloud data to segment components (using MaskR-CNN algorithm, mIOU≥0.92), extract the spatial coordinates of key components such as modules, shells, connectors, and screws, and construct a topology graph through an undirected graph model, with nodes representing components (such as "Module A" and "Screw B") and edge weights representing connection strength (physical connection in torque threshold kg·cm, electrical connection in contact resistance mΩ); for example, in the topology graph of a certain ternary lithium battery pack, the "shell" node is connected to the "module 1-4" node through an edge weight of "5 kg·cm" (indicating that a torque of 5 kg·cm is required for disassembly), and the "module 1" is connected to the "high-voltage connector" through an edge weight of "2 mΩ" (contact resistance 2 mΩ). The disassembly path planning scheme is a structured disassembly execution plan generated based on the topology relationship and risk level, including disassembly sequence, tools used, operation area, force parameter, and emergency plan. The disassembly force parameter is a quantitative value of the force or torque applied to the component during disassembly, used to balance disassembly efficiency and component protection, to avoid excessive force causing cell damage or insufficient force causing disassembly stagnation; the disassembly force parameter can be calculated according to the connection strength in the topology relationship (such as screw torque threshold) and the risk level matrix (such as a 20% reduction in force for high-risk areas), the formula is: force parameter = basic force × (1-risk level coefficient), where the basic force comes from the component manual (such as M5 screw basic torque 2.5 N·m), and the risk level coefficient (such as 1 level 0.05, 2 level 0.1, 3 level 0.2, 4 level 0.3). Example: for a M5 screw in a risk level 2 area, the force parameter = 2.5 × (1-0.1) = 2.25 N·m; for a module connected by a glue layer, the cutting force = basic cutting force 80 N × (1-0.3) = 56 N (due to risk level 4). The disassembly path priority refers to the execution order sorting rule of multiple potential disassembly paths, and the comprehensive score of each path can be calculated (risk value × 0.4 + time consumption × 0.3 + component damage probability × 0.3), and arranged in ascending order of score; for example, there are 3 paths for a certain battery pack, path A (risk level 1, time consumption 80 s, damage rate 5%) score = 1 × 0.4 + 80 × 0.3 + 5 × 0.3 = 25.9; path B (risk level 3, time consumption 60 s, damage rate 10%) score = 3 × 0.4 + 60 × 0.3 + 10 × 0.3 = 22.2; path C (risk level 2, time consumption 70 s, damage rate 8%) score = 2 × 0.4 + 70 × 0.3 + 8 × 0.3 = 24.2; the final priority is path B > path C > path A.

[0047] In the implementation process, the multi-modal perception data can be pre-processed first; the 3D visual point cloud data is down-sampled (key feature points are retained) and coordinate-normalized (with the geometric center of the battery pack as the origin), the risk level matrix is converted into a grid risk value array (1-4 levels correspond to numerical values 1-4); then input into the preset adaptive decision model, the topological analysis module generates a topological graph containing 200+ nodes based on GNN (edge weight accuracy ± 0.1 kg·cm), the path search module performs path traversal with risk value as the cost (search step length 5 cm), and the parameter optimization module outputs 8-12 groups of force parameters in combination with the connection strength and risk level; the model reasoning time is ≤3 seconds, and the final output disassembly path planning scheme includes: preferentially disassembling the low-risk area at the lower left corner of the battery pack (X: 0-30 cm, Y: 50-80 cm), using a collaborative robot to disassemble the M6 screw with a torque of 1.8 N·m, and the subsequent path is executed in the order of increasing risk level, the scheme synchronously generates tool switching instructions (such as switching from an electric wrench to a laser cutting head) and emergency force adjustment thresholds (such as automatically reducing the force by 30% when the risk level suddenly rises).

[0048] Through the implementation of step 103, the adaptive decision model is used to analyze and process the topological relationship of the battery pack in combination with the multi-modal perception data and the risk level matrix, and a disassembly path planning scheme containing disassembly force parameters and disassembly path priority is generated, which can dynamically optimize the disassembly sequence and operation mode according to the actual structure and risk distribution of the battery pack, thereby significantly improving the intelligence, pertinence and execution efficiency of the disassembly process.

[0049] Step 104, based on the disassembly force parameters and the disassembly path priority, controlling the heterogeneous robot cooperative execution unit to execute the disassembly operation on the battery pack to be disassembled, wherein the heterogeneous robot cooperative execution unit can include at least one collaborative robot and one precision cutting robot;

[0050] In some examples, the heterogeneous robot cooperative execution unit is a cooperative work system composed of different types of robots with complementary functions, which realizes task division and action synchronization through a distributed control architecture, and is used to complete the composite disassembly operation of the power battery pack. The disassembly operation is a set of physical separation actions performed according to a disassembly path planning scheme, including but not limited to screw disassembly, shell separation, module disassembly, electrode cutting, etc., and needs to meet three requirements of no excessive damage to the battery cell, no secondary pollution and operation timing matching; the specific type in the disassembly operation process can be determined according to the battery pack structure, for the shell connected by bolts, a torque controllable loosening operation is performed; for the glue-sealed module, a hot knife assisted peeling operation is performed; for the laser welded Busbar, a precision controllable cutting operation is performed. The cooperative robot arm is a multi-joint robot with force control feedback and human-machine cooperation function, mainly responsible for high-precision grabbing, carrying and low-load disassembly actions, and its end effector can be quickly replaced, such as electric screwdriver, vacuum suction cup, gripper. The precision cutting robot is a special robot focusing on material separation, integrating high-precision cutting tools (laser head / milling cutter) and visual positioning system, used for processing high-strength connection structures such as welding and riveting in the battery pack; the cutting precision of the precision cutting robot needs to reach ±0.05mm, the laser cutting power can be continuously adjusted at 50-300W (for different thickness metals), the milling speed can reach 10000rpm, and it has a spark collection and smoke purification device.

[0051] In the specific implementation process, the control system can first parse the disassembly path planning scheme: assign the "shell screw disassembly" task of priority 1 to the cooperative robot arm, configure the electric screwdriver torque parameter (3.5N·m, rotation speed 300rpm); assign the "tab cutting" task of priority 2 to the precision cutting robot, set the laser power to 150W and the cutting speed to 50mm / s. Through the ROS2 distributed cooperative controller, the action timing error of the robot arm and the cutting robot is controlled within ±100ms, and the cutting robot starts tab cutting only after the robot arm completes the first group of screw disassembly and moves to the safe area. During execution, the 3D vision sensor monitors the tool position in real time (sampling frequency 100Hz), and if the robot arm screw torque exceeds the ±0.3N·m deviation range or the cutting area gas sensor detects that the HF concentration is ≥1ppm, a safety interrupt is triggered immediately, and all robots stop moving and reset to the safe position within 0.5 seconds.

[0052] Through the implementation of step 104, based on the foregoing disassembly force parameters and disassembly path priorities, the heterogeneous robot cooperative execution unit composed of cooperative robot arms and precision cutting robots is controlled to disassemble, which can realize efficient cooperation of various work tasks such as grabbing and carrying, positioning and fixing, and precise cutting while ensuring safety, not only improving the speed and precision of disassembly work, but also reducing the demand for human work in high-risk environments.

[0053] In summary, the embodiment of the present application realizes all-round monitoring of the appearance structure, electrical state and internal gas environment of the battery pack to be disassembled by collecting multi-modal perception data including 3D vision point cloud data, real-time voltage detection data and gas sensing data, can discover potential risks such as deformation, short circuit and leakage in time before and during disassembly, and can improve the safety of disassembly operation from the source; risk assessment is performed on the multi-modal perception data based on a preset safety threshold to generate a risk level matrix, which can finely classify the safety of different parts of the battery pack, can avoid blind operation on high-risk areas in traditional disassembly, and thus effectively reduces the probability of safety accidents; the adaptive decision-making model is used to analyze the topological relationship of the battery pack by combining the multi-modal perception data and the risk level matrix, and generate a disassembly path planning scheme including disassembly force parameters and disassembly path priority, which realizes the intelligentization and individualization of the disassembly process and significantly improves the disassembly efficiency; by dynamically adjusting the disassembly force parameters, appropriate force can be selected for parts with different risk levels and structural characteristics, which can not only ensure smooth disassembly, but also reduce secondary damage to battery monomers and surrounding structures, and can improve the recycling value of battery materials; the heterogeneous robot cooperative execution unit is composed of at least one collaborative manipulator and one precision cutting robot, which can realize cooperative operation of tasks such as grabbing and carrying, positioning and fixing, and accurate cutting, which not only can improve the work efficiency, but also can reduce the demand for manual operation in high-risk environment. In summary, the power battery disassembly method provided by the embodiment of the present application realizes high safety, high efficiency and high precision in the disassembly process of the power battery pack by multi-modal perception, risk assessment, adaptive path planning and heterogeneous robot cooperative operation, and improves the recycling value of materials.

[0054] In some embodiments, the foregoing step 101 can include: scanning the battery pack to be disassembled by a three-dimensional laser scanner to obtain 3D vision point cloud data; collecting voltage of the battery pack to be disassembled by a distributed voltage sensor array to obtain real-time voltage detection data; and detecting gas of the battery pack to be disassembled by a multi-channel gas mass spectrometer to obtain gas sensing data, wherein the detection range of the multi-channel gas mass spectrometer can include electrolyte volatiles and thermal runaway products.

[0055] In some examples, the three-dimensional laser scanner is a high-precision measurement device that acquires three-dimensional coordinate information of an object surface by emitting a laser beam, and is used to construct a three-dimensional structure model of the battery pack to be disassembled. The selection needs to meet the scanning accuracy ≤ 0.1 mm, the point cloud density ≥ 100 points / mm2, the scanning rate ≥ 100 million points / second, and support 360° omnidirectional scanning. The scanning process is the whole process of data acquisition and preprocessing of the three-dimensional laser scanner on the battery pack, including scanning path planning, point cloud data generation, denoising and splicing, etc. First, the scanner can be carried by an industrial robot to move along the preset trajectory, and the scanning time of each surface is ≥ 2 seconds to ensure data integrity; then the original point cloud data is executed Gaussian filtering (σ = 0.5) to remove noise points, and multi-view point cloud splicing is completed through ICP algorithm (iteration times ≤ 50 times), and finally a global point cloud model is generated. The distributed voltage sensor array is a detection network formed by arranging multiple voltage sensors according to a spatial distribution rule, which is used to synchronously collect voltage signals at different positions of the battery pack, and needs to meet the measurement range 0-1000V, the accuracy ±0.2%FS, the sampling rate ≥ 1kHz, and support expansion (8-32 sensors can be integrated in a single array) through modular design. Voltage acquisition is the real-time detection process of the sensor array on the electrical parameters of the battery pack, including signal acquisition, amplification, AD conversion and data transmission. The multi-channel gas mass spectrometer is a high-precision analysis instrument that can simultaneously detect multiple gas components, which identifies the mass of gas molecules through ionization technology to realize qualitative and quantitative analysis of volatile gases around the battery pack. Gas detection is the sampling and analysis process of the multi-channel gas mass spectrometer on the gas samples around the battery pack, including sampling, ionization, mass separation and signal detection. Electrolyte volatile matter is an organic compound and electrolyte composition generated by the volatilization of power battery electrolyte due to sealing failure or temperature rise, mainly including carbonate solvents (such as dimethyl carbonate, methyl ethyl carbonate) and lithium salt decomposition products (such as lithium hexafluorophosphate). Thermal runaway product is a characteristic gas generated by thermal runaway reaction of battery cells due to internal short circuit, overcharge, etc., mainly including carbon monoxide (CO), hydrogen (H2), hydrogen fluoride (HF), methane (CH4), etc.

[0056] In the implementation process, after the to-be-disassembled battery pack is fixed by the positioning tool (repeated positioning accuracy ±0.5 mm), the control system synchronously starts the three-dimensional laser scanner, the distributed voltage sensor array and the multi-channel gas mass spectrometer: the scanner completes omnidirectional scanning along the preset 5 tracks, and the generated point cloud data is compared with the CAD model after preprocessing, and the structural deviation area is marked; the sensor array collects voltage data of 32 monitoring points at a frequency of 1 kHz, and removes high-frequency noise through sliding window filtering (window size 100 ms); three gas sampling probes monitor and record the concentrations of 15 characteristic gases in real time, and when the concentration of lithium hexafluorophosphate is detected to be ≥3 ppm, the scanning time of the scanner on the corresponding area is automatically extended to 5 seconds. After all the data are aligned by timestamp (accuracy 1 ms), they are packaged and stored as a multi-modal data set containing spatial coordinates, voltage values and gas concentrations, providing original input for subsequent risk assessment.

[0057] Through the implementation of the above embodiments, the collection method of multi-modal data is determined, the three-dimensional laser scanner can provide millimeter-level structural details and solve the misjudgment problem of traditional visual recognition on complex battery pack structures; the distributed voltage sensor array realizes voltage monitoring at the cell level and can avoid the omission of local short-circuit risk by overall voltage detection; the multi-channel gas mass spectrometer accurately captures electrolyte volatiles and thermal runaway products, provides early warning of potential explosion risk, and thus can improve the accuracy and timeliness of data, providing reliable input for subsequent risk assessment and path planning, which not only strengthens the timeliness of safety control, but also lays a data foundation for precise disassembly.

[0058] In some embodiments, the foregoing step 102 can include: generating a voltage risk matrix based on the comparison result of real-time voltage detection data and a voltage safety threshold, in combination with 3D visual point cloud data; generating a chemical risk matrix based on the comparison result of gas sensing data and a gas concentration threshold, in combination with 3D visual point cloud data; identifying a structural weak area based on 3D visual point cloud data to obtain a structural risk matrix; and fusing the voltage risk matrix, the chemical risk matrix and the structural risk matrix to obtain a risk level matrix.

[0059] In some examples, the voltage safety threshold is a preset voltage critical value based on the power battery electrical characteristics, used to determine whether the real-time voltage detection data is safe or not, including total voltage threshold, single cell voltage threshold and voltage change rate threshold. The voltage risk matrix is a two-dimensional risk distribution matrix formed by mapping the comparison results of real-time voltage detection data and voltage safety threshold according to the spatial coordinates of 3D visual point cloud data, used to present the electrical safety risk of different areas. The gas concentration threshold is a preset concentration critical value for electrolyte volatiles and thermal runaway products, used to determine the safety state of gas sensing data, set respectively according to the gas type; it can be determined based on electrolyte composition analysis (such as boiling point of carbonate, decomposition temperature of lithium salt) and thermal runaway experimental data (such as accelerated calorimeter test results), for example, the dimethyl carbonate threshold is set to 5ppm (exceeding this value is determined as a slight leakage), 20ppm (serious leakage), the CO threshold is set to 10ppm (initial stage of thermal runaway), 100ppm (active stage of thermal runaway), and the HF threshold is set to 1ppm (warning required), 5ppm (emergency treatment required). The chemical risk matrix is a two-dimensional risk distribution matrix generated by combining the comparison results of gas sensing data and gas concentration threshold with the spatial coordinates of 3D visual point cloud data, used to present the chemical safety risk of different areas; the gas sampling point concentration can be calculated to the whole area grid through the Gaussian diffusion model, combined with the potential gas leakage path (such as shell gap, interface) identified by the point cloud data, and each grid is assigned a chemical risk level (1-4 levels). The structure weak area identification is based on 3D visual point cloud data, and the battery pack structure integrity is analyzed through algorithm to identify the areas with defects such as deformation, crack, loose connection, etc.; specifically, the RANSAC algorithm can be used to fit the deviation of point cloud data and standard CAD model, and when the local area point cloud deviation exceeds the structure parameter threshold (such as shell deformation≥3mm, screw loose displacement≥1mm, crack length≥5mm), it is determined as a structure weak area. The structure risk matrix is a two-dimensional risk distribution matrix formed by dividing the structure weak area identification results according to the spatial grid, used to present the physical structure safety risk of different areas. The voltage risk matrix, chemical risk matrix and structure risk matrix, which are three single-dimensional risk matrices, can be integrated into a global risk level matrix through a multi-modal data fusion algorithm, to comprehensively reflect the composite risk of each area.

[0060] In the implementation process, the control system first performs grid processing (generates 480 5 cm x 5 cm grids) on the 3D vision point cloud data, and associates the voltage and gas data through timestamp alignment (accuracy 1 ms); for the voltage risk matrix, the detection values of the 32 sensors are interpolated to the full grid, and after comparison with the voltage safety threshold, levels 1-4 are assigned; for the chemical risk matrix, the concentration of each grid is calculated through the gas diffusion model, and the level is corrected in combination with the leakage path identified by the point cloud; for the structural risk matrix, the level is automatically classified based on the deviation of the point cloud and the CAD model; then the fusion algorithm is started, the influence factors of each matrix are adjusted according to the spatial weight, the comprehensive risk value is calculated through the coupling formula, and finally the risk level matrix containing 480 grids is generated, the time consumption is ≤1.5 seconds, and the red marked 4-level risk area (such as a battery module with high voltage, HF leakage and shell deformation) in the matrix will be the key object to be avoided in subsequent disassembly path planning.

[0061] Through the implementation of the above embodiments, the risk matrix is generated and fused in multiple dimensions, which can realize the fine identification of safety risks. The voltage risk matrix accurately locuses the high-voltage dangerous area through visual positioning, the chemical risk matrix identifies the leakage source by associating the gas concentration with the spatial position, and the structural risk matrix focuses on the weak part to avoid secondary damage during disassembly; this mode of domain evaluation and global fusion solves the one-sidedness of traditional single risk evaluation, makes the risk identification more three-dimensional, reduces the risk of missed judgment, and provides multi-dimensional risk coordinates for path planning, ensuring that the planned path can avoid high-risk areas and improving the disassembly efficiency.

[0062] In some embodiments, the aforementioned fusion of the voltage risk matrix, the chemical risk matrix, and the structural risk matrix to obtain the risk level matrix can include: registering the voltage risk matrix, the chemical risk matrix, and the structural risk matrix based on a spatial position weight to obtain a voltage registration matrix, a chemical registration matrix, and a structural registration matrix, wherein the weight coefficient of the core area of the spatial position weight is higher than that of the edge area; performing multi-modal risk coupling processing based on the voltage risk matrix, the chemical risk matrix, and the structural risk matrix to obtain the risk level matrix.

[0063] In some examples, the spatial position weight is a weight coefficient set according to the safety influence degree of different regions of the battery pack in the disassembly process and the functional importance, used to adjust the contribution degree of each risk matrix when fused; the spatial position weight can be based on the battery pack structure design drawing (such as cell arrangement, high-voltage component position) and failure mode analysis (FMEA) result, and the weight values of different regions are calibrated by an expert system to form a spatial weight distribution atlas. The core region is a key region in the battery pack to be disassembled which is sensitive to safety risk and directly bears the function of storing and transmitting electric energy, mainly including the space where the cell module cluster, high-voltage connector, bus bar and other components are located, and specifically, the region where the cell arrangement density is ≥50 / ㎡ and there is an electrical connection with a voltage ≥36V can be defined as the core region. The edge region is an auxiliary region in the battery pack to be disassembled which has lower structural strength and smaller influence on the overall safety risk, mainly including the edge of the shell, the non-connected section of the cooling pipeline, the decorative parts and other parts. The multi-modal risk coupling processing is to nonlinearly fuse the registered voltage, chemical and structural risk matrices through mathematical algorithms, comprehensively consider the synergistic effect of different types of risks (such as nonlinear superposition of risks when gas leakage exists in high-voltage regions at the same time), and generate a quantitative result reflecting the combined risk. Based on spatial registration, the risk levels of the voltage risk matrix, the chemical risk matrix and the structural risk matrix can be integrated through the multi-modal risk coupling algorithm to generate a global risk level matrix.

[0064] In the specific implementation process, the core and edge regions can be automatically divided based on the semantic segmentation result (cell region / non-cell region) of the 3D visual point cloud data, and a spatial weight (core region 0.7, transition region 0.5, edge region 0.3) is assigned to each 5cm×5cm grid; then the voltage, chemical and structural risk matrices are multiplied by the spatial weight element by element to obtain the registration matrix (such as voltage registration matrix A'=A×spatial weight); then the multi-modal coupling algorithm is started, and the pre-stored coupling coefficient matrix (obtained by training 500 sets of experimental data) is called to perform nonlinear calculation on the registration matrix; finally, the calculation result of each grid is mapped to a 1-4 level risk to generate a risk level matrix, and the whole process takes ≤0.8 seconds; for example, a grid in the core region of the battery pack is upgraded to level 4 due to the coupling of voltage level 3 and chemical level 3, which is highlighted in red in the matrix, directly triggering the avoidance strategy of the subsequent disassembly path.

[0065] Through the implementation of the foregoing embodiments, the spatial position weight registration ensures the risk assessment accuracy of core areas (such as cell clusters and connectors), and avoids the interference of edge area data on key risks; the multi-modal risk coupling processing comprehensively considers the synergistic effects of electricity, chemistry and structure, solves the problem that a single risk assessment cannot deal with complex risks, makes the risk level matrix more suitable for actual disassembly scenes, improves the accuracy of safety warnings, and provides a more accurate risk heat map for path planning, reduces the path redundancy caused by risk assessment deviation, and improves the disassembly efficiency.

[0066] In some embodiments, the aforementioned multi-modal risk coupling processing based on the voltage risk matrix, the chemical risk matrix and the structural risk matrix to obtain the risk level matrix can include: based on the voltage risk matrix, the chemical risk matrix and the structural risk matrix, the multi-modal risk coupling processing is performed through the following formula to obtain the risk level matrix:

[0067] Y = a·A x b·B + g·C2

[0068] In the formula, Y is the risk level matrix, A is the voltage risk matrix, B is the chemical risk matrix, C is the structural risk matrix, a is a first adjustment coefficient, b is a second adjustment coefficient, and g is a third adjustment coefficient.

[0069] In some examples, the first adjustment coefficient is a parameter for adjusting the weight proportion of the voltage risk matrix in the coupling calculation, the value of which reflects the influence degree of the voltage risk on the overall risk level, and can be determined based on high-voltage fault statistical data of the power battery, through an orthogonal experiment method to determine the optimal value in different scenarios. The core area has greater high-voltage risk hazards, and the value of a is higher than that of the edge area. For example, for a ternary lithium battery pack, the value of a in the core area is set to 0.6 (the voltage risk weight is higher), and the value of a in the edge area is set to 0.3 (the voltage risk influence is weaker). For a lithium iron phosphate battery pack, because the thermal runaway risk is relatively low, the value of a in the core area can be lowered to 0.5. The second adjustment coefficient is a parameter for adjusting the weight proportion of the chemical risk matrix in the coupling calculation, the value of which is directly related to the electrolyte characteristics and the thermal runaway probability, and can be determined by fitting through a least squares method in combination with an electrolyte volatile concentration and thermal runaway correlation experiment (such as 500 groups of gas leakage tests at different temperatures). When strong corrosive gases such as lithium hexafluorophosphate and HF are detected, the value of β is dynamically increased. For example, at room temperature (25°C), the default value of β is 0.4. When the gas sensing data shows that the CO concentration exceeds 50 ppm (in the active period of thermal runaway), β is automatically increased to 0.6 to strengthen the influence weight of the chemical risk. The third adjustment coefficient is a parameter for adjusting the weight proportion of the square term of the structure risk matrix in the coupling calculation. Because the structure risk has a cumulative effect, such as a small deformation that may cause a chain damage, the square term is used to amplify its influence. The value of γ can be optimized based on structural mechanics simulation (such as ANSYS simulation of the extrusion effect of shell deformation on the battery cell) and fatigue experiment data through a genetic algorithm, and the value of γ in the structure weak area (such as the welding seam and screw connection position) is higher. For example, the value of γ in the battery pack module connecting beam area (structure failure is easy to cause short circuit of the battery cell) is set to 0.3, and the value of γ in the non-bearing area of the shell is set to 0.1. When the 3D visual point cloud data identifies that the crack length exceeds 8 mm, γ is temporarily increased to 0.5.

[0070] Through the implementation of the above embodiments, the weight of each risk factor is dynamically balanced by adjusting the coefficients in the quantitative coupling formula, for example, the value of a is increased in the high-voltage area to strengthen the influence of the voltage risk, and the value of γ is increased in the structure weak area to highlight the structure risk, which can avoid the subjective randomness of risk fusion, make the risk level matrix have explainability and adjustability, optimize the risk assessment for different battery pack types, and provide clear quantitative basis for path planning, to ensure the optimal balance between safety and efficiency in path selection.

[0071] In some embodiments, the foregoing step 103 can include: constructing a battery pack module connection topology graph based on the 3D visual point cloud data, wherein the nodes of the battery pack module connection topology graph represent the battery modules, and the edges represent the connection relationship strength between the battery modules; mapping the risk level matrix to the battery pack module connection topology graph to obtain a risk weight topology graph; determining a minimum risk propagation path from the starting disassembly point to the target disassembly point in the risk weight topology graph based on a preset risk propagation algorithm, wherein the risk propagation algorithm dynamically updates the node risk value by simulating the propagation process of the risk in the topology graph, and backtracks to generate the disassembly path; determining the disassembly strength parameter based on the connection relationship strength on the minimum risk propagation path; and determining the disassembly path priority based on the risk degree of each path in the risk weight topology graph.

[0072] In some examples, the battery pack module connection topology graph is a directed graph model constructed based on the internal module space connection relationship of the battery pack, which is used to abstractly express the physical connection logic of each module; the module boundaries can be identified by semantic segmentation of 3D visual point cloud data, and the connection features (such as bolts, welding points, and glue layers) between the modules can be extracted to model in a node-edge structure; for example, the battery pack module connection topology graph contains 12 nodes (corresponding to 12 cell modules) and 8 edges (corresponding to the Busbar connection between the modules), the direction of the edges is determined by the current flow direction, and the whole presents a series-parallel hybrid structure. The battery module is an independent functional unit composed of multiple cells combined in series / parallel, equipped with a shell and electrode lead-out end, and is the core component of the battery pack. The connection relationship strength is a quantitative parameter representing the connection firmness degree between the battery modules, reflecting the force or energy required during disassembly, and is related to the connection method (welding / bolt / gluing) and material properties, which can be assigned based on the connection type identified by 3D visual point cloud recognition and combined with a material database (such as laser welding strength 8-10 N / mm 2 , M6 bolt connection strength 5-7 N / mm 2 , and glue joint strength 2-3 N / mm 2). The risk weight topology map is a weighted graph formed by mapping the risk values of the risk level matrix to the battery pack module connection topology map, both nodes and edges carry risk parameters, the comprehensive risk value (level 1-4) of each battery module node can be associated with the corresponding area in the risk level matrix, and the risk weight of the connection edge is set as the product of the average value of the risk values of the two end nodes and the connection strength. The preset risk propagation algorithm is an iterative algorithm for simulating the propagation process of risk between nodes in the topology map, and the path cumulative risk is calculated based on the risk diffusion coefficient; the core modules of the preset risk propagation algorithm include: a risk initialization module (assigning an initial risk value to the node), a propagation coefficient matrix (defining the risk transfer probability between nodes, such as an adjacent node transfer probability of 0.8), an iterative update module (updating the node risk according to the propagation coefficient in each iteration), and a convergence determination module (stopping iteration when the risk change rate is less than 5%). Through the training of 500 sets of historical disassembly data, it is determined that the propagation attenuation coefficient is 0.95 (the risk is attenuated by 5% for each node transmission), and the upper limit of iteration is set to 20 rounds. The starting disassembly point is the initial position of the disassembly operation, which can select a region with low risk level and easy-to-separate structure, can select the boundary node with a risk level ≤2 and a connection strength ≤4 in the risk weight topology map, and preferentially selects the module close to the battery pack shell. The target disassembly point is the core module or component that needs to be finally separated in the disassembly operation, which can be a high-value recovery component (such as a cell cluster) or a high-risk component to be processed (such as a bulge module). The minimum risk propagation path is the path with the minimum risk accumulation value among all possible paths from the starting disassembly point to the target disassembly point; for example, there are 3 paths from the starting point S to the target point T, the total weight of path 1 is 58, the total weight of path 2 is 45, and the total weight of path 3 is 39, and path 3 is determined as the minimum risk propagation path. The risk propagation algorithm first assigns an initial risk value (equal to the node risk level) to the starting point, and then spreads to adjacent nodes according to the propagation coefficient (for example, node A risk level 3, spread 3x0.8=2.4 risk value to adjacent node B); after each iteration, the node risk value = initial risk + sum of adjacent node propagation risk, and the iteration is repeated until the risk value is stable; finally, the minimum risk contribution predecessor node is selected by backtracking from the target node, and the disassembly path is formed. According to the strength value of each connection edge in the path, the force parameter required can be calculated according to the intensity = intensity x 1.2 (safety factor), and the mechanical arm operation (such as bolt disassembly torque) and cutting robot parameters (such as laser power) are distinguished; for example, the connection strength of a certain segment in the path is 7 (screw connection), and the disassembly intensity parameter is set to 7x1.2=8.4N.m (torque); the strength of a certain laser welding edge is 9, and it is set to 9x1.2=10.8W / mm 2 (activation energy). All feasible paths can be sorted in ascending order of total risk weight, with the smallest weight being priority 1, and increasing in turn; when the total weight difference is <10%, the path with smaller connection strength sum is preferentially selected.

[0073] By implementing the above embodiments, the physical structure of the battery pack is converted into a calculable topological relationship by using the topological graph construction and risk propagation algorithm, which can solve the problem that the traditional path planning is difficult to adapt to complex module connections; the minimum risk propagation path optimizes the path by simulating the risk diffusion, which can avoid the risk accumulation caused by the fixed path; the force parameter based on the connection strength and the priority setting based on the risk degree make the disassembly operation accurately match the structural characteristics and orderly proceed according to the risk level, which breaks through the single path and improves the safety and efficiency of battery disassembly.

[0074] In some embodiments, the foregoing step 104 can include: according to the disassembly path priority, decomposing the disassembly operation sequence into a grabbing task and a cutting task, and assigning the grabbing task to the collaborative robot and the cutting task to the precision cutting robot; based on the disassembly force parameter, setting the grabbing force of the collaborative robot and the cutting force of the precision cutting robot; determining the task execution order according to the disassembly path priority, and synchronizing the operation timing of the collaborative robot and the precision cutting robot through the distributed collaborative controller; in the process of executing the disassembly operation, real-time monitoring of the multi-modal perception data, when the monitoring data in the multi-modal perception data exceeds the preset safety threshold, triggering the safety interrupt mechanism.

[0075] In some examples, the disassembly operation sequence is an ordered disassembly step set generated according to the disassembly path planning scheme, containing all operation actions from the starting point to the target point, arranged in time sequence; the path can be decomposed into atomic operations to form a time-sequenced operation list based on the minimum risk propagation path and the disassembly path priority through a task decomposition algorithm. The grabbing task is an operation executed by the collaborative robot, which is the core of clamping, carrying and light disassembly, mainly involving non-destructive separation actions, i.e. tasks that require controllable clamping force (≤50N) and do not require material separation, such as module carrying, screw disassembly, connector plugging. The cutting task is an operation executed by the precision cutting robot, which is the core of material separation, mainly involving separation of high-strength connections such as welding and riveting, i.e. tasks that require destruction of the connection structure through laser, milling, etc. such as Busbar laser cutting, metal shell milling. The grabbing force of the collaborative robot can be calculated according to the connection strength in the disassembly force parameter, i.e. grabbing force = module weight x 1.5 (safety factor) + 0.2 x connection strength, unit: N; for example, the M2 module weighs 3 kg, the connection strength is 3, the grabbing force = 3 x 9.8 x 1.5 + 0.2 x 3 ≈ 44.1 + 0.6 = 44.7 N, set to 45 N (accuracy ±1 N). The cutting force of the precision cutting robot is the power density (W / mm 2) is characterized by power density = disassembly intensity parameter x 0.8, and is characterized by spindle torque (N.m) during milling, and is set as torque = disassembly intensity parameter x 0.1; for example, disassembly intensity parameter 10.8 W / mm2(laser cutting), set power density to 10.8 x 0.8 = 8.64 W / mm2; disassembly intensity parameter 8.4 N.m (milling), set torque to 8.4 x 0.1 = 0.84 N.m. The task execution sequence is sorted from high to low according to the path priority, and tasks with the same priority are executed according to the operation sequence time sequence; the time sequence synchronization process is that the distributed collaborative controller assigns a timestamp (accuracy 1 ms) to each task, and the robot and the cutting robot exchange state information through the EtherCAT bus (period 10 ms), to ensure that the previous task completion signal triggers the start of the subsequent task. The safety interrupt mechanism triggers three levels of response after execution; level 1 (warning): the robot slows down and issues an audible and visual alarm; level 2 (warning): the robot suspends operation and maintains the current posture; level 3 (emergency): the robot immediately resets to a safe position (coordinates X = 0, Y = 0, Z = 500 mm), and starts the inert gas fire extinguishing device; for example, when cutting the M4 module, the gas sensor detects that the CO concentration is 120 ppm (exceeds the 100 ppm threshold), triggers level 3 interruption, the cutting robot resets within 0.5 seconds, and the robot stops in the safety area.

[0076] Through the implementation of the above-mentioned embodiments, the cooperation logic of the heterogeneous robot is clear, and the special task allocation of the cooperation robot and the cutting robot avoids the efficiency loss caused by functional overlap; the accurate setting of the intensity parameter solves the problem of cell damage or disassembly stagnation caused by improper intensity in traditional mechanical operation; the timing synchronization and safety interrupt mechanism realize dynamic safety control, avoid operation conflicts and risk diffusion in multi-machine cooperation, not only improve the disassembly precision (such as accurate control of cutting intensity) and efficiency (such as parallel operation), but also build a safety line through real-time monitoring and interrupt mechanism.

[0077] Further, as an implementation of the foregoing method embodiment, the application also provides a power battery disassembly device for implementing the foregoing method embodiment. The device embodiment corresponds to the foregoing method embodiment, and for the sake of reading, the details of the foregoing method embodiment will not be described one by one, but it should be clear that the device in the application embodiment can correspondingly implement all the contents in the foregoing method embodiment. For example, Figure 2As shown, the power battery disassembly device 20 comprises a data acquisition unit 201, a risk assessment unit 202, a disassembly planning unit 203, and a battery disassembly unit 204. The data acquisition unit 201 is configured to acquire multi-modal perception data of a battery pack to be disassembled, wherein the multi-modal perception data can include 3D visual point cloud data, real-time voltage detection data, and gas sensing data. The risk assessment unit 202 is configured to perform risk assessment on the multi-modal perception data based on a preset safety threshold to generate a risk level matrix. The disassembly planning unit 203 is configured to analyze and process the battery pack topology relationship of the battery pack to be disassembled based on the multi-modal perception data and the risk level matrix through a preset adaptive decision model to obtain a disassembly path planning scheme, wherein the disassembly path planning scheme can include a disassembly strength parameter and a disassembly path priority. The battery disassembly unit 204 is configured to control a heterogeneous robot cooperative execution unit to perform disassembly operation on the battery pack to be disassembled based on the disassembly strength parameter and the disassembly path priority, wherein the heterogeneous robot cooperative execution unit can include at least one collaborative robot arm and one precision cutting robot.

[0078] In some embodiments, the data acquisition unit 201 is further configured to obtain 3D visual point cloud data by scanning the battery pack to be disassembled through a three-dimensional laser scanner; obtain real-time voltage detection data by collecting voltage of the battery pack to be disassembled through a distributed voltage sensor array; and obtain gas sensing data by detecting gas of the battery pack to be disassembled through a multi-channel gas mass spectrometer, wherein the detection range of the multi-channel gas mass spectrometer includes electrolyte volatiles and thermal runaway products.

[0079] In some embodiments, the risk assessment unit 202 is further configured to generate a voltage risk matrix based on a comparison result of the real-time voltage detection data and a voltage safety threshold, in combination with the 3D visual point cloud data; generate a chemical risk matrix based on a comparison result of the gas sensing data and a gas concentration threshold, in combination with the 3D visual point cloud data; identify a structural weak area based on the 3D visual point cloud data to obtain a structural risk matrix; and fuse the voltage risk matrix, the chemical risk matrix, and the structural risk matrix to obtain the risk level matrix.

[0080] In some embodiments, the risk assessment unit 202 is further configured to register the voltage risk matrix, the chemical risk matrix, and the structural risk matrix based on a spatial position weight to obtain a voltage registration matrix, a chemical registration matrix, and a structural registration matrix, wherein the weight coefficient of a core area of the spatial position weight is higher than that of an edge area; and perform multi-modal risk coupling processing based on the voltage risk matrix, the chemical risk matrix, and the structural risk matrix to obtain the risk level matrix.

[0081] In some embodiments, the risk assessment unit 202 is further configured to perform a multi-modal risk coupling process based on the voltage risk matrix, the chemical risk matrix and the structure risk matrix to obtain a risk level matrix by the following formula:

[0082] Y = a A x b B + g C2

[0083] wherein Y is the risk level matrix, A is the voltage risk matrix, B is the chemical risk matrix, C is the structure risk matrix, a is a first adjustment coefficient, b is a second adjustment coefficient, and g is a third adjustment coefficient.

[0084] In some embodiments, the disassembly planning unit 203 is further configured to construct a battery pack module connection topology graph based on the 3D visual point cloud data, wherein the nodes of the battery pack module connection topology graph represent the battery modules, and the edges represent the connection relationship strength between the battery modules; map the risk level matrix to the battery pack module connection topology graph to obtain a risk weight topology graph; determine a minimum risk propagation path from a starting disassembly point to a target disassembly point in the risk weight topology graph based on a preset risk propagation algorithm, wherein the risk propagation algorithm simulates the propagation process of the risk in the topology graph, dynamically updates the node risk value, and backtracks to generate a disassembly path; determine a disassembly intensity parameter based on the connection relationship strength on the minimum risk propagation path; and determine a disassembly path priority based on the risk degree of each path in the risk weight topology graph.

[0085] In some embodiments, the battery disassembly unit 204 is further configured to decompose the disassembly operation sequence into a grasping task and a cutting task according to the disassembly path priority, assign the grasping task to the collaborative robot arm, and assign the cutting task to the precision cutting robot; set the grasping intensity of the collaborative robot arm and the cutting intensity of the precision cutting robot based on the disassembly intensity parameter; determine a task execution order according to the disassembly path priority, and synchronize the operation timing of the collaborative robot arm and the precision cutting robot through a distributed collaborative controller; and monitor the multi-modal perception data in real time during the execution of the disassembly operation, and trigger a safety interruption mechanism when the monitored data in the multi-modal perception data exceeds a preset safety threshold.

[0086] The application also provides a computer-readable storage medium having stored computer-executable instructions or computer programs, which, when executed by a processor, cause the processor to perform any step of the power battery disassembly method provided by the application.

[0087] In some embodiments, the computer-readable storage media can be random access memory (RAM), read-only memory (ROM), flash memory, magnetic surface memory, optical disc, or Compact Disc Read-Only Memory (CD-ROM), etc. storage, or various devices including one or any combination of the above storage.

[0088] In some embodiments, the computer-executable instructions can be in the form of programs, software, software modules, scripts, or code, written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages; and can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0089] In some embodiments, the computer-executable instructions can or can not correspond to files in a file system, can be stored in a portion of a file that holds other programs or data, for example, one or more scripts stored in a Hypertext Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files, for example, files that store one or more modules, sub programs, or portions of code.

[0090] In some embodiments, the computer-executable instructions can be deployed to execute on one electronic device, or on multiple electronic devices located at one site, or on multiple electronic devices distributed across multiple sites and interconnected by a communication network.

[0091] As shown in Figure 3 The present application also provides an electronic device 30, which includes a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, any step of the power battery disassembly method described above is implemented.

[0092] The present application also provides a computer program product, which includes a computer program or computer-executable instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer program or computer-executable instructions from the computer-readable storage medium, and the processor executes the computer program or computer-executable instructions, so that the electronic device performs any step of the power battery disassembly method described above.

[0093] The above examples are only used to illustrate the technical solutions of the present application, but not to limit the same; although the present application has been described in detail with reference to the foregoing examples, it should be understood by those of ordinary skill in the art that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for disassembling a power battery, characterized in that, include: Acquire multimodal sensing data of the battery pack to be disassembled, wherein the multimodal sensing data includes 3D visual point cloud data, real-time voltage detection data and gas sensing data; A risk assessment is performed on the multimodal sensing data based on a preset safety threshold to generate a risk level matrix; Based on the multimodal sensing data and the risk level matrix, the battery pack topology of the battery pack to be disassembled is analyzed and processed by a preset adaptive decision model to obtain a disassembly path planning scheme, wherein the disassembly path planning scheme includes disassembly force parameters and disassembly path priority. Based on the disassembly force parameter and the disassembly path priority, the heterogeneous robot collaborative execution unit is controlled to perform disassembly operation on the battery pack to be disassembled, wherein the heterogeneous robot collaborative execution unit includes at least one collaborative robotic arm and a precision cutting robot.

2. The method for disassembling a power battery according to claim 1, characterized in that, The acquisition of multimodal sensing data of the battery pack to be disassembled includes: The battery pack to be disassembled is scanned using a 3D laser scanner to obtain the 3D visual point cloud data; The voltage of the battery pack to be disassembled is collected by a distributed voltage sensor array to obtain the real-time voltage detection data; The gas of the battery pack to be disassembled is detected by a multi-channel gas mass spectrometer to obtain gas sensing data. The detection range of the multi-channel gas mass spectrometer includes electrolyte volatiles and thermal runaway products.

3. The method for disassembling a power battery according to claim 1, characterized in that, The step of performing a risk assessment on the multimodal sensing data based on a preset security threshold to generate a risk level matrix includes: Based on the comparison results between the real-time voltage detection data and the voltage safety threshold, and combined with the 3D visual point cloud data, a voltage risk matrix is ​​generated. Based on the comparison results between the gas sensing data and the gas concentration threshold, and combined with the 3D visual point cloud data, a chemical risk matrix is ​​generated. Based on the 3D visual point cloud data, weak structural regions are identified to obtain a structural risk matrix. The risk level matrix is ​​obtained by fusing the voltage risk matrix, the chemical risk matrix, and the structural risk matrix.

4. The method for disassembling a power battery according to claim 3, characterized in that, The process of fusing the voltage risk matrix, the chemical risk matrix, and the structural risk matrix to obtain the risk level matrix includes: The voltage risk matrix, chemical risk matrix, and structural risk matrix are registered based on spatial location weights to obtain voltage registration matrices, chemical registration matrices, and structural registration matrices, wherein the weight coefficients of the core region of the spatial location weights are higher than the weight coefficients of the edge region. Based on the voltage risk matrix, the chemical risk matrix, and the structural risk matrix, multimodal risk coupling processing is performed to obtain the risk level matrix.

5. The method for disassembling a power battery according to claim 4, characterized in that, The risk level matrix is ​​obtained by performing multimodal risk coupling processing based on the voltage risk matrix, the chemical risk matrix, and the structural risk matrix, including: Based on the voltage risk matrix, the chemical risk matrix, and the structural risk matrix, the risk level matrix is ​​obtained by performing multimodal risk coupling processing using the following formula: Y = α·A × β·B + γ·C² In the formula, Y is the risk level matrix, A is the voltage risk matrix, B is the chemical risk matrix, C is the structural risk matrix, α is the first adjustment coefficient, β is the second adjustment coefficient, and γ is the third adjustment coefficient.

6. The method for disassembling a power battery according to claim 1, characterized in that, The step involves analyzing and processing the battery pack topology of the battery pack to be disassembled using the multimodal sensing data and the risk level matrix, through a preset adaptive decision model, to obtain a disassembly path planning scheme, including: Based on the 3D visual point cloud data, a battery pack module connection topology graph is constructed, wherein the nodes of the battery pack module connection topology graph represent battery modules, and the edges represent the connection strength between the battery modules; The risk level matrix is ​​mapped to the battery pack module connection topology to obtain the risk weight topology. Based on a preset risk propagation algorithm, the minimum risk propagation path from the starting split point to the target split point is determined in the risk weight topology graph. The risk propagation algorithm dynamically updates the node risk value by simulating the risk propagation process in the topology graph and backtracks to generate a splitting path. The dismantling force parameter is determined based on the strength of the connection relationship on the minimum risk propagation path; The priority of the decomposition path is determined based on the risk level of each path in the risk weight topology graph.

7. The method for disassembling a power battery according to claim 1, characterized in that, The step of controlling the heterogeneous robot collaborative execution unit to perform disassembly operations on the battery pack to be disassembled based on the disassembly force parameter and the disassembly path priority includes: Based on the disassembly path priority, the disassembly operation sequence is decomposed into grasping tasks and cutting tasks, and the grasping tasks are assigned to the cooperative robotic arm and the cutting tasks are assigned to the precision cutting robot. Based on the disassembly force parameters, the gripping force of the collaborative robotic arm and the cutting force of the precision cutting robot are set. The task execution order is determined according to the priority of the disassembly path, and the operation timing of the cooperative robotic arm and the precision cutting robot is synchronized through a distributed collaborative controller. During the disassembly process, the multimodal sensing data is monitored in real time. When the monitored data in the multimodal sensing data exceeds a preset safety threshold, a safety interruption mechanism is triggered.

8. A power battery dismantling device, characterized in that, include: The data acquisition unit is used to acquire multimodal sensing data of the battery pack to be disassembled, wherein the multimodal sensing data includes 3D visual point cloud data, real-time voltage detection data and gas sensing data; The risk assessment unit is used to assess the risks of the multimodal sensing data based on a preset safety threshold and generate a risk level matrix. The disassembly planning unit is used to analyze and process the topological relationship of the battery pack to be disassembled based on the multimodal sensing data and the risk level matrix through a preset adaptive decision model to obtain a disassembly path planning scheme, wherein the disassembly path planning scheme includes disassembly force parameters and disassembly path priority. A battery disassembly unit is used to control a heterogeneous robot collaborative execution unit to perform disassembly operations on the battery pack to be disassembled based on the disassembly force parameters and the disassembly path priority. The heterogeneous robot collaborative execution unit includes at least one collaborative robotic arm and a precision cutting robot.

9. An electronic device, comprising: The memory and processor are characterized in that the processor, when executing a computer program stored in the memory, implements the steps of the power battery disassembly method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the power battery disassembly method as described in any one of claims 1 to 7.

Images