An ai-based coal mill steel ball dynamic proportioning and robot transfer method and system

Abstract

The present application relates to the technical field of coal mill operation control and material transfer, and particularly relates to a coal mill steel ball dynamic proportioning method and system based on AI and a robot transfer method and system. The method comprises the following steps: obtaining coal mill operation parameters, performing historical window arrangement and coal powder fineness related parameter registration, performing coal powder fineness prediction, steel ball wear rate prediction and trigger determination; generating a quality proportion based on the trigger determination result, arranging the steel ball specification corresponding relationship and completing five-channel material distribution valve channel mapping to form a proportioning scheme and a transfer task; and then performing path planning, obstacle avoidance path generation, coal mill platform in-place docking, material distribution valve channel distribution, and after ball supplementing, operation parameter acquisition, comparison calibration and proportioning scheme updating processing. The present application realizes continuous connection of steel ball dynamic proportioning, robot transfer and calibration after ball supplementing.

Description

Technical Field

[0001] This invention relates to the field of coal mill operation control and material transfer technology, and in particular to an AI-based method and system for dynamic proportioning of steel balls in a coal mill and robotic transfer. Background Technology

[0002] In the field of coal mill operation control and material transfer technology, existing solutions usually revolve around coal mill operation parameter acquisition, ball replenishment mechanism control, coal powder fineness adjustment, steel ball inventory management, and transfer equipment scheduling. These solutions have limitations such as insufficient dynamic steel ball ratio, disconnect between robot transfer and ball replenishment control, and lack of continuous calibration of the operating status after ball replenishment.

[0003] Existing methods often rely on the separate monitoring of coal mill operating parameters, fixed ball replenishment rhythm, fixed ball replenishment mechanism actions, or independent execution by transfer equipment. In scenarios where the coal mill platform and feeding position are connected, unclear correspondence between steel ball specifications and discontinuous mapping of the five-channel material distribution valve channels can easily occur, making it difficult to achieve stable dynamic proportioning of steel balls and robotic transfer based on coal mill operating parameters, proportioning scheme packages, and transfer task packages.

[0004] Existing technologies generally suffer from common shortcomings in the joint processing of coal mill operating parameters, coal powder fineness prediction, steel ball wear rate prediction, mass ratio generation, steel ball specification correspondence organization, five-channel material distribution valve channel mapping, transfer task package, in-place addition package, and operating parameter package. These shortcomings include the separation of trigger judgment and proportioning scheme generation, weak connection between proportioning scheme package and transfer task package, disconnection between in-place addition package and acquisition of coal mill operating parameters after ball replenishment, and insufficient comparison calibration and proportioning scheme updates. These shortcomings make it difficult to form a consistent process of data collection, judgment, transfer, addition, and updating in the application scenario of dynamic steel ball proportioning and robotic transfer in coal mills. As a result, there is a lack of continuous link between dynamic steel ball proportioning and robotic transfer, and the operating status after ball replenishment is difficult to write back to the subsequent processing stage. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an AI-based method for dynamic proportioning and robotic transfer of steel balls in a coal mill, comprising:

[0006] S100: Obtain the operating parameters of the coal mill, organize the 24-hour historical window and register the parameters related to coal powder fineness, generate a prediction input package, and perform coal powder fineness prediction, steel ball wear rate prediction and trigger judgment processing to generate a trigger judgment package.

[0007] S200. Based on the trigger judgment package, perform mass ratio generation, steel ball specification correspondence organization and five-channel material distribution valve channel mapping processing to generate a proportioning scheme package; based on the proportioning scheme package, perform task field assembly, target floor registration and material feeding location registration processing to generate a transfer task package;

[0008] S300. Based on the transfer task package, perform A* algorithm path planning and DWA algorithm obstacle avoidance path generation, and perform docking processing to reach the coal mill platform and feeding position to generate the in-place addition package.

[0009] S400. Based on the aforementioned in-place addition package, perform five-channel material distribution valve channel allocation processing, and execute the acquisition, comparison, calibration, and ratio scheme update processing of the coal mill operating parameters after ball replenishment to generate an operating parameter package.

[0010] Furthermore, the process of organizing the 24-hour historical data and registering parameters related to coal powder fineness includes:

[0011] The 24-hour historical window processing includes time segmentation, segment order reorganization, continuity check of the same field, and window boundary registration. The coal powder fineness related parameter registration processing includes mapping the coal powder fineness test record to the 24-hour historical window according to the sampling time, so that each segment of operating parameters carries coal powder fineness related parameters and generates a prediction input package.

[0012] Furthermore, the process of performing coal powder fineness prediction, steel ball wear rate prediction, and trigger determination includes:

[0013] The coal powder fineness prediction process includes reading the current, vibration, inlet and outlet pressure difference, temperature and coal powder fineness related parameters in the prediction input package and feeding them into the LSTM-GRU hybrid model segment by segment in time order to obtain the coal powder fineness prediction result.

[0014] The steel ball wear rate prediction process includes reading current, vibration, inlet and outlet pressure difference and continuous running time records from the same prediction input packet and sending them into the LSTM-GRU hybrid model in the same time sequence to obtain the steel ball wear rate prediction result. Then, trigger judgment processing is performed on the coal powder fineness prediction result and the steel ball wear rate prediction result.

[0015] The trigger determination process includes threshold comparison, anomaly source verification, and trigger status registration, generating a trigger determination package.

[0016] Furthermore, the process of generating mass ratios, organizing the correspondence between steel ball specifications, and mapping the channels of the five-channel material distribution valve includes:

[0017] The mass ratio generation includes generating a replenishment ratio for steel balls of different diameters participating in this replenishment based on the steel ball specification set, the allocable inventory set, and the trigger type correspondence in the ratio constraint package.

[0018] The steel ball specification correspondence organization includes establishing a fixed relationship between steel balls of different diameters and storage locations, replenishment order, and addition order;

[0019] The five-channel material distribution valve channel mapping process includes mapping each steel ball specification participating in this round of ball replenishment to an independent channel of the five-channel material distribution valve, and simultaneously registering the stepper motor-driven screw feeder and weighing sensor reading positions corresponding to the independent channel, thereby generating a mixing scheme package.

[0020] Furthermore, the process of assembling task fields, registering target floors, and registering material feeding locations includes:

[0021] The task field assembly includes assembling the mass ratio records, steel ball specification correspondence, and five-channel material distribution valve channel mapping relationship in the proportioning scheme package into a set of task fields that can be executed by the tracked laser navigation robot.

[0022] The task field set includes task number, trigger time, steel ball specification order, silo source, ball retrieval order, target floor, feeding location, channel mapping relationship, and status record bit;

[0023] The target floor registration includes registering the elevator target floor that the tracked laser navigation robot will subsequently reach, and includes the current floor, the target floor, the coal mill platform number corresponding to the target floor, and the elevator call status;

[0024] The material feeding location registration includes registering the final docking location of the tracked laser navigation robot after it completes cross-floor transfer, and includes the coal mill number, material feeding port location, material feeding port orientation, docking boundary and stopping direction, generating a transfer task package.

[0025] Furthermore, the process of performing A* algorithm path planning and DWA algorithm obstacle avoidance path generation includes:

[0026] The path planning of the A algorithm includes generating a path from the robot's current position to the warehouse ball retrieval position in the constructed map based on the warehouse ball retrieval position, target floor record, and material feeding position record;

[0027] Take the ball from the warehouse to the elevator entrance;

[0028] The path sequence from the elevator exit on the target floor to the coal mill platform feeding position;

[0029] The obstacle avoidance path generation process of the DWA algorithm includes generating a local obstacle avoidance path based on the path sequence and the obstacle positions detected in real time by the lidar.

[0030] The local obstacle avoidance path is concatenated with the path sequence.

[0031] Furthermore, the process of docking the coal mill platform with the feeding location includes:

[0032] The arrival at the coal mill platform includes, after the elevator has finished running and reached the target floor, exiting the elevator car and entering the coal mill platform passage according to the target floor record and platform direction record in the cross-floor transfer package;

[0033] Arrive at the coal mill platform docking area according to the path sequence and local obstacle avoidance path in the navigation task package;

[0034] The material feeding location docking process includes performing end-point approach, attitude adjustment, and stop locking processes based on the material feeding location record, docking direction, and docking boundary in the transfer task package, and generating a docking addition package.

[0035] Furthermore, the process of allocating the channels of the five-channel material distribution valve includes:

[0036] The five-channel material distribution valve channel allocation process includes matching the specifications of the steel balls participating in the ball replenishment with the five independent discharge channels of the five-channel material distribution valve one by one according to the channel mapping relationship in the proportioning scheme package.

[0037] Independent discharge channels that are not involved in this ball replenishment will be registered as closed.

[0038] Furthermore, the process of acquiring, comparing, calibrating, and updating the proportioning scheme of the coal mill after ball replenishment includes:

[0039] The acquisition of the mill operating parameters after ball replenishment includes reading the mill operating parameters after ball replenishment from the programmable logic controller again through the open platform communication unified architecture protocol after the completion of mass ratio addition, and binding them with the task number and completion time in the current addition result package;

[0040] The comparison calibration includes comparing the operating parameters of the coal mill after ball replenishment with the trigger judgment package, proportioning scheme package and addition result package before ball replenishment, item by item.

[0041] The comparison calibration includes comparing the actual amount of each steel ball specification added in the added result package with the mass ratio record in the proportioning scheme package;

[0042] The operating parameters of the coal mill after ball replenishment are compared with the predicted results of coal powder fineness and steel ball wear rate in the trigger judgment package;

[0043] The ratio scheme update process includes writing the current round of comparison and calibration results, the actual amount of each steel ball specification added, the coal mill operating parameters after ball replenishment, and the abnormal fluctuation correlation registration results back to the ratio scheme storage area, and updating the next round of task call boundaries according to the order of steel ball specifications to generate an operating parameter package.

[0044] Furthermore, an AI-based dynamic proportioning and robotic transfer system for steel balls in a coal mill includes: a coal mill operating parameter acquisition module, a prediction triggering module, a proportioning scheme generation module, a transfer task generation module, a navigation transfer module, a placement addition module, and a feedback calibration module; the modules are connected in sequence to implement the method described in any of the above-mentioned embodiments.

[0045] The key innovations of this invention include:

[0046] (1) Generate a prediction input package based on the coal mill operating parameters, and perform coal powder fineness prediction, steel ball wear rate prediction and trigger judgment processing to form a trigger judgment package. Then, based on the trigger judgment package, generate the mass ratio, organize the steel ball specification correspondence and map the five-channel material distribution valve channel to generate a proportioning scheme package, and directly connect the coal mill operating parameters with the dynamic proportioning of steel balls.

[0047] (2) Based on the proportioning scheme package, the task field assembly, target floor registration and feeding location registration are performed to generate a transfer task package. Then, based on the transfer task package, path planning and obstacle avoidance path generation are performed, and the connection processing of reaching the coal mill platform and feeding location is executed to form a continuous transfer link from the proportioning scheme package to the in-place addition package.

[0048] (3) Based on the in place addition package, perform five-channel material distribution valve channel allocation processing, and perform coal mill operation parameter acquisition, comparison calibration and ratio scheme update processing after ball replenishment, generate operation parameter package, and organize ball replenishment action, operation status and ratio scheme update into a feedback calibration link.

[0049] The following are its main beneficial effects:

[0050] (1) In response to the problems of insufficient dynamic proportion of steel balls, unclear correspondence of steel ball specifications and discontinuous channel mapping of the five-channel material distribution valve in the existing scheme, the continuous organization of trigger judgment package, proportioning scheme package and five-channel material distribution valve channel mapping processing is used to keep the mass ratio of steel balls of different diameters consistent with the execution channel, and the dynamic proportion of steel balls no longer stays at the level of replenishment rhythm or replenishment amount.

[0051] (2) In response to the problems of the disconnect between robot transfer and ball replenishment control and the difficulty in incorporating the target floor and feeding position into the same task chain in the existing scheme, the continuous link of transfer task package, path planning, obstacle avoidance path generation and docking processing of reaching the coal mill platform and feeding position is used to connect the dynamic proportion of steel balls with robot transfer under the same task field, and the docking process of the coal mill platform and feeding position is consistent with the previous proportion scheme.

[0052] (3) In response to the problem that the existing scheme lacks continuous calibration of the operating status after ball replenishment and that the proportioning scheme is difficult to write back to the subsequent processing link, the operating parameter package is sent back to the preceding processing link by acquiring, comparing and calibrating the operating parameters of the coal mill after ball replenishment and updating the proportioning scheme, so that a closed-loop update relationship is formed between the trigger judgment package, the proportioning scheme package and the operating parameter package.

[0053] (4) Around the continuous processing chain of coal mill operating parameters, proportioning scheme package, transfer task package, arrival addition package and operating parameter package, the present invention organizes the collection, judgment, transfer, addition and update into a consistent process, reducing the interruption caused by the separation of each link in the existing scheme.

[0054] (5) In the scenario of connecting the coal mill platform and the feeding position, this invention incorporates the generation of mass ratio, the collation of steel ball specifications, the mapping of five-channel material distribution valve channels, path planning, obstacle avoidance path generation and comparison calibration into the same method link, so that the ball replenishment action and the running state after ball replenishment are kept in correspondence, which is convenient for subsequent calls and processing. Attached Figure Description

[0055] Figure 1 A flowchart illustrating an AI-based method for dynamic proportioning and robotic transfer of steel balls in a coal mill, provided as an embodiment of this application;

[0056] Figure 2 This is a structural block diagram of an AI-based dynamic proportioning and robotic transfer system for coal mill steel balls, provided as an embodiment of this application. Detailed Implementation

[0057] Example 1: Refer to Figure 1 This is a flowchart illustrating an AI-based method for dynamic proportioning and robotic transfer of steel balls in a coal mill, provided by an embodiment of the present invention. The process may include at least steps S100-S400:

[0058] S100: Obtain the operating parameters of the coal mill, organize the 24-hour historical window and register the parameters related to coal powder fineness, generate a prediction input package, and perform coal powder fineness prediction, steel ball wear rate prediction and trigger judgment processing to generate a trigger judgment package.

[0059] S200. Based on the trigger judgment package, perform mass ratio generation, steel ball specification correspondence organization and five-channel material distribution valve channel mapping processing to generate a proportioning scheme package; based on the proportioning scheme package, perform task field assembly, target floor registration and material feeding location registration processing to generate a transfer task package;

[0060] S300. Based on the transfer task package, perform A* algorithm path planning and DWA algorithm obstacle avoidance path generation, and perform docking processing to reach the coal mill platform and feeding position to generate the in-place addition package.

[0061] S400. Based on the aforementioned in-place addition package, perform five-channel material distribution valve channel allocation processing, and execute the acquisition, comparison, calibration, and ratio scheme update processing of the coal mill operating parameters after ball replenishment to generate an operating parameter package.

[0062] Step S100 includes at least steps S110-S130:

[0063] S110. Obtain the operating parameters of the coal mill, perform field sorting and time sequence sorting to obtain the operating parameter package.

[0064] Specifically, the operating parameters of the coal mill come from the programmable logic controller (PLC) in the field control link of the coal mill and are accessed via the Open Platform Communications Unified Architecture (OPC UA) protocol.

[0065] The operating parameters of the coal mill include at least current, vibration, inlet and outlet pressure difference, and temperature. The current represents the driving load state of the coal mill, the vibration represents the operating state of the cylinder and transmission parts, the inlet and outlet pressure difference represents the changes in ventilation and internal resistance, and the temperature represents the hot operating state of the coal mill.

[0066] During acquisition, the AI ​​coal mill performance analysis module polls the PLC register area according to the sampling cycle and reads the current, vibration, inlet and outlet pressure difference and temperature in groups at the same sampling time.

[0067] When any sample item is missing, retain the same field value from the previous sampling time and register the missing flag;

[0068] When any sampling item exceeds the device's range, an anomaly flag is registered and the most recent valid value in the same field is called to participate in this round of field sorting.

[0069] Field organization specifically includes standardizing field naming, standardizing field order, registering dimensional identifiers, binding sampling time, and binding equipment source. Each record after field organization includes sampling time, equipment number, field name, field value, and status flag. Time order organization specifically includes reordering by sampling time in ascending order, merging duplicate sampling times, removing duplicates of abnormal times, and registering breakpoint time periods, so that current, vibration, inlet / outlet pressure difference, and temperature on the same time axis are arranged continuously in a unified order.

[0070] After processing, an operating parameter package is generated. The operating parameter package is recorded as an output field name and sent to the "operating parameter package" call in S120. At the same time, it serves as the feedback reference basis for subsequent acquisition, comparison and calibration of coal mill operating parameters and updating of proportioning scheme after ball replenishment.

[0071] S120. Based on the operating parameter package, perform 24-hour historical window sorting and coal powder fineness related parameter registration processing to generate a prediction input package.

[0072] Specifically, the operating parameter package serves as the input source for this step. The AI ​​coal mill performance analysis module traces back 24 hours from the current sampling time, extracts continuous operating segments, and forms a 24-hour historical window.

[0073] When there is a downtime segment, maintenance segment, or manual data entry segment within the window, the time period will be registered separately, the window position will be retained but it will not be included in the continuous segment splicing;

[0074] If the data at the beginning of the window is less than 24 hours old, call the most recent complete runtime segment to complete it and register the source of the completion.

[0075] The 24-hour historical window processing includes time segmentation, segment order reorganization, continuity check of the same field, and window boundary registration, thereby obtaining a historical input sequence suitable for subsequent coal powder fineness prediction and steel ball wear rate prediction.

[0076] The registration and processing of coal powder fineness-related parameters are performed based on the 24-hour historical window. The coal powder fineness-related parameters include coal powder fineness inspection records, coal powder fineness sampling time, corresponding operating parameter segments, and current shift markers. When there are multiple sources for coal powder fineness inspection records, the online record corresponding to the sampling time of the coal mill operating parameters is called first. If there is no online record, the most recent test record is called, and the alternative source is registered.

[0077] Furthermore, the coal powder fineness inspection records are mapped to the corresponding positions in the 24-hour historical window according to the sampling time, so that each segment of operating parameters carries the registration results of coal powder fineness-related parameters. After processing, a prediction input package is generated. The prediction input package contains the organized 24-hour historical window, coal powder fineness-related parameters, missing markers, anomaly markers, and window boundary records. This prediction input package is recorded as an output field name and sent to the "prediction input package" call in S130. At the same time, the window boundary records and coal powder fineness-related parameters in the prediction input package provide a preliminary basis for the trigger judgment package source in S200.

[0078] S130. Based on the predicted input package, perform coal powder fineness prediction, steel ball wear rate prediction and trigger determination processing to generate a trigger determination package.

[0079] Specifically, the predicted input package serves as the input source for this step. The AI ​​coal mill performance analysis module calls the LSTM-GRU hybrid model for continuous processing. The LSTM-GRU hybrid model consists of a long short-term memory network and a gated recurrent unit. The front end receives the time-sequential data from the 24-hour historical window, and the back end receives coal powder fineness-related parameters and anomaly markers, thereby outputting the coal powder fineness prediction result and steel ball wear rate prediction result corresponding to the current operating segment.

[0080] When processing the coal powder fineness prediction, the current, vibration, inlet and outlet pressure difference, temperature, and coal powder fineness-related parameters in the prediction input package are first read, and then sent into the LSTM-GRU hybrid model segment by segment in chronological order to obtain the coal powder fineness prediction result corresponding to the current sampling time. When processing the steel ball wear rate prediction, the current, vibration, inlet and outlet pressure difference, temperature, and continuous running time records in the same prediction input package are read and sent into the LSTM-GRU hybrid model in the same chronological order to obtain the steel ball wear rate prediction result corresponding to the current sampling time.

[0081] The trigger determination process is executed after the coal powder fineness prediction results and steel ball wear rate prediction results are generated. Specifically, it includes threshold comparison, anomaly source verification, and trigger status registration. When the coal powder fineness prediction result reaches the coal powder fineness exceeding the standard condition, or the steel ball wear rate prediction result reaches the steel ball wear rate trigger condition, the trigger status, trigger type, and trigger time are registered. When both conditions are met simultaneously, the joint trigger status is registered and the trigger is promoted to the priority processing queue.

[0082] Furthermore, the predicted results of coal powder fineness, steel ball wear rate, trigger status, trigger type, trigger time, and source window number are written into the same data carrier to generate a trigger determination package. This trigger determination package is recorded as an output field name and is called by S210's "trigger determination package." Based on this, S210 performs registration of steel balls of different diameters, steel ball inventory registration, and proportion constraint sorting. Subsequently, S220 and S230 form proportion scheme packages and transfer task packages around this trigger determination package, thus maintaining a continuous transmission relationship between S100 and S200, S300, and S400.

[0083] In summary, this step consolidates the operating parameters and coal powder fineness-related parameters from the 24-hour historical window into a single predictive input path. It then merges the predicted coal powder fineness and steel ball wear rate into a single trigger decision package, rewriting the processing chain that was previously triggered only by a single ball replenishment signal. This processing method ensures that the subsequent dynamic steel ball proportioning scheme generation directly corresponds to the current operating state and steel ball wear state, providing a unified input source for proportioning constraint processing and transfer task assembly.

[0084] Step S200 includes at least steps S210-S230:

[0085] S210. Based on the trigger determination package, perform registration of steel balls of different diameters, registration of steel ball inventory, and sorting of proportion constraints to obtain a proportion constraint package.

[0086] Specifically, the trigger determination package originates from S130, and includes at least the coal powder fineness prediction result, the steel ball wear rate prediction result, the trigger status, the trigger type, and the trigger time. S210 is executed by the proportioning constraint processing unit in the AI ​​coal mill performance analysis module, which is connected to the steel ball inventory registration unit, the steel ball specification registration unit, and the proportioning rule registration unit.

[0087] The registration of steel balls of different diameters refers to the unified registration of the specifications of steel balls used for replenishing coal mills in the warehouse. The steel balls of different diameters shall include at least the specification identification of the corresponding particle size, the location of the warehouse, the current inventory status and the availability status.

[0088] In one engineering embodiment, five types of steel ball storage locations are pre-set on the warehouse side. Each type of steel ball storage location corresponds to a steel ball specification. The entrance of the storage location is equipped with a specification nameplate and an inventory collection terminal. The inventory collection terminal can be composed of a weighing device, a manual input terminal, or a storage location counting component. After receiving the trigger judgment package, the AI ​​coal mill performance analysis module retrieves the steel ball specification registration table for the current shift and performs specification name verification, storage location verification, and replenishment availability verification for each type of steel ball.

[0089] The steel ball inventory registration refers to registering the current transferable quantity of steel balls of each specification. The registration content includes the current inventory quantity, minimum inventory quantity, frozen inventory quantity, and allocable inventory quantity. The current inventory quantity corresponds to the number of steel balls currently in the warehouse, the minimum inventory quantity corresponds to the reserved inventory boundary, the frozen inventory quantity corresponds to the inventory quantity that has been occupied by the previous task and will not be called up temporarily, and the allocable inventory quantity corresponds to the inventory quantity that can be used in the calculation of this allocation scheme.

[0090] When the available inventory of a certain specification of steel ball is lower than the minimum inventory, the steel ball of that specification is registered as restricted and its recall priority is reduced in subsequent allocation constraint processing.

[0091] Furthermore, the ratio constraint sorting process is performed after the registration of steel balls of different diameters and the registration of steel ball inventory are completed. Specifically, it includes trigger type correspondence matching, steel ball specification boundary matching, and inventory boundary matching. Trigger type correspondence matching refers to reading the coal powder fineness prediction result and steel ball wear rate prediction result from the trigger determination package, and mapping the coal powder fineness exceeding the standard state and the steel ball wear rate trigger state to the calling relationship of steel balls of different diameters. Steel ball specification boundary matching refers to filtering the registered steel balls of different diameters by specification group and retaining the specification group that meets the current trigger state. Inventory boundary matching refers to checking the allocable inventory quantity against the current trigger state item by item, and eliminating the specification group in the restricted state or registering it as a standby calling state.

[0092] Understandably, if the predicted results of coal powder fineness and steel ball wear rate both reach the trigger condition at the same time, joint registration will be performed on steel balls of different diameters at the same trigger time, and synchronous verification will be performed on the inventory boundary; if there is only a single trigger type, the call boundary will be generated according to the single trigger type.

[0093] After processing, a proportion constraint package is obtained. The proportion constraint package is recorded as an output field name. The proportion constraint package contains at least a set of steel ball specifications, a set of warehouse locations, a set of allocable inventory, a set of restricted states, and a correspondence of trigger types. It is used by the "proportion constraint package" in S220 and provides a source of pre-constraints for the warehouse ball retrieval parsing in S300.

[0094] S220. Based on the aforementioned proportion constraint package, perform mass ratio generation, steel ball specification correspondence organization, and five-channel material distribution valve channel mapping processing to generate a proportion scheme package.

[0095] Specifically, the proportioning constraint package originates from S210, and S220 is executed by the proportioning scheme generation unit in the AI ​​coal mill performance analysis module. The proportioning scheme generation unit is connected to the five-channel material distribution valve configuration unit. The mass ratio generation refers to generating the replenishment ratio for steel balls of different diameters participating in this replenishment based on the steel ball specification set, the allocable inventory set, and the trigger type correspondence in the proportioning constraint package.

[0096] The mass ratio refers to the distribution relationship of steel balls of various specifications in this ball replenishment task. The recorded content includes specification identification, ratio order, ratio value, corresponding storage location, and call status.

[0097] Specifically, the steel ball specifications in an allocable state are first extracted from the proportioning constraint package. Then, the specification group that should participate in replenishment is read according to the trigger type correspondence. Proportional allocation is then performed within the specification group. When a steel ball of a certain specification is in a restricted state, the proportioning scheme generation unit retains that specification in the proportioning record and registers it as restricted, preventing it from participating in this round of mass proportion generation. The steel ball specification correspondence is organized after the mass proportion is generated. This correspondence establishes a fixed relationship between steel balls of different diameters and their storage locations, replenishment order, and addition order.

[0098] Furthermore, for each steel ball specification, its storage location, target replenishment sequence, material distribution valve access channel, and weighing reading location are recorded separately to prevent specification confusion during subsequent transfer and addition of the mixing scheme. When there are multiple storage locations for the same specification of steel ball, the storage location with the shorter transfer path to the current target floor is registered first, and the remaining storage locations are registered as backup sources.

[0099] The five-channel dispensing valve channel mapping process is performed after the steel ball specification correspondence is organized. The five-channel dispensing valve is a dispensing execution component mounted on the robot, containing five independent channels, each corresponding to a steel ball addition path. During the mapping process, the steel ball specification correspondence is first read, and then each steel ball specification participating in this round of replenishment is mapped to an independent channel, while simultaneously registering the reading positions of the stepper motor-driven screw feeder and the weighing sensor corresponding to that independent channel. When the number of steel ball specifications participating in this round of replenishment is less than five, the empty channels are registered as closed. When the number of steel ball specifications participating in this round of replenishment reaches five, they are mapped sequentially according to the steel ball specification correspondence.

[0100] Understandably, the five-channel material distribution valve channel mapping process does not change the steel ball specifications themselves, but rather transforms the mass ratio generation result into a subsequently executable channel configuration relationship. After processing, a proportioning scheme package is generated, which is recorded as an output field name. The proportioning scheme package includes at least the mass ratio record, the steel ball specification correspondence, the five-channel material distribution valve channel mapping relationship, the stepper motor driven screw feeder correspondence, and the weighing sensor reading attribution relationship, and is called by the "proportioning scheme package" in S230. At the same time, the channel mapping relationship in the proportioning scheme package will be called in S410 for the five-channel material distribution valve channel allocation, the stepper motor driven screw feeder startup, and the weighing sensor reading processing.

[0101] S230. Based on the aforementioned proportioning scheme package, perform task field assembly, target floor registration, and material feeding location registration to generate a transfer task package.

[0102] Specifically, the proportioning scheme package comes from S220, and S230 is executed by the task assembly unit in the robot control system. The task assembly unit is connected to the target floor registration unit, the feeding position registration unit, and the transfer status recording unit.

[0103] The task field assembly refers to assembling the mass ratio records, steel ball specification correspondences, and five-channel material distribution valve channel mapping relationships from the proportioning scheme package into a set of task fields that can be directly executed by the tracked laser navigation robot. Each task field includes at least the task number, trigger time, steel ball specification sequence, bin source, ball retrieval sequence, target floor, material feeding location, channel mapping relationship, and status record bit. During task field assembly, each steel ball specification and its corresponding bin source are first read from the proportioning scheme package. Then, the ball retrieval sequence and channel mapping relationship corresponding to that steel ball specification are written. Finally, the trigger time and status record bit for this ball replenishment task are written, forming a complete task field set.

[0104] When multiple coal mill ball replenishment requests exist at the same trigger time, the task assembly unit sorts the task fields according to the order of trigger time and floor affiliation, and the task field corresponding to the current coal mill is written with the current task number first.

[0105] The target floor registration refers to registering the elevator target floor that the tracked laser navigation robot will subsequently reach. The registration content includes the current floor, the target floor, the coal mill platform number corresponding to the target floor, and the elevator call status.

[0106] In one engineering embodiment, the warehouse is located on the ground floor, and the coal mill platform is located on the upper equipment platform. After reading the proportioning scheme package, the task assembly unit first parses the target floor from the coal mill platform number, and then writes the target floor into the transfer control field for subsequent target floor sending and door opening signal reception processing by S320. The feeding position registration refers to registering the final docking position after the tracked laser navigation robot completes cross-floor transfer. The registration content includes the coal mill number, feeding port position, feeding port orientation, docking boundary, and stopping direction. When registering the feeding position, the task assembly unit reads the current coal mill fixed position record and feeding port configuration record, binds the feeding port position with the current task number, and writes the docking boundary and stopping direction into the task field set, so that S330 can directly perform the entry into the car, arrival at the coal mill platform, and feeding position docking processing based on the feeding position registration content. Furthermore, after the task field assembly, target floor registration, and material feeding location registration are completed, the transfer status recording unit adds a pending execution mark to this task and archives the pending execution mark along with the task number, target floor, and material feeding location to form a traceable transfer record. After processing, a transfer task package is generated, which is recorded as an output field name. The transfer task package contains at least a set of task fields, a target floor record, a material feeding location record, and a pending execution mark, and is called by S310's "transfer task package". Subsequently, S310 will perform warehouse ball retrieval position analysis, A* algorithm path planning, and DWA algorithm obstacle avoidance path generation based on this. S320 and S330 will then continue to complete cross-floor transfer and docking along the transfer task package.

[0107] Summary of the technical effects of this step: This step further refines the trigger decision package generated by S130 to execution fields such as steel ball specifications, steel ball inventory, mass ratio, five-channel material distribution valve channel mapping, target floor, and feeding position. It rewrites the processing method that only outputs replenishment decision values ​​without forming a complete transfer task chain. Through the continuous generation of the aforementioned proportioning constraint package, proportioning scheme package, and transfer task package, a direct connection is formed between the dynamic proportioning of steel balls and the transfer by the tracked laser navigation robot. Subsequent on-site addition processing can directly call the aforementioned task fields.

[0108] Step S300 includes at least steps S310-S330:

[0109] S310. Obtain the transfer task package, perform warehouse ball retrieval position analysis, A* algorithm path planning and DWA algorithm obstacle avoidance path generation processing to obtain the navigation task package.

[0110] Specifically, the transfer task package originates from S230, and includes at least a task field set, target floor record, material feeding location record, and execution marker. S310 is executed by the navigation task generation unit in the robot control system, which is connected to the warehouse location database, map building unit, path planning unit, and obstacle avoidance control unit. The warehouse ball retrieval position analysis refers to resolving the warehouse location, stopping direction, and ball retrieval order that the tracked laser navigation robot needs to enter from the steel ball specification order, warehouse location source, and ball retrieval order in the transfer task package.

[0111] Specifically, the aforementioned transfer task package is used as input. Through task field reading and processing, the warehouse location source corresponding to the steel ball specification is first extracted. Then, the warehouse location source is matched with the ball retrieval coordinates, channel width, turning boundary and parking orientation in the warehouse location database to obtain the warehouse ball retrieval location corresponding to the current task. When there is a backup source for the same steel ball specification, the main warehouse location source is called first. If the main warehouse location source is occupied or the channel is blocked, the backup source is switched and the switching result is recorded as a path adjustment mark.

[0112] The A* algorithm path planning refers to the process by which the path planning unit generates a sequence of paths in the constructed map, from the robot's current position to the warehouse ball-retrieving position, from the warehouse ball-retrieving position to the elevator entrance, and from the elevator exit of the target floor to the coal mill platform feeding position, based on the records of the warehouse ball-retrieving position, the target floor, and the feeding position. The A* algorithm is also known as A* Algorithm.

[0113] The map building unit calls the map data collected by the lidar and industrial inertial measurement unit (IMU) carried by the tracked laser navigation robot to form a high-precision map including warehouse passages, equipment passages, elevator entrances and coal mill platforms. The path planning unit then extracts the starting point, turning point and ending point from the high-precision map to generate a path sequence arranged in order.

[0114] The obstacle avoidance path generation process using the DWA algorithm is executed after the A* algorithm path planning. The DWA algorithm stands for Dynamic Window Approach. The obstacle avoidance control unit generates a local obstacle avoidance path based on the path sequence and the obstacle positions detected in real-time by the lidar, and then splices the local obstacle avoidance path with the aforementioned path sequence. When temporary material stockpiles, personnel occupying positions, or narrowing of the turning area are detected, the obstacle avoidance control unit performs detour adjustments to the local path and records the detour start point, detour end point, and detour reason. If the path still cannot be passed after detour, the navigation task generation unit outputs a mark indicating manual verification is required. Further, in one engineering embodiment, the warehouse is located in the north passageway on the ground floor, the elevator entrance is located in the central equipment passageway, and the coal mill platform is located on the east platform on the upper floor. The navigation task generation unit first reads the warehouse ball-retrieving position, then generates the first path segment to the warehouse ball-retrieving position, the second path segment to the elevator entrance, and the third path segment to the coal mill platform feeding position according to a high-precision map. The local obstacle avoidance path generated by the DWA algorithm is then superimposed on the second and third paths, respectively.

[0115] After processing, a navigation task package is obtained and recorded as an output field name for S320 to call. At the same time, the path sequence, local obstacle avoidance path, path adjustment mark and manual review mark in the navigation task package are continuously available for S330 to call for docking processing when entering the car, arriving at the coal mill platform and feeding position.

[0116] S320. Based on the navigation task package, perform arrival at elevator entrance, target floor sending and door opening signal receiving processing to generate a cross-floor transfer package.

[0117] Specifically, the navigation task package originates from S310, and S320 is executed by the cross-floor scheduling unit in the robot control system. This cross-floor scheduling unit is connected to the tracked laser navigation robot body, the elevator controller, the wireless communication unit, and the status feedback unit. Reaching the elevator entrance refers to the tracked laser navigation robot retrieving a ball from the warehouse and moving towards the elevator entrance parking area according to the path sequence and local obstacle avoidance path in the navigation task package, and performing position correction after entering the parking area.

[0118] During the position correction process, the tracked laser navigation robot uses LiDAR to identify the elevator entrance boundary, uses an industrial-grade IMU to correct the current orientation, and then makes end-effector adjustments based on the elevator entrance stopping position and stopping direction in the navigation task package. When there is a temporary obstacle in front of the elevator entrance, it first calls a local obstacle avoidance path to enter a waiting position, and then performs position correction after the obstacle is removed.

[0119] The target floor transmission refers to the process whereby, after the tracked laser navigation robot arrives at the elevator entrance, the cross-floor scheduling unit extracts the target floor record from the navigation task package and transmits it to the elevator controller via the wireless communication unit. The wireless communication unit can use a fourth-generation mobile communication module or a Modbus TCP protocol link, where Modbus TCP stands for Transmission Control Protocol over Modbus.

[0120] The transmitted content must include at least the task number, current floor, target floor, and robot call status. The door opening signal reception refers to the cross-floor scheduling unit continuously monitoring the elevator status signal returned by the elevator controller after sending the signal to the target floor. Upon receiving a status signal indicating that the elevator has arrived and the car door is open, the unit writes this status signal into the current transfer record. If no door opening signal is received within the predetermined waiting period, the cross-floor scheduling unit retransmits the target floor record and records the number of retransmissions as a scheduling anomaly. If the door opening signal is still not received after continuous retransmissions, a "pending manual review" flag is output, and the process is paused.

[0121] Furthermore, in the engineering embodiment, after retrieving the ball from the warehouse, the tracked laser-guided robot proceeds to the elevator entrance according to the navigation task package. It decelerates at a preset distance from the elevator stop line, corrects its direction after entering the stopping area, and then sends the target floor to the elevator controller. It then waits for the return door opening signal. When the elevator door opens and the interior space is in an enterable state, the cross-floor scheduling unit simultaneously writes the enterable state and task number into the transfer record. After processing, a cross-floor transfer package is generated and recorded as an output field name for use in S330's "Cross-floor Transfer Package" function. Simultaneously, the elevator status signal, enterable state, scheduling anomaly flag, and transfer record in the cross-floor transfer package continue to be used in S400 for adding execution pre-state verification.

[0122] S330. Based on the cross-floor transfer package, perform docking processing for entering the car, arriving at the coal mill platform and feeding position, and generate the in-place addition package.

[0123] Specifically, the inter-floor transfer package originates from S320, and S330 is executed by the docking unit in the robot control system. The docking unit is connected to the drive control unit, platform positioning unit, and material feeding position correction unit. Entering the elevator car refers to the tracked laser navigation robot reading the enterable status in the inter-floor transfer package, then driving at low speed into the elevator car according to the elevator entrance stopping direction and the car entry direction in the inter-floor transfer package, and stopping after all tracks have entered the car. During the entry process, the drive control unit continuously reads the attitude data from the lidar and industrial-grade IMU to verify whether the vehicle body deviates from the center channel of the car. If the deviation exceeds the preset channel boundary, a mid-journey correction is performed. If a temporary occupancy occurs in the car, the robot immediately stops and waits for the occupancy to be cleared.

[0124] The arrival at the coal mill platform refers to the process after the elevator has finished running and reached the target floor. The tracked laser navigation robot reads the target floor record and platform direction record from the cross-floor transfer package, drives out of the elevator car and enters the coal mill platform passage, and then reaches the coal mill platform docking area according to the third segment path and local obstacle avoidance path in the navigation task package.

[0125] After the platform positioning unit arrives at the coal mill platform docking area, it uses lidar to re-identify the platform boundary, equipment outline and channel opening position, and compares the identification results with the aforementioned feeding position record;

[0126] When there is a deviation between the platform boundary recognition result and the feeding position record, the feeding position correction unit is invoked to perform position correction. The feeding position docking refers to the process by which the tracked laser navigation robot, after arriving at the coal mill platform docking area, performs end-effector approach, attitude adjustment, and stop locking according to the feeding position record, docking direction, and docking boundary in the transfer task package. Specifically, it first moves towards the coal mill feeding port according to the orientation in the feeding position record, then gradually shortens the distance according to the docking boundary, stops driving after reaching the predetermined docking position, and writes the current vehicle attitude, the relative position of the feeding port, and the channel mapping relationship into the docking record; if there is an obstacle in front of the feeding port or the position recognition result does not meet the docking boundary, it first retreats to the adjustment position and then re-executes the end-effector approach.

[0127] Furthermore, in one engineering embodiment, after the tracked laser navigation robot completes its entry into the elevator car, it exits the elevator car at the target floor, travels along the east side passage of the coal mill platform to the front of the designated coal mill feeding port, and then completes position correction at low speed and stops at the docking position. At this time, the five-channel material distribution valve on the robot faces the feeding port, and the stepper motor drives the screw feeder and weighing sensor to enter the ready-to-start state.

[0128] After processing, a position addition package is generated and recorded as an output field name for S410 to call. The position addition package includes at least position records, feeding position docking status, channel mapping relationship, waiting to start status and position correction records, so that S410 can directly execute the five-channel material distribution valve channel allocation, stepper motor driven screw feeder start and weighing sensor reading processing.

[0129] Summary of the technical effects of this step: This step further refines the transfer task package generated by S230 into a navigation task package, a cross-floor transfer package, and a location addition package, enabling the tracked laser navigation robot to form a continuous transfer chain from the warehouse ball retrieval position, elevator entrance, elevator car to the coal mill platform feeding position. Compared to processing methods that only describe handling or only describe the ball replenishment mechanism, this step continuously writes A* algorithm path planning, DWA algorithm obstacle avoidance path generation, target floor sending, door opening signal reception, and feeding position docking into the same execution chain. Subsequently, the five-channel material distribution valve and the stepper motor-driven screw feeder can directly enter the addition process according to the arrival status.

[0130] Step S400 includes at least steps S410-S430:

[0131] S410. Based on the aforementioned addition package, perform five-channel material distribution valve channel allocation, stepper motor-driven screw feeder startup, and weighing sensor reading processing to obtain the addition execution package.

[0132] Specifically, the in-place addition package comes from S330, and the in-place addition package includes at least an in-place record, a feeding position docking status, a channel mapping relationship, a pending start status, and a position correction record.

[0133] S410 is accomplished by the execution control unit in the precision steel ball proportioning and adding module. The execution control unit is connected to the five-channel material distribution valve, the stepper motor-driven screw feeder, the weighing sensor, and the robot body controller.

[0134] The five-channel material distribution valve is a material distribution component installed on the tracked laser navigation robot. It includes five independent discharge channels, each of which corresponds to the passage path of a steel ball specification. The channel allocation refers to matching the steel ball specifications participating in the current ball replenishment with the five independent discharge channels one by one according to the channel mapping relationship formed by S220, and registering the independent discharge channels that do not participate in the current ball replenishment as closed.

[0135] Specifically, the aforementioned in-place addition package is used as input. The control unit reads the channel mapping relationship and feeding position docking status. First, the opening and closing checks of each channel of the five-channel material distribution valve are performed. Then, the specifications of the steel balls participating in this ball replenishment are written into the control position of the corresponding channel.

[0136] If the control bit of a certain independent discharge channel is inconsistent with the channel mapping relationship, the channel is first reset, then the channel allocation result is rewritten, and the reset result is recorded as a channel correction record.

[0137] The start-up of the stepper motor-driven screw feeder refers to the execution control unit sending a start command to each stepper motor-driven screw feeder corresponding to each independent discharge channel after completing the allocation of the five-channel material distribution valve. The stepper motor-driven screw feeder is used to feed the corresponding steel ball specifications to the feeding port along the independent discharge channel. During start-up, the state to be started is checked first, and then the docking state of the feeding position is checked. When both are in an executable state, the corresponding stepper motor-driven screw feeder is started one by one according to the steel ball specifications.

[0138] Furthermore, in the engineering embodiment, after the tracked laser-guided robot arrives at the feeding position on the coal mill platform, the execution control unit first reads the position correction record in the addition package to confirm that the robot body is within the docking boundary of the feeding position. Then, it sequentially opens the independent discharge channels involved in ball replenishment and starts the corresponding stepper motor-driven screw feeder according to the steel ball specifications. If any channel fails to respond during the startup process, the startup of subsequent channels is stopped and a startup anomaly flag is registered. The weighing sensor reading processing is performed after the stepper motor-driven screw feeder is started. The weighing sensor is set in the material distribution execution path to record the current addition amount of each steel ball specification. During reading processing, the execution control unit reads the current value, cumulative value, and reading time of the weighing sensor one by one according to the independent discharge channel and binds the reading result with the corresponding steel ball specification. When the reading of a certain weighing sensor is interrupted or remains unchanged, the execution control unit registers the channel as an abnormal reading state and suspends the subsequent reading writing of that channel. After processing, an additional execution package is obtained and recorded as an output field name. The additional execution package includes at least the channel allocation result, the start status of the stepper motor driven screw feeder, the weighing sensor reading result, the channel correction record, and the start-up anomaly flag. The additional execution package is called by the "Add Execution Package" function of S420. At the same time, the weighing sensor reading result and the channel allocation result are continued to be called by the comparison calibration and proportioning scheme update processing in S430.

[0139] S420. Based on the added execution package, perform mass ratio addition, steel ball stacking height detection, and pause addition alarm processing to generate an addition result package.

[0140] Specifically, the addition execution package originates from S410, and S420 is completed by the addition scheduling unit in the precise steel ball proportioning addition module. The addition scheduling unit is connected to the five-channel material distribution valve, the stepper motor-driven screw feeder, the weighing sensor, and the infrared sensor. The mass proportion addition refers to controlling each independent discharge channel to deliver steel balls to the coal mill feed inlet according to the corresponding proportion based on the mass proportion record formed in S220 and the channel allocation result formed in S410. Specifically, the channel allocation result and the weighing sensor reading result are extracted from the addition execution package, and then the corresponding mass proportion record in the proportioning scheme package is read. Addition control is performed on each independent discharge channel according to the steel ball specification order. During the addition control process, the addition scheduling unit continuously compares the current weighing sensor reading with the corresponding mass proportion record for each independent discharge channel. When the current reading reaches the addition amount that the channel should complete, the corresponding independent discharge channel is closed, and the stepper motor-driven screw feeder for that channel is stopped.

[0141] Understandably, the mass ratio addition is not a single total addition, but rather an addition process that is sequentially controlled, reading-controlled, and channel-controlled for different steel ball specifications. The steel ball stacking height detection is performed synchronously during the mass ratio addition process. The steel ball stacking height refers to the current height of the steel balls stacked within the feeding port area, and the detection component is an infrared sensor located at the feeding port.

[0142] Specifically, the infrared sensor continuously monitors the passage and accumulation status of the steel balls in front of the feeding port and sends the detection results back to the addition scheduling unit. When the infrared sensor detects that the steel balls are passing through normally, the addition scheduling unit maintains the current addition control. When the infrared sensor detects that the steel balls in front of the feeding port are stagnant or the accumulation height is rising continuously, the addition scheduling unit registers the current channel as an accumulation monitoring state.

[0143] The pause-addition alarm process is executed when the steel ball stacking height reaches the preset boundary. Specifically, it includes pausing the current channel addition, stopping the corresponding stepper motor-driven screw feeder, closing the corresponding independent discharge channel, and outputting an alarm signal. If multiple independent discharge channels share the same feeding port area, when one channel shows a stacking monitoring state, other started channels are paused simultaneously, and the reason for the pause is written into the alarm record.

[0144] Furthermore, in an engineering embodiment, when a certain steel ball specification forms a short-term accumulation in front of the feeding port, and the infrared sensor reports an abnormal accumulation height, the addition scheduling unit immediately stops the stepper motor-driven screw feeder of the corresponding independent discharge channel and closes that channel, then synchronously pauses other started channels; after the feeding port area returns to normal, the addition scheduling unit reads the weighing sensor readings before the pause and continues to add the remaining material according to the proportion of unfinished mass. After processing, an addition result package is generated and recorded as an output field name. The addition result package includes at least the actual addition amount of each steel ball specification, completion status, accumulation height detection result, pause addition record, and alarm record. The addition result package is called by the "addition result package" in S430, and the actual addition amount and completion status of each steel ball specification in the package are also called by the operation parameter feedback calibration chain in S100.

[0145] S430. Based on the added result package, perform the acquisition, comparison, calibration, and ratio scheme update processing of the coal mill operating parameters after ball replenishment to generate an operating parameter package.

[0146] Specifically, the added result package comes from S420, and S430 is completed by the feedback calibration unit in the AI ​​coal mill performance analysis module. The feedback calibration unit is connected to the PLC, OPC UA protocol interface, proportioning scheme storage area and operating parameter storage area.

[0147] The acquisition of the mill operating parameters after ball replenishment refers to the process of reading the mill operating parameters after ball replenishment from the PLC again via the OPC UA protocol after the mass ratio of each steel ball specification has been added or after the remaining addition has been resumed following a pause. The mill operating parameters after ball replenishment include at least current, vibration, inlet and outlet pressure difference, and temperature, and are bound to the task number and completion time in the current addition result package.

[0148] Specifically, the aforementioned addition result package is used as input, and the completion status and completion time are read through the feedback calibration unit. When the conditions for collecting the operating parameters after ball replenishment are met, the PLC is started to read them. During the collection, the operating parameters of the coal mill after ball replenishment are sorted by field, time sequence and task number, so that the operating parameters of the coal mill after ball replenishment correspond one-to-one with the actual amount of steel balls added.

[0149] The comparison calibration is performed after the operating parameters of the coal mill are obtained after ball replenishment. The comparison calibration refers to comparing the operating parameters of the coal mill after ball replenishment with the trigger judgment package, proportioning scheme package and addition result package before ball replenishment item by item.

[0150] Furthermore, the actual addition amount and completion status of each steel ball specification are extracted from the addition result package; the coal powder fineness prediction result and steel ball wear rate prediction result are extracted from the trigger judgment package; and the mass ratio record and channel mapping relationship are extracted from the proportioning scheme package. The operating parameters of the coal mill after ball replenishment are compared before and after the comparison. When there is a deviation between the actual addition amount of a certain steel ball specification and the corresponding mass ratio record, the deviation is recorded as the comparison calibration result of this round. When the operating parameters of the coal mill fluctuate abnormally after ball replenishment, the abnormal fluctuation is associated with the pause addition record and alarm record.

[0151] The ratio scheme update process is performed after the comparison calibration. Specifically, the comparison calibration results of this round, the actual amount of each steel ball specification added, the operating parameters of the coal mill after replenishing the balls, and the abnormal fluctuation correlation registration results are written back to the ratio scheme storage area, and the next round of task call boundaries are updated according to the order of steel ball specifications.

[0152] If the actual addition amount of a certain steel ball specification deviates in multiple consecutive rounds of tasks, its calling order will be reduced in the next round of proportion constraint sorting.

[0153] If a single discharge channel experiences paused data additions in multiple consecutive task rounds, that single discharge channel will be registered as a priority review channel in the next round of channel mapping processing.

[0154] Understandably, the ratio scheme update process does not operate independently of the aforementioned trigger judgment package, ratio constraint package, and ratio scheme package. Instead, it rewrites the actual addition results of this round back to the operating parameter storage area and the ratio scheme storage area, so that the operating parameter package read in the next round S110, S120, and S130 includes the continuous operating status after ball replenishment. Further, in an engineering embodiment, after this round of ball replenishment is completed, the feedback calibration unit reads the current, vibration, inlet / outlet pressure difference, and temperature within a preset acquisition period, binds them with the actual addition amount of each steel ball specification in this round, paused addition records, and alarm records, and then generates an updated operating parameter package. After processing, a running parameter package is generated and recorded as an output field name. The running parameter package is called by the "running parameter package" of S120, thereby forming a closed loop chain of S110, S120, S130, S210, S220, S230, S310, S320, S330, S410, S420, and S430.

[0155] Summary of the technical effects of this step: This step connects the five-channel material distribution valve channel allocation, the stepper motor-driven screw feeder startup, the weighing sensor reading, the infrared sensor accumulation detection, and the feedback of the coal mill's operating parameters after ball replenishment into a single execution chain, rewriting the processing method that only completes the ball replenishment action without writing back the operating status. By continuously transmitting the added execution package, the added result package, and the updated operating parameter package, the dynamic steel ball ratio, the timely addition, and the next round of trigger determination form a closed loop, and subsequent ratio scheme updates no longer depend on a fixed ball replenishment rhythm.

[0156] Example 2: Figure 2 This diagram illustrates a structural block diagram of an AI-based dynamic proportioning and robotic transfer system for steel balls in a coal mill, according to an embodiment of the present invention. Figure 2 As shown, the structure may include:

[0157] The coal mill operating parameter acquisition module 01 is used to acquire coal mill operating parameters and organize these parameters by field and time sequence to obtain an operating parameter package. Specifically, the coal mill operating parameter acquisition module is connected to the coal mill field control link and receives coal mill operating parameters from the programmable logic controller. These operating parameters include at least current, vibration, inlet / outlet pressure difference, and temperature. The coal mill operating parameter acquisition module performs field name unification, sampling time binding, equipment source binding, and status flag writing processing on the collected data. It then performs duplicate record removal, breakpoint record registration, and abnormal record retention processing according to the sampling time sequence, forming an operating parameter package arranged in a unified order. The operating parameter package contains the sampled value, sampling time, equipment source, and status flag corresponding to each field. The coal mill operating parameter acquisition module transmits the operating parameter package to the prediction trigger module for use as the operating parameter package, and retains the sampling records within this module for subsequent data correspondence after the calibration module generates the operating parameter package.

[0158] The prediction trigger module 02 receives the operating parameter package, performs 24-hour historical window processing and registers coal powder fineness-related parameters in the operating parameter package, generates a prediction input package, and performs coal powder fineness prediction, steel ball wear rate prediction, and trigger determination on the prediction input package to generate a trigger determination package. Specifically, the prediction trigger module receives the operating parameter package from the coal mill operating parameter acquisition module, extracts continuous operating records from the operating parameter package, and performs 24-hour historical window interception, window boundary registration, shutdown segment marking, and abnormal segment marking processing on the continuous operating records to form historical window data corresponding to the current sampling time. The prediction trigger module also performs coal powder fineness-related parameter registration on the historical window data, writes the coal powder fineness inspection record, sampling time, and corresponding operating segment into the same data carrier to generate a prediction input package. Subsequently, the prediction trigger module calls the internal prediction processing unit to sequentially perform coal powder fineness prediction, steel ball wear rate prediction, and trigger status comparison on the prediction input package, and writes the coal powder fineness prediction result, steel ball wear rate prediction result, trigger type, and trigger time into the trigger determination package. The prediction triggering module transmits the trigger determination package to the ratio scheme generation module for use as a trigger determination package, and retains the correspondence between the prediction input package and the trigger determination package for use by the feedback calibration module.

[0159] The proportioning scheme generation module 03 receives the trigger judgment package, registers steel balls of different diameters, registers steel ball inventory, and organizes proportioning constraints to obtain a proportioning constraint package. It then generates a proportioning scheme package by generating a mass ratio, organizing the steel ball specification correspondence, and mapping the five-channel material distribution valve channels. Specifically, the proportioning scheme generation module receives the trigger judgment package from the prediction trigger module and retrieves the steel ball specification records and inventory records corresponding to the current shift based on the trigger type and trigger time. The proportioning scheme generation module first performs specification registration and storage location registration for steel balls of different diameters, then performs inventory status verification, allocable inventory verification, and call status registration for each specification of steel ball. Subsequently, it combines the coal powder fineness prediction results and steel ball wear rate prediction results from the trigger judgment package to perform proportioning constraint organization, forming the proportioning constraint package. The proportioning scheme generation module continues to generate mass ratios from the proportioning constraint package, obtaining a record of the proportions of steel balls of various specifications participating in this round of replenishment. It then organizes the steel ball specification correspondences within these records, establishing a correspondence between steel ball specifications, storage location, replenishment order, material distribution channel, and weighing location. Next, the proportioning scheme generation module performs a five-channel material distribution valve channel mapping on the steel ball specification correspondences, generating a proportioning scheme package containing mass ratio records, steel ball specification correspondences, and the five-channel material distribution valve channel mapping. The proportioning scheme generation module transmits this package to the transfer task generation module for use as a proportioning scheme package, while retaining the five-channel material distribution valve channel mapping for use by the arrival addition module.

[0160] The transfer task generation module 04 receives the proportioning scheme package, assembles task fields, registers target floors, and registers feeding positions to generate a transfer task package. Specifically, the transfer task generation module receives the proportioning scheme package from the proportioning scheme generation module and extracts the mass ratio record, steel ball specification correspondence, and five-channel material distribution valve channel mapping relationship from the proportioning scheme package. The transfer task generation module performs task field assembly on the above data, writing the task number, trigger time, steel ball specification sequence, bin source, ball retrieval sequence, target floor, feeding position, channel mapping relationship, and status record bit into the same task field set. Subsequently, the transfer task generation module performs target floor registration based on the coal mill platform position record and feeding position registration based on the coal mill feeding port position record, ensuring that the target floor record and feeding position record are bound to the task field set. After processing, a transfer task package is formed, which includes the task field set, target floor record, feeding position record, and pending execution marker. The transfer task generation module transmits the transfer task package to the navigation transfer module for use as a transfer task package, and at the same time writes the task number and target floor record into the storage area for subsequent cross-floor status verification and retrieval.

[0161] The navigation and transfer module 05 receives the transfer task package, performs warehouse ball-retrieval position analysis, path planning, and obstacle avoidance path generation on the transfer task package to obtain a navigation task package. It then sends arrival signals to the elevator entrance, target floor, and receives door opening signals from the navigation task package to generate a cross-floor transfer package. Furthermore, it connects the cross-floor transfer package to the elevator car, coal mill platform, and feeding position to generate an in-place addition package. Specifically, the navigation and transfer module receives the transfer task package from the transfer task generation module and extracts the warehouse source, target floor record, and feeding position record from the transfer task package. The navigation and transfer module first performs warehouse ball-retrieval position analysis to obtain the ball-retrieval position, stopping direction, and ball-retrieval order corresponding to the current task of the tracked laser navigation robot. Then, it combines map data to perform path planning and obstacle avoidance path generation to form a navigation task package. Afterward, the navigation and transfer module calls the cross-floor scheduling processing chain to send arrival signals to the elevator entrance, target floor, and receive door opening signals from the navigation task package to form a cross-floor transfer package. The navigation and transfer module continues to execute the process of entering the elevator car, arriving at the coal mill platform, and docking with the feeding position for the cross-floor transfer package. It writes the robot's current position, docking status, position correction record, and pending start status into the positioning add package. The navigation and transfer module transmits the positioning add package to the positioning add module for use as a positioning add package, and retains the elevator status record and docking status record from the cross-floor transfer package for subsequent status traceability.

[0162] The positioning addition module 06 receives the positioning addition package, performs channel allocation for the five-channel distribution valve, starts the stepper motor-driven screw feeder, and collects weighing sensor readings to obtain an addition execution package. It then performs mass ratio addition, steel ball stacking height detection, and pause addition alarm on the addition execution package to generate an addition result package. Specifically, the positioning addition module receives the positioning addition package from the navigation and transfer module and reads the channel mapping relationship, the waiting-to-start status, and the material feeding position docking status. The positioning addition module first performs channel allocation for the five-channel distribution valve, ensuring that the steel ball specifications participating in this round of ball replenishment correspond to the corresponding distribution channels. Then, it sends start control to the stepper motor-driven screw feeder corresponding to each distribution channel, while simultaneously collecting weighing sensor readings to form an addition execution package. Subsequently, the in-place addition module performs mass ratio addition according to the mass ratio record retained in the proportioning scheme package, and calls the detection component at the feeding port to perform steel ball accumulation height detection during the addition process. When an abnormal steel ball accumulation height is detected, the in-place addition module performs a pause addition alarm for the current material distribution channel and the related stepper motor driven screw feeder, and writes the pause status, alarm status, and completed addition amount into the addition result package. The in-place addition module transmits the addition result package to the feedback calibration module for use as the addition result package, and at the same time writes the actual addition amount and completion status of each specification of steel ball into the local recording area for comparison and retrieval of the before and after data acquisition of the coal mill operating parameters after ball replenishment.

[0163] The feedback calibration module 07 receives the added result package, performs post-ball replenishment mill operating parameter acquisition, comparison calibration, and proportioning scheme update on the added result package, generates an operating parameter package, and transmits the operating parameter package to the prediction trigger module. Specifically, the feedback calibration module receives the added result package from the in-place added module and initiates post-ball replenishment mill operating parameter acquisition processing based on the task number, completion time, and actual added quantity in the added result package. The feedback calibration module rereads the post-ball replenishment mill operating parameters from the programmable logic controller, performs field sorting, time sequence sorting, and task number binding on the post-ball replenishment mill operating parameters, and then compares and calibrates them with the preceding trigger judgment package, proportioning scheme package, and added result package to form a comparison record for this round of ball replenishment tasks. Based on the comparison record, the feedback calibration module performs a proportioning scheme update, writing the actual added quantity of each specification of steel ball, the post-ball replenishment mill operating parameters, and the comparison record into the proportioning scheme update data, and then generates a new operating parameter package. The feedback calibration module transmits the operating parameter package to the prediction trigger module as the input for the next round of 24-hour historical window sorting and coal powder fineness related parameter registration. At the same time, this module completes the associated storage of task number, completion time, comparison record and proportioning scheme update record, so that the coal mill operating parameter acquisition module, the prediction trigger module, the proportioning scheme generation module, the transfer task generation module, the navigation transfer module, the arrival addition module and the feedback calibration module form a closed loop link.

Claims

1. A method for dynamic proportioning and robotic transfer of steel balls in a coal mill based on AI, characterized in that, include: S100: Obtain the operating parameters of the coal mill, organize the 24-hour historical window and register the parameters related to coal powder fineness, generate a prediction input package, and perform coal powder fineness prediction, steel ball wear rate prediction and trigger judgment processing to generate a trigger judgment package. S200. Based on the trigger determination package, perform mass ratio generation, steel ball specification correspondence organization and five-channel material distribution valve channel mapping processing to generate a proportioning scheme package. Based on the aforementioned proportioning scheme package, task field assembly, target floor registration, and material feeding location registration are performed to generate a transfer task package; S300. Based on the transfer task package, perform A* algorithm path planning and DWA algorithm obstacle avoidance path generation, and perform docking processing to reach the coal mill platform and feeding position to generate the in-place addition package. S400. Based on the aforementioned in-place addition package, perform five-channel material distribution valve channel allocation processing, and execute the acquisition, comparison, calibration, and ratio scheme update processing of the coal mill operating parameters after ball replenishment to generate an operating parameter package.

2. The method according to claim 1, characterized in that, The process of organizing the 24-hour historical data and registering parameters related to coal powder fineness includes: The 24-hour historical window processing includes time segmentation, segment order reorganization, continuity check of the same field, and window boundary registration. The coal powder fineness related parameter registration processing includes mapping the coal powder fineness test record to the 24-hour historical window according to the sampling time, so that each segment of operating parameters carries coal powder fineness related parameters and generates a prediction input package.

3. The method according to claim 1, characterized in that, The process of performing coal powder fineness prediction, steel ball wear rate prediction, and trigger determination includes: The coal powder fineness prediction process includes reading the current, vibration, inlet and outlet pressure difference, temperature and coal powder fineness related parameters in the prediction input package and feeding them into the LSTM-GRU hybrid model segment by segment in time order to obtain the coal powder fineness prediction result. The steel ball wear rate prediction process includes reading current, vibration, inlet and outlet pressure difference and continuous running time records from the same prediction input packet and sending them into the LSTM-GRU hybrid model in the same time sequence to obtain the steel ball wear rate prediction result. Then, trigger judgment processing is performed on the coal powder fineness prediction result and the steel ball wear rate prediction result. The trigger determination process includes threshold comparison, anomaly source verification, and trigger status registration, generating a trigger determination package.

4. The method according to claim 1, characterized in that, The process of generating mass ratios, organizing steel ball specification correspondences, and mapping the channels of the five-channel material distribution valve includes: The mass ratio generation includes generating a replenishment ratio for steel balls of different diameters participating in this replenishment based on the steel ball specification set, the allocable inventory set, and the trigger type correspondence in the ratio constraint package. The steel ball specification correspondence organization includes establishing a fixed relationship between steel balls of different diameters and storage locations, replenishment order, and addition order; The five-channel material distribution valve channel mapping process includes mapping each steel ball specification participating in this round of ball replenishment to an independent channel of the five-channel material distribution valve, and simultaneously registering the reading positions of the stepper motor driven screw feeder and weighing sensor corresponding to the independent channel, thereby generating a proportioning scheme package.

5. The method according to claim 1, characterized in that, The process of assembling task fields, registering target floors, and registering material feeding locations includes: The task field assembly includes assembling the mass ratio records, steel ball specification correspondence, and five-channel material distribution valve channel mapping relationship in the proportioning scheme package into a set of task fields that can be executed by the tracked laser navigation robot. The task field set includes task number, trigger time, steel ball specification order, silo source, ball retrieval order, target floor, feeding location, channel mapping relationship, and status record bit; The target floor registration includes registering the elevator target floor that the tracked laser navigation robot will subsequently reach, and includes the current floor, the target floor, the coal mill platform number corresponding to the target floor, and the elevator call status; The material feeding location registration includes registering the final docking location of the tracked laser navigation robot after it completes cross-floor transfer, and includes the coal mill number, material feeding port location, material feeding port orientation, docking boundary and stopping direction, generating a transfer task package.

6. The method according to claim 1, characterized in that, The process of performing A* algorithm path planning and DWA algorithm obstacle avoidance path generation includes: The A* algorithm path planning includes generating a path sequence in the constructed map based on the warehouse ball retrieval position, target floor record, and material feeding position record, from the robot's current position to the warehouse ball retrieval position, from the warehouse ball retrieval position to the elevator entrance, and from the elevator exit of the target floor to the coal mill platform material feeding position; The obstacle avoidance path generation process of the DWA algorithm includes generating a local obstacle avoidance path based on the path sequence and the obstacle positions detected in real time by the lidar. The local obstacle avoidance path is concatenated with the path sequence.

7. The method according to claim 1, characterized in that, The process of docking the coal mill platform and the feeding position includes: The arrival at the coal mill platform includes, after the elevator has finished running and reached the target floor, exiting the elevator car and entering the coal mill platform passage according to the target floor record and platform direction record in the cross-floor transfer package; Arrive at the coal mill platform docking area according to the path sequence and local obstacle avoidance path in the navigation task package; The material feeding location docking process includes performing end-point approach, attitude adjustment, and stop locking processes based on the material feeding location record, docking direction, and docking boundary in the transfer task package, and generating a docking addition package.

8. The method according to claim 1, characterized in that, The process of allocating channels for the five-channel material distribution valve includes: The five-channel material distribution valve channel allocation process includes matching the specifications of the steel balls participating in the ball replenishment with the five independent discharge channels of the five-channel material distribution valve one by one according to the channel mapping relationship in the proportioning scheme package. Independent discharge channels that are not involved in this ball replenishment will be registered as closed.

9. The method according to claim 1, characterized in that, The process of acquiring, comparing, calibrating, and updating the proportioning scheme of the coal mill after ball replenishment includes: The acquisition of the mill operating parameters after ball replenishment includes reading the mill operating parameters after ball replenishment from the programmable logic controller again through the open platform communication unified architecture protocol after the completion of mass ratio addition, and binding them with the task number and completion time in the current addition result package; The comparison calibration includes comparing the operating parameters of the coal mill after ball replenishment with the trigger judgment package, proportioning scheme package and addition result package before ball replenishment, item by item. The comparison calibration includes comparing the actual amount of each steel ball specification added in the added result package with the mass ratio record in the proportioning scheme package; The operating parameters of the coal mill after ball replenishment are compared with the predicted results of coal powder fineness and steel ball wear rate in the trigger judgment package; The ratio scheme update process includes writing the current round of comparison and calibration results, the actual amount of each steel ball specification added, the coal mill operating parameters after ball replenishment, and the abnormal fluctuation correlation registration results back to the ratio scheme storage area, and updating the next round of task call boundaries according to the order of steel ball specifications to generate an operating parameter package.

10. An AI-based dynamic proportioning and robotic transfer system for steel balls in a coal mill, characterized in that, include: The pulverizer includes a coal mill operating parameter acquisition module, a prediction triggering module, a proportioning scheme generation module, a transfer task generation module, a navigation transfer module, a location addition module, and a feedback calibration module; these modules are connected in sequence to implement the method described in any one of claims 1-9.

Images