Intelligent loading and unloading system for carbon blocks

Abstract

The invention relates to the technical field of carbon block loading and unloading, and discloses an intelligent carbon block loading and unloading system, which comprises a sensing layer, a decision scheduling layer, an execution layer and a safety protection layer, and the four-layer architecture realizes differentiated collaborative loading and unloading of raw and cooked carbon blocks based on unified data bus real-time communication and linkage. The sensing layer accurately distinguishes raw and cooked carbon blocks and detects special features through a multi-modal data acquisition and recognition algorithm; the decision scheduling layer depends on a three-dimensional scheduling model, a dynamic path optimization algorithm and a digital twinning mapping module to realize task dynamic allocation and equipment collaboration; the execution layer adapts to the loading and unloading requirements of two types of carbon blocks through equipment such as a self-adaptive flexible clamping jaw; and the safety protection layer guarantees the operation safety through a partition sensing network, pre-judgment type response and multi-device interlocking. The system solves the problems of poor adaptability of raw and cooked carbon blocks, low efficiency and the like, improves the loading and unloading accuracy, efficiency and safety, and is suitable for heavy industry carbon block warehouse scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon block loading and unloading, and particularly relates to an intelligent carbon block loading and unloading system. BACKGROUND

[0002] As a core raw material in heavy industry fields such as electrolytic aluminum and steel, the loading, unloading and transfer operation of carbon blocks is a key link in the production process, which directly affects the production efficiency, product loss and operation safety. At present, carbon blocks in the industry are mainly divided into two categories: green carbon blocks and cooked carbon blocks. The green carbon block is a product after forming without baking, and the size is usually (1570±10) x (665±5) x (640±10) mm. It has characteristics such as large size tolerance, brittle material and easy to produce micro-cracks. The cooked carbon block is usually stable in size after baking, with a size of 1550 x 645 x 630 mm, high structural strength but easy to form an oxidation layer on the surface. The differences in physical properties of the two types of carbon blocks put forward different requirements for loading and unloading operations.

[0003] Traditional carbon block loading and unloading operations mainly rely on manual cooperation with forklifts, fixed mechanical arms and other equipment, which has many technical bottlenecks and industry pain points. First, the operation efficiency is low and the coordination is poor. In the traditional mode, the truck unloading, carbon block transfer and rack conveying need to be coordinated by manual operation. When green and cooked carbon blocks are mixed and loaded, path conflicts and equipment waiting problems often occur. The single carbon block loading and unloading cycle is long, which is difficult to meet the needs of large-scale production. At the same time, cooked carbon blocks can be batched, but due to the single operation mode, they cannot fully exert their advantages of uniform specifications. The green carbon block has large size tolerance, and manual grabbing is prone to positioning deviation.

[0004] Secondly, the product loss rate is high. The green carbon block is fragile, and it is difficult to accurately control the force during manual operation or traditional mechanical grabbing, which may cause extrusion damage or micro-crack expansion. Although the cooked carbon block has stable structure, it lacks precise positioning and guidance when stacking, which may cause collapse risk due to tilting, resulting in material waste and equipment damage. In addition, the traditional equipment has poor adaptability. Special mechanical arms and racks can only meet the operation of single type of carbon block. When facing the mixed storage of green and cooked carbon blocks or the storage of purchased baked blocks, the equipment parameters need to be frequently adjusted or the tooling needs to be replaced, which is complicated and affects the production continuity.

[0005] Moreover, the safety protection and intelligent level are insufficient, the single block weight of the carbon block is greater than or equal to 1000kg, manual operation is easy to cause safety accidents such as extrusion and falling when heavy load operation is performed; the existing safety system mainly relies on single emergency stop button, limit switch and other passive protection means, and it is difficult to deal with dynamic risks such as personnel misentry, vehicle overrun and carbon block stacking and tilting, the probability of false triggering is high, and the operation efficiency is seriously affected. At the same time, the traditional system lacks the perception and traceability ability of the whole process state of the carbon block, and cannot accurately identify the detail features such as the surface microcracks of the green carbon block and the oxidation layer of the mature carbon block, and it is also difficult to realize real-time monitoring and dynamic scheduling of the equipment operation state, and the intelligent level is far from meeting the needs.

[0006] In addition, the traditional operation also has environmental protection and energy consumption problems, the carbon powder generated in the loading and unloading process lacks effective global collection device, resulting in dust pollution exceeding the standard; the equipment driving mainly uses ordinary motor, and lacks targeted energy saving control strategy, and the energy consumption is high when heavy load operation is performed. Although some enterprises introduce AGV and simple visual system to realize local automation, there are still problems such as no differentiated operation of green and mature carbon blocks, rigid multi-device collaborative scheduling and insufficient sensing accuracy, and a full-process closed loop of identification, scheduling, execution and safety cannot be formed, so it is difficult to meet the multiple needs of accurate loading and unloading of green and mature carbon blocks, efficient collaboration and safety protection, and the transformation of carbon block loading and unloading operation to intelligent, flexible and green is restricted. Therefore, it is an urgent need to develop an intelligent loading and unloading system which is adapted to the differentiated characteristics of green and mature carbon blocks and deeply integrates multiple systems to solve the industry pain points. SUMMARY

[0007] The present application aims to provide an intelligent carbon block loading and unloading system to solve the problems of poor loading and unloading adaptability, low scheduling efficiency and insufficient safety protection caused by the difference in physical characteristics of green and mature carbon blocks.

[0008] To achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: An intelligent carbon block loading and unloading system, comprising a perception layer, a decision-making and scheduling layer, an execution layer and a safety protection layer, the four-layer architecture realizes real-time communication based on a unified data bus; The perception layer is used for multi-modal data acquisition and differentiated identification of green and mature carbon blocks; The decision-making and scheduling layer is used for dynamic scheduling and multi-device collaboration of green and mature carbon blocks; The execution layer is used for adaptive loading and unloading operation of green and mature carbon blocks; The safety protection layer is used for global safety monitoring and predictive protection; The four-layer architecture realizes differentiated collaborative loading and unloading of green and mature carbon blocks: The perception layer also configures a green and mature briquette identification algorithm, the green and mature briquette identification algorithm inputs include size tolerance, weight distribution, surface texture, thermal imaging features, and the output is a classification result; wherein, the green briquette special algorithm identifies the size deviation within ±10mm through edge detection and gray contrast, and marks the position of micro-cracks with a length greater than 3mm; the mature briquette special algorithm filters the reflection interference of the surface oxidation layer and locates the stacking barycenter; The decision scheduling layer includes a differentiated scheduling engine of green and mature briquettes, a multi-device collaborative algorithm, and a digital twin mapping module. The differentiated scheduling engine establishes a three-dimensional scheduling model of the priority, efficiency, and loss of green and mature briquettes, dynamically allocates job tasks according to production plans, device load rates, and briquette characteristics, and divides the briquette library into a green briquette exclusive area, a mature briquette exclusive area, and a transfer collaborative area. The multi-device collaborative algorithm is used for dynamic path optimization, the optimization goal of green briquettes is a low-vibration, short-distance path, and the optimization goal of mature briquettes is a high-speed, multi-task combined path; at the same time, a device load balancing strategy is adopted, and green briquette tasks are preferentially allocated to AGVs equipped with flexible grippers, and mature briquette tasks are allocated to AGVs with low load rates. The digital twin mapping module constructs digital twins of the briquette library, devices, and briquettes, synchronizes the positions of green and mature briquettes, device operating parameters, and safety states in real time, and supports virtual simulation scheduling and risk prediction. The execution layer includes an adaptive flexible gripper, a variable-diameter power roller rack, and an omnidirectional heavy AGV. The safety protection layer includes a partitioned safety sensor network, a pre-judgment emergency response module, and a multi-device interlocking mechanism. The partitioned safety sensor network includes an AI human body detection camera and a double laser beam sensor in the green briquette work area, an AI human body detection camera and a single laser beam sensor in the mature briquette work area, and an AGV-triggered lifting and moving safety fence in the transfer collaborative area. The pre-judgment emergency response module is used to control the device parameters in the execution layer. The multi-device interlocking mechanism synchronously stops the devices associated with a device that triggers a safety warning.

[0009] The principle and advantages of the scheme are: in actual application, the differentiated characteristics of raw and cooked carbon blocks are accurately identified through the perception layer to provide data support for subsequent operations; the decision scheduling layer realizes the zoning operation and dynamic path planning of raw and cooked carbon blocks based on the three-dimensional scheduling model and digital twin technology, avoids path conflicts and equipment idling; the execution layer balances the needs of raw carbon block damage prevention and efficient loading and unloading of cooked carbon blocks through adaptive structure and driving mode; the zoning monitoring and predictive response of the safety protection layer realize the full-scene risk prevention and control. Finally, the multiple problems of poor adaptability, low operation efficiency, high carbon block loss and passive safety protection in the traditional system are solved, and the integrated operation effect of accurate identification, dynamic scheduling, flexible operation and global safety is achieved.

[0010] Preferably, as an improvement, the perception layer comprises a multi-modal vision module, a weight and vibration sensor group, and an environmental monitoring sensor; The multi-modal vision module includes an industrial camera and an infrared thermal imager, a high-definition 3D camera is configured in the raw carbon block operation area to collect raw carbon block size deviation and surface micro-crack data; a high-speed 2D camera is configured in the cooked carbon block operation area to quickly identify the stacking posture of cooked carbon blocks; The weight and vibration sensor group includes a high-precision weight sensor installed on the adaptive flexible gripper and a vibration sensor built-in the drum rack, which is used to feedback the carbon block weight and transfer vibration data in real time; The environmental monitoring sensor includes a dust concentration sensor and a temperature and humidity sensor, which is used to link the multi-modal vision module to adjust the exposure parameters and eliminate dust interference.

[0011] Technical effect: The combination of multi-modal sensors realizes the all-round and high-precision collection of raw and cooked carbon block characteristics, the environmental sensors and vision modules are linked to effectively resist the interference of carbon block dust, temperature and humidity fluctuations on recognition accuracy, ensuring the accuracy of key data such as size deviation, micro-cracks and stacking posture, and providing a reliable data foundation for differentiated scheduling and adaptive execution.

[0012] Preferably, as an improvement, the raw and cooked carbon block recognition algorithm realizes: The collected multi-modal data is preprocessed; the size tolerance, weight distribution, surface texture and thermal imaging features are extracted through edge detection and gray level co-occurrence matrix method; the normalized weighted fusion is used to generate a feature vector, which is input into the CNN model with improved ResNet-18 architecture, and the classification result is output; special feature optimization is performed, the size deviation and micro-crack length of raw carbon blocks are calculated, the oxidation layer of cooked carbon blocks is filtered and the center of gravity is located; the result output and feedback are synchronized to the decision scheduling layer and digital twin.

[0013] Technical effect: The integrated processing of raw and cooked carbon block classification and special feature detection provides accurate parameters for differentiated operation of the execution layer.

[0014] Preferably, as an improvement, the raw and cooked carbon block recognition algorithm adopts a normalized weighted fusion method to realize a multi-feature fusion formula:

[0015]

[0016] wherein, is the feature normalization result, , are the maximum and minimum values of the ith feature, respectively, is the feature weight.

[0017] Technical effect: By normalization processing, the interference of different dimensional features is eliminated, and dynamic weight distribution can focus on key features according to the characteristics of raw and cooked carbon blocks, so that the fused feature vector is more distinguishable, and the accuracy and generalization ability of raw and cooked carbon block classification are significantly improved, avoiding the limitations of single feature recognition.

[0018] Preferably, as an improvement, the raw carbon block special algorithm adopts a Euclidean distance formula to calculate the size deviation, and the formula is:

[0019] wherein, is the actual physical coordinates of the two points on the edge of the carbon block after perspective transformation, is the standard length, width or height of the raw carbon block, is the size deviation, and the threshold is ±10mm.

[0020] Technical effect: Quickly identify abnormal carbon blocks that exceed the tolerance range, provide quantitative basis for adjusting the jaw opening and gripping force of the execution layer, and avoid carbon block extrusion damage or loose gripping caused by size deviation.

[0021] Preferably, as an improvement, the raw carbon block special algorithm adopts a pixel and physical scale conversion formula to calculate the micro-crack length:

[0022] wherein, is the pixel and physical scale conversion coefficient obtained by camera calibration, are the pixel coordinates of the two endpoints of the micro-crack; when , it is marked as a defective carbon block.

[0023] Technical effect: Realize accurate detection and length quantification of micro-cracks on the surface of raw carbon blocks, facilitate early screening of carbon blocks with damage risk, avoid carbon block breakage caused by crack expansion during subsequent transportation and handling, and reduce the damage rate of raw carbon blocks.

[0024] Preferably, as an improvement, the cooked carbon block special algorithm adopts a gray weighted barycenter algorithm to locate the stacking barycenter, with the formula being:

[0025] wherein, is the physical coordinates of the (m, n)th pixel in the stacking carbon block region; is the gray value of the corresponding pixel; M and N are respectively the number of pixel rows and columns of the stacking carbon block region.

[0026] Technical effect: Ensure balanced force during grabbing, avoid stacking carbon block tilting and collapse, and improve stacking neatness.

[0027] Preferably, as an improvement, the task allocation step of the differentiated scheduling engine includes: receiving production plan tasks, extracting task type, quantity, priority; obtaining real-time data such as current carbon block warehouse partition occupancy, device load rate, idle device type; inputting task information and state data into a three-dimensional scheduling model, calculating the comprehensive evaluation index of each candidate scheme; selecting the scheduling scheme with the largest comprehensive evaluation index value to determine the work device, work area and task execution order; updating the device state and carbon block information in real time during the work process, and when device failure or task change occurs, recalculating the comprehensive evaluation index and adjusting the scheduling scheme.

[0028] Technical effect: Realize dynamic and intelligent task allocation, balance the three goals of priority, efficiency and loss through comprehensive evaluation index, and avoid the defects of single-target-oriented decision-making.

[0029] Preferably, as an improvement, the dynamic path optimization algorithm introduces a vibration cost factor and a task merging penalty term to adapt to the path requirements of raw and cooked carbon blocks respectively: Raw carbon block path optimization objective function:

[0030] Cooked carbon block path optimization objective function:

[0031] wherein, is the path length, is the path vibration integral value, is the AGV travel speed, is the task merging penalty term, , , , is the weight coefficient.

[0032] Technical effects: The green charcoal block path controls the transfer vibration amplitude through the vibration cost factor, reduces the risk of micro-crack expansion by combining short-distance optimization, and improves the transfer efficiency by optimizing the speed and merging tasks. The single batch task operation cycle is shortened. At the same time, it avoids congestion caused by cross-operation of multiple devices, and further improves the overall operation efficiency and stability of the system. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 It is a structural schematic diagram of an intelligent carbon block loading and unloading system. DETAILED DESCRIPTION

[0034] The following will be further described in detail through specific embodiments: The embodiment is basically as shown in the accompanying drawings: Figure 1 An intelligent carbon block loading and unloading system includes a perception layer, a decision-making and scheduling layer, an execution layer, and a safety protection layer. The four-layer architecture realizes real-time communication based on a unified data bus.

[0035] The perception layer is used for multi-modal data acquisition and differential identification of green and cooked carbon blocks.

[0036] The decision-making and scheduling layer is used for dynamic scheduling and multi-device collaboration of green and cooked carbon blocks.

[0037] The execution layer is used for adaptive loading and unloading operation of green and cooked carbon blocks.

[0038] The safety protection layer is used for global safety monitoring and predictive protection.

[0039] The four-layer architecture realizes differential collaborative loading and unloading of green and cooked carbon blocks. The work flow includes: the control system receives the production plan, the digital twin loads the current carbon block library state and device load data, and the scheduling engine initializes the operation mode; after the vehicle stops, the perception layer starts the vision module and sensor group, collects carbon block data, judges the green and cooked types through classification algorithm, and outputs the differential identification result; the scheduling engine allocates operation devices and plans exclusive paths according to the identification result, and synchronously updates the operation process in the digital twin; the execution layer operates according to the differential parameters, the driving system receives the sensor feedback in real time, and dynamically adjusts the action parameters; the safety protection layer monitors the whole process, predicts the risk and responds in time; after the operation is completed, the carbon block library data is updated to form a closed loop. Specifically: The perception layer includes a multi-modal vision module, a weight and vibration sensor group, and an environment monitoring sensor.

[0040] The multi-modal vision module includes an industrial camera and an infrared thermal imager. The green carbon block operation area is configured with a high-definition 3D camera for collecting green carbon block size deviation and surface micro-crack data. The cooked carbon block operation area is configured with a high-speed 2D camera for quickly identifying the cooked carbon block stacking posture.

[0041] ​The weight and vibration sensor group includes a high-precision weight sensor mounted on the adaptive flexible clamp jaw and a vibration sensor built into the roller rack, which is used to feed back the carbon block weight and transfer vibration data in real time. The environmental monitoring sensor includes a dust concentration sensor and a temperature and humidity sensor, which are used to link the multi-modal vision module to adjust the exposure parameters and eliminate dust interference.

[0042] The perception layer also configures a green and cooked carbon block recognition algorithm, which is a multi-feature fusion classification model based on deep learning. The input includes size tolerance, weight distribution, surface texture, and thermal imaging features, and the output is a classification result.

[0043] The green and cooked carbon block recognition algorithm includes: preprocessing the collected multi-modal data; the multi-modal data includes carbon block images, weight, thermal imaging, and vibration data collected by the perception layer camera and sensor; the data preprocessing includes Gaussian filter denoising and grayscale conversion for images, 3σ criterion for removing outliers for weight data, and temperature normalization for thermal imaging data. The size tolerance, weight distribution, surface texture, and thermal imaging features are extracted by edge detection and gray level co-occurrence matrix method; the normalized weighted fusion is used to generate a feature vector, which is input into the improved ResNet-18 architecture CNN model, and the classification result is output; special feature optimization is performed, the size deviation and micro-crack length of green carbon block are calculated, and the oxidation layer of cooked carbon block is filtered and the center of gravity is located; the result output and feedback are synchronized to the decision-making and scheduling layer and the digital twin.

[0044] The green and cooked carbon block recognition algorithm uses a normalized weighted fusion method to realize a multi-feature fusion formula:

[0045]

[0046] wherein, is the feature normalization result, , are the maximum and minimum values of the ith feature, respectively, is the feature weight. In the green carbon block scenario, the size tolerance weight and surface texture weight are high, and in the cooked carbon block scenario, the weight distribution weight and thermal imaging feature weight are high. The specific data is obtained according to historical data statistics.

[0047] The green carbon block special algorithm uses the Euclidean distance formula to calculate the size deviation, and the formula is:

[0048] wherein, is the actual physical coordinates of the two points on the edge of the carbon block after perspective transformation, is the standard length, width, or height of the green carbon block, For dimensional deviation, the threshold is ±10mm.

[0049] The green coke lump special algorithm uses a pixel and physical scale conversion formula to calculate the micro crack length:

[0050] wherein, is the pixel and physical scale conversion coefficient obtained by camera calibration, is the pixel coordinates of the two endpoints of the micro crack; when , it is marked as a defective coke lump.

[0051] The gray-scale weighted barycenter algorithm is used in the special algorithm for cooked coke lumps to locate the stacking barycenter, and the formula is:

[0052] wherein, is the physical coordinates of the (m, n)th pixel in the stacked coke lump region; is the gray value of the corresponding pixel; M and N are respectively the number of pixel rows and columns of the stacked coke lump region.

[0053] The decision scheduling layer includes a differentiated scheduling engine for green and cooked coke lumps, a multi-device collaborative algorithm, and a digital twin mapping module.

[0054] The differentiated scheduling engine establishes a three-dimensional scheduling model for the priority, efficiency, and loss of green and cooked coke lumps, dynamically allocates job tasks according to production plans, device load rates, and coke lump characteristics, divides the coke lump warehouse into a green coke lump exclusive area, a cooked coke lump exclusive area, and a transfer collaborative area, and avoids job path conflicts. The three-dimensional scheduling model is:

[0055] wherein, is the comprehensive evaluation index of the scheduling scheme, and the larger the value, the better the scheme; P is the task priority coefficient, which is set according to the production plan, and the cooked coke lump out-of-warehouse priority value is 0.8-1.0, and the green coke lump into-warehouse priority value is 0.5-0.7; T is the single batch task operation cycle; is the coke lump loss rate, and the threshold value of green coke lump is ≤0.8%, and the threshold value of cooked coke lump is ≤0.1%; , , is the weight coefficient, which is dynamically adjusted in the range of: ∈[0.4, 0.6], ∈[0.3, 0.4], ∈[0.1, 0.2].

[0056] The task allocation step of the differential scheduling engine includes: receiving production plan tasks, extracting task type, quantity, priority; obtaining real-time data such as current coke block library partition occupancy, equipment load rate, idle equipment type; inputting task information and state data into a three-dimensional scheduling model, calculating the comprehensive evaluation index of each candidate scheme; selecting the scheduling scheme with the maximum comprehensive evaluation index value to determine the work equipment, work area and task execution order; updating the equipment state and coke block information in real time during the work process, and when equipment failure or task change occurs, recalculating the comprehensive evaluation index and adjusting the scheduling scheme.

[0057] The multi-device cooperative algorithm is used for dynamic path optimization, a dynamic path optimization algorithm is adopted, the green coke optimization target is low vibration and short distance path, and the mature coke optimization target is high speed and multi-task combined path; the dynamic path optimization algorithm introduces a vibration cost factor and a task combination penalty term, which are respectively adapted to the green and mature coke path requirements: Green coke path optimization objective function:

[0058] Mature coke path optimization objective function:

[0059] Wherein, is the path length, is the path vibration integral value, is the AGV driving speed, is the task combination penalty term, , , , is a weight coefficient.

[0060] A device load balancing strategy is adopted, green coke tasks are preferentially allocated to AGVs equipped with flexible clamps, and mature coke tasks are preferentially allocated to AGVs with low load rates. In this embodiment, the load rate threshold for allocation is 60%, and when the load rate is less than 60%, the mature coke multi-task combined work is preferentially allocated; green coke tasks are only allocated to AGVs with a load rate less than 80% and equipped with adaptive flexible clamps.

[0061] The digital twin mapping module constructs a carbon block library, equipment, and digital twin of carbon blocks, synchronizes real-time green and cooked carbon block positions, equipment operating parameters, and safety states, supports virtual simulation scheduling and risk prediction; the implementation steps of the digital twin mapping module include: based on the three-dimensional model of the carbon block library, the CAD drawings of the equipment, and the carbon block parameters, a virtual model that is 1:1 mapped with the physical scene is built, and state parameters such as position, speed, and load are defined; real-time data synchronization, through a unified data bus, sensor data from the perception layer, equipment operating data from the execution layer, and state data from the safety protection layer are collected, and the MQTT communication protocol is used to realize millisecond-level data synchronization between physical entities and virtual models; virtual simulation scheduling, after a new task is assigned, the working process of different scheduling schemes is simulated in the digital twin, the path conflict probability, equipment load rate, and carbon block loss risk are calculated, and the optimal scheme is output; risk prediction and feedback, through the virtual model, the equipment operating trend is monitored in real time, when it is predicted that the AGV load rate will exceed the threshold or the path will be congested, adjustment instructions are sent to the differentiated scheduling engine in advance.

[0062] The execution layer includes an adaptive flexible gripper, a variable-diameter power roller rack, and an omnidirectional heavy AGV.

[0063] The adaptive flexible gripper adopts a telescopic multi-claw structure, is equipped with a green carbon block special silica gel non-slip pad and a cooked carbon block special hard wear-resistant pad, and the pads can be quickly replaced; based on the feedback data of the perception layer, the pressure of the hydraulic servo proportional valve is dynamically adjusted to control the green and cooked carbon block grabbing force, supporting single green carbon block accurate grabbing and double cooked carbon block synchronous grabbing.

[0064] The roller spacing of the variable-diameter power roller rack is dynamically adjusted by a servo motor, the adjustment range is 50-100mm, the green carbon block operation time interval is reduced to 50mm, and the cooked carbon block operation time interval is expanded to 100mm; the roller speed is adaptively adjusted, and the roller surface adopts differential materials, the green carbon block area is soft rubber, and the cooked carbon block area is hard alloy.

[0065] The omnidirectional heavy AGV adopts a servo motor and hydraulic drive composite mode, the hydraulic buffer device is enabled during green carbon block transfer, and the servo motor high-speed mode is switched to during cooked carbon block transfer, and magnetic stripe navigation and visual calibration are fused.

[0066] The safety protection layer includes a partitioned safety sensor network, a predictive emergency response module, and a multi-device interlocking mechanism. The partitioned safety sensor network includes an AI human body detection camera and a double laser beam sensor in the green carbon block operation area, an AI human body detection camera and a single laser beam sensor in the cooked carbon block operation area, and an AGV triggered lifting and moving safety fence in the transfer coordination area.

[0067] The pre-judgment emergency response module is used for controlling the parameters of each device in the execution layer; for example, based on the vibration sensor data, when the vibration amplitude of the carbon block transfer exceeds ±1 mm, the AGV is controlled to slow down and the clamping jaw is locked; when the vision system detects that the inclination angle of the stack of carbon blocks exceeds 15°, the subsequent transfer task is suspended, and the mechanical arm is linked to adjust the stack posture.

[0068] The multi-device interlocking mechanism synchronously stops the devices associated with a device triggering a safety warning.

[0069] The above is only an embodiment of the present application, and common technical solutions and / or characteristics in the scheme are not described in detail. It should be noted that, for those skilled in the art, without departing from the technical solutions of the present application, a number of modifications and improvements can be made, which should also be considered as the protection scope of the present application, and these will not affect the effect and practicality of the present application. The protection scope claimed in the present application should be subject to the content of its claims, and the specific implementation mode and the like recorded in the specification can be used to explain the content of the claims.

Claims

1. A smart loading and unloading system for charcoal blocks, characterized in that, The architecture comprises a perception layer, a decision-making and scheduling layer, an execution layer, and a security protection layer. It is based on a unified data bus to achieve real-time communication. The sensing layer is used for multimodal data acquisition and differential identification of raw and cooked charcoal blocks; The decision-making and scheduling layer is used for dynamic scheduling of raw and cooked charcoal blocks and multi-device collaboration. The execution layer is used for adaptive loading and unloading operations of raw and cooked charcoal blocks; The security protection layer is used for comprehensive security monitoring and predictive protection. A four-layer architecture enables differentiated and collaborative loading and unloading of raw charcoal blocks and cooked charcoal blocks: The perception layer is also equipped with a raw and cooked charcoal block recognition algorithm. The input of the raw and cooked charcoal block recognition algorithm includes dimensional tolerance, weight distribution, surface texture, and thermal imaging features, and the output is a classification result. Among them, the raw charcoal block-specific algorithm identifies dimensional deviations within ±10mm through edge detection and grayscale comparison, and marks the location of microcracks with a length greater than 3mm; the cooked charcoal block-specific algorithm filters out the reflection interference of the surface oxide layer and locates the stacking center of gravity. The decision-making and scheduling layer includes a differentiated scheduling engine for raw and cooked charcoal blocks, a multi-device collaborative algorithm, and a digital twin mapping module; The differentiated scheduling engine establishes a three-dimensional scheduling model of priority, efficiency, and loss of raw and cooked charcoal blocks. It dynamically allocates work tasks according to production plans, equipment load rates, and charcoal block characteristics. The charcoal block library is divided into a dedicated area for raw charcoal blocks, a dedicated area for cooked charcoal blocks, and a transit and collaboration area. The multi-device collaborative algorithm is used for dynamic path optimization. The optimization target for raw charcoal blocks is low vibration and short distance path, while the optimization target for cooked charcoal blocks is high speed and multi-task merging path. At the same time, a device load balancing strategy is adopted, prioritizing the allocation of raw charcoal block tasks to AGVs equipped with flexible grippers, and assigning cooked charcoal block tasks to AGVs with low load rates. The digital twin mapping module constructs digital twins of the charcoal block library, equipment, and charcoal blocks, and synchronizes the location of raw and cooked charcoal blocks, equipment operating parameters, and safety status in real time, supporting virtual simulation scheduling and risk prediction. The execution layer includes an adaptive flexible gripper, a variable diameter powered roller rack, and an omnidirectional heavy-duty AGV; The security protection layer includes a partitioned security sensor network, a predictive emergency response module, and a multi-device interlocking mechanism. The zoned safety sensor network includes an AI human detection camera and dual laser beam sensors in the raw charcoal block operation area, an AI human detection camera and single laser beam sensors in the cooked charcoal block operation area, and an AGV-triggered lifting and moving safety fence in the transfer and coordination area. The predictive emergency response module is used to control the parameters of each device in the execution layer; The multi-device interlocking mechanism causes associated devices to shut down synchronously when a device triggers a safety warning.

2. The intelligent loading and unloading system for charcoal blocks according to claim 1, characterized in that: The perception layer includes a multimodal vision module, a weight and vibration sensor group, and an environmental monitoring sensor; The multimodal vision module includes an industrial camera and an infrared thermal imager. The raw charcoal block working area is equipped with a high-definition 3D camera to collect data on raw charcoal block size deviation and surface micro-cracks; the cooked charcoal block working area is equipped with a high-speed 2D camera to quickly identify the stacking posture of cooked charcoal blocks. The weight and vibration sensor group includes a high-precision weight sensor mounted on an adaptive flexible gripper and a vibration sensor built into the roller feeder, used to provide real-time feedback on the weight of the carbon blocks and the vibration data during transport. The environmental monitoring sensors include a dust concentration sensor and a temperature and humidity sensor, which are used to link with the multimodal vision module to adjust exposure parameters and eliminate dust interference.

3. The intelligent loading and unloading system for charcoal blocks according to claim 1, characterized in that, The algorithm for identifying raw and cooked charcoal blocks includes: The collected multimodal data is preprocessed; dimensional tolerance, weight distribution, surface texture, and thermal imaging features are extracted using edge detection and gray-level co-occurrence matrix methods; normalized weighted fusion is used to generate feature vectors, which are then input into a CNN model with an improved ResNet-18 architecture to output classification results; specific feature optimization is performed, including calculating dimensional deviations and microcrack lengths for raw charcoal blocks and filtering the oxide layer and locating the center of gravity for cooked charcoal blocks; the results are output and fed back, synchronizing the recognition results to the decision scheduling layer and the digital twin.

4. The intelligent loading and unloading system for charcoal blocks according to claim 1, characterized in that, The algorithm for identifying raw and cooked charcoal blocks uses a normalized weighted fusion method to achieve multi-feature fusion. The formula is as follows: in, For the feature normalization results, , Let be the maximum and minimum values ​​of the i-th feature, respectively. These are the feature weights.

5. The intelligent loading and unloading system for charcoal blocks according to claim 1, characterized in that, The specific algorithm for calculating dimensional deviations using the Euclidean distance formula is as follows: in, These are the actual physical coordinates of two points on the edge of the charcoal block after perspective transformation. The standard length, width, or height of the charcoal block. For dimensional deviation, the threshold is ±10mm.

6. The intelligent loading and unloading system for charcoal blocks according to claim 1, characterized in that, The specific algorithm for raw charcoal blocks uses a pixel-to-physical scale conversion formula to calculate the microcrack length: in, The pixel-to-physical scale conversion coefficients obtained for camera calibration. The pixel coordinates of the two ends of the microcrack; when At that time, it was marked as a defective carbon block.

7. The intelligent loading and unloading system for charcoal blocks according to claim 1, characterized in that, The specific algorithm for calcined charcoal blocks uses a gray-scale weighted centroid algorithm to locate the stacking centroid, and the formula is as follows: in, Let m be the physical coordinates of the (m,n)th pixel within the stacked carbon block region; represents the grayscale value of the corresponding pixel; M and N are the number of rows and columns of pixels in the stacked carbon block area, respectively.

8. The intelligent loading and unloading system for charcoal blocks according to claim 1, characterized in that, The task allocation steps of the differentiated scheduling engine include: receiving production plan tasks and extracting task type, quantity, and priority; obtaining real-time data such as the current occupancy status of the charcoal block warehouse partitions, equipment load rate, and idle equipment type; inputting task information and status data into the three-dimensional scheduling model and calculating the comprehensive evaluation index of each candidate scheme; selecting the scheduling scheme with the largest comprehensive evaluation index value and determining the operating equipment, operating area, and task execution order; updating equipment status and charcoal block information in real time during operation; and recalculating the comprehensive evaluation index and adjusting the scheduling scheme when equipment failure or task change occurs.

9. The intelligent loading and unloading system for charcoal blocks according to claim 1, characterized in that, The dynamic path optimization algorithm introduces a vibration cost factor and a task merging penalty term to adapt to the path requirements of raw and cooked charcoal blocks respectively: Objective function for biochar block path optimization: Objective function for path optimization of charcoal blocks: in, For path length, This is the integral value of path vibration. The speed at which the AGV travels. To merge the penalty items for the task, , , , These are the weighting coefficients.

Images