Reaction kettle cleaning and discharge control method and system based on multi-sensor fusion

Abstract

The application discloses a reaction kettle cleaning and discharge control method and system based on multi-sensor fusion, which adopts visual and ultrasonic sensors that can extend into the interior of the reaction kettle to perform multi-source data fusion detection, obtains the global distribution state of the kettle wall scraping and the large block residues at the kettle bottom, and adjusts the spray angle, pressure and walking track of the high-pressure water gun based on the same, so that the residual area is targetedly cleaned; during the discharge process, the discharge flow data is collected in real time, the stirring speed is dynamically adjusted through a stirring frequency adaptive algorithm, and the discharge flow is maintained stable; when the work requirement of feeding into the kettle is triggered, the power-off interlocking, gas purging and concentration detection are automatically performed according to the preset safety time sequence, the work permission is granted only when all indexes reach the standard, and the whole process interlocking control of feeding into the kettle is realized. The application effectively solves the problems of low cleaning automation, insufficient residue detection precision, non-adaptive stirring adjustment and imperfect safety interlocking system in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of chemical reactor automation technology, and particularly relates to a reactor cleaning and discharge control method and system based on multi-sensor fusion. Background Technology

[0002] Reactors are process equipment used in process industries such as chemical, pharmaceutical, fine chemical, and food processing for material mixing, reaction, and synthesis. During production batch changeovers or process maintenance, cleaning of material adhering to the reactor walls, discharging of residual material from the reactor bottom, and personnel entering the reactor for maintenance are indispensable procedures. Their operational efficiency and safety management level directly affect production continuity and the personal safety of on-site workers.

[0003] Currently, most chemical production sites still rely on manual high-pressure flushing as the primary method for cleaning reactors. Workers must enter the confined space inside the reactor and use handheld high-pressure water guns to clean the reactor walls and bottom area by area. This method is not only physically demanding and inefficient, but also exposes workers to multiple safety risks, including exposure to toxic and hazardous materials, accidental injury from high-pressure water, and suffocation in confined spaces. It falls under the category of high-risk operations in the chemical industry and is prone to causing personal injury accidents. Some large-scale enterprises use fixed high-pressure cleaning devices for semi-automatic reactor cleaning. However, these devices generally employ a pre-set fixed trajectory and constant spray pressure for full-area flushing, failing to dynamically adjust cleaning parameters based on the actual distribution and thickness of residual material inside the reactor. This results in both excessive cleaning of thin areas leading to water and energy waste, and incomplete cleaning of thick areas and corners requiring rework, making it difficult to balance cleaning quality and operational efficiency.

[0004] Existing automated reactor cleaning solutions mostly employ a single visual sensor or a single ultrasonic sensor for residue identification. Single visual detection is easily affected by environmental factors such as fog, dim lighting, and material reflection within the reactor, making it difficult to accurately obtain information on the thickness of residual material. While single ultrasonic detection can measure residue thickness, it cannot accurately delineate the planar distribution boundaries of the residue area. Neither of these detection methods can independently provide complete data on the overall distribution of residue within the reactor, failing to provide a reliable basis for targeted and precise cleaning decisions. Regarding material discharge control, existing reactor discharge processes often use a fixed-speed stirring device, unable to adaptively adjust the stirring frequency according to real-time changes in discharge flow. When the material in the reactor has high viscosity or contains large agglomerates, material bridging and blockage at the discharge port are prone to occur, leading to drastic fluctuations in discharge flow. Insufficient flow significantly prolongs discharge time and slows down production, while excessive flow can easily cause overload and overflow problems in subsequent material conveying pipelines, affecting the stability of the production process. In terms of safety management and control for operations in reactors, most existing enterprises rely on manual verification and confirmation for each step, such as power outage confirmation, reactor gas purging, hazardous gas detection, and work permit issuance. The standardization of process execution depends on the responsibility of personnel, which leads to safety hazards such as lack of interlock control, missed detections and misjudgments, and reversed processes. Safety accidents are easily caused by residual hazardous gases, accidental equipment startup, etc., making it difficult to meet the requirements for fully automated interlock control of confined space operations.

[0005] In summary, existing reactor cleaning and discharge control technologies suffer from technical defects such as low automation in cleaning operations, insufficient accuracy in residual detection, inability to adaptively adjust discharge stirring, and imperfect safety interlocking systems for reactor entry operations. These technologies cannot simultaneously meet the dual requirements of efficient production and inherent safety. Summary of the Invention

[0006] To address these issues, this invention provides a method and system for controlling reactor cleaning and discharge based on multi-sensor fusion, which solves the problems of low automation in existing reactor cleaning and discharge control technologies, insufficient accuracy in residual detection, inability to adaptively adjust discharge stirring, and imperfect safety interlocking system for reactor entry.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a reaction vessel cleaning and discharge control method based on multi-sensor fusion, comprising the following steps:

[0008] Reactor interior status perception: Multi-source data fusion detection is performed by visual sensors and ultrasonic sensors that can extend into the reactor interior to obtain the overall distribution status of material scraping residue on the reactor wall and large material residue on the reactor bottom, and the overall distribution status is used as the decision input for reactor cleaning operation;

[0009] Targeted cleaning control: Based on the global distribution state, the corresponding cleaning execution parameters are matched and the spray angle, spray pressure and travel trajectory of the high-pressure water gun are controlled to perform targeted cleaning operations on the residual areas on the inner surface of the reactor.

[0010] Adaptive stirring during material discharge: During the material discharge operation of the reactor, the material discharge flow rate data at the discharge port is collected in real time. Using the material discharge flow rate data as feedback, the operating frequency of the stirring mechanism is dynamically matched through an adaptive stirring frequency algorithm to maintain a stable flow rate during the material discharge process.

[0011] Safety interlock for entering the vessel: When the need for personnel to enter the vessel is triggered, the power-off interlock, gas purging inside the vessel, and harmful gas concentration detection are executed in sequence according to the preset safety sequence. Once all safety indicators meet the standards, the permission to enter the vessel is granted, so as to control the interlock for entering the vessel.

[0012] As a preferred embodiment of the reactor cleaning and discharge control method based on multi-sensor fusion, the visual sensor and the ultrasonic sensor are integrated at the end of a liftable and rotatable detection bracket, and the spatial coverage of the full-domain scanning satisfies the cylindrical coordinate model:

[0013]

[0014] In the formula, The scannable spatial region of the sensor; These are the radial, circumferential, and axial coordinates in the cylindrical coordinate system inside the vessel, respectively. The inner radius of the reactor; The height is the internal height of the reactor; the detection support achieves full-area scanning inside the reactor through a combination of circumferential rotation and axial lifting.

[0015] As a preferred method for reactor cleaning and discharge control based on multi-sensor fusion, the multi-source data fusion is performed through spatial coordinate registration and a weighted fusion model, specifically including:

[0016] Establish the projection mapping relationship between visual pixel coordinates and three-dimensional spatial coordinates within the vessel. The coordinate transformation formula is as follows:

[0017]

[0018] In the formula, The depth scale factor is determined jointly by the geometric constraints of the reactor body and ultrasonic ranging data. The projection matrix of the 3×4D vision sensor includes sensor intrinsic parameters, installation pose extrinsic parameters, and vessel coordinate system transformation parameters; u and v are the pixel coordinates of the residual points in the visual image; The three-dimensional spatial coordinates of the residual points inside the reactor in the reactor body coordinate system;

[0019] The residual thickness data detected by the ultrasonic sensor and the position data recognized by vision are spatially registered and then weighted and fused to obtain the fused estimate of the residual thickness:

[0020]

[0021] In the formula, Spatial coordinates The estimated value of the residual thickness of the material at the location; The residual thickness at this location as identified by the visual sensor; The residual thickness at this location, as detected by the ultrasonic sensor; These are the weighting coefficients for visual inspection and ultrasonic inspection, respectively, and they satisfy... .

[0022] As a preferred scheme for the reactor cleaning and discharge control method based on multi-sensor fusion, the cleaning execution parameters are matched according to the global distribution state. During the process of adjusting the spray angle, spray pressure and travel trajectory of the high-pressure water gun, the spray pressure and travel feed speed of the high-pressure water gun are adaptively matched according to the residual characteristics in the global distribution state, satisfying the following control model:

[0023] Injection pressure control formula:

[0024]

[0025] In the formula, Spatial coordinates The high-pressure water jet pressure at the location; Based on the cleaning pressure; This is the pressure gain coefficient; This represents the residual fusion thickness of the material at that location;

[0026] Formula for controlling the feed rate of the injection trajectory:

[0027]

[0028] In the formula, Spatial coordinates The travel speed of the high-pressure water gun; The reference feed rate; This is the velocity attenuation coefficient; This represents the residual coverage rate within the neighborhood of that location.

[0029] As a preferred embodiment of the reactor cleaning and discharge control method based on multi-sensor fusion, the end nozzle posture of the high-pressure water gun is controlled by a multi-degree-of-freedom robotic arm, and the spatial posture transformation of the nozzle satisfies a homogeneous transformation relationship:

[0030]

[0031] In the formula, This is the homogeneous transformation matrix of the nozzle relative to the robotic arm base, used to describe the spatial position and spraying attitude of the nozzle; For the first robotic arm The homogeneous transformation matrix corresponding to each joint; For the first The rotation angle of each joint; This represents the total number of degrees of freedom of the robotic arm.

[0032] As a preferred embodiment of the reactor cleaning and discharge control method based on multi-sensor fusion, the adaptive stirring frequency algorithm adopts an incremental closed-loop adjustment model, and the real-time adjustment increment of the stirring frequency satisfies the following formula:

[0033]

[0034] ;

[0035] In the formula, For the first Time relative to the first Adjust the stirring frequency increment at any given time; The first The deviation value of the material discharge flow rate at any given time; For flow deviation, Set the target feed rate; For the first The actual discharge flow rate detected at all times; This is the proportional adjustment coefficient; This is the integral adjustment coefficient; This is the differential adjustment coefficient.

[0036] As a preferred solution for the reactor cleaning and discharge control method based on multi-sensor fusion, the safety qualification determination of the reactor entry operation adopts a multi-index joint determination model, and the determination logic satisfies the following relation:

[0037]

[0038] In the formula, The result of the safety permit determination for entering the vessel. This indicates that the operation has been deemed successful and the work authorization has been granted. This indicates that the product is deemed unqualified and is prohibited from entering the reactor. This represents the actual concentration of harmful gases inside the vessel. The safe concentration threshold for harmful gases; This represents the actual oxygen concentration inside the reactor. These are the lower and upper limits of safe oxygen concentration, respectively. This is the power-off status flag. This indicates that the power circuit of the reactor has been disconnected.

[0039] As a preferred solution for the reactor cleaning and discharge control method based on multi-sensor fusion, it also includes performing an effect verification after the cleaning operation is completed, and judging it through cleanliness quantification indicators, the calculation formula of which is as follows:

[0040] Residual coverage inside the vessel:

[0041]

[0042] Maximum residual thickness inside the vessel:

[0043]

[0044] The criteria for judging cleanliness are as follows:

[0045] and

[0046] In the formula, This refers to the coverage rate of residual materials inside the vessel; The entire inner surface area of ​​the reactor; coordinates The residual region indicator function takes a value of 1 when there is a residual and a value of 0 when there is no residual. This represents the total surface area of ​​the reactor's inner surface. This represents the maximum residual thickness across the entire region. The threshold for acceptable residual coverage; This is the maximum acceptable threshold for residual thickness.

[0047] As a preferred solution for the reactor cleaning and discharge control method based on multi-sensor fusion, it also includes a continuous flow zero-value time-series detection model for determining the completion of the discharge operation, which satisfies the following judgment logic:

[0048]

[0049] In the formula, The material feeding is complete. This indicates that the material feeding operation is complete; for Instantaneous flow rate of material discharge at any given moment; This is the moment when the flow rate first drops to zero; The preset duration for continuous judgment.

[0050] This invention also provides a reactor cleaning and discharge control system based on multi-sensor fusion, employing the aforementioned reactor cleaning and discharge control method based on multi-sensor fusion, including:

[0051] The reactor interior state sensing unit is used to perform multi-source data fusion detection through visual sensors and ultrasonic sensors that can extend into the reactor interior, to obtain the overall distribution of scraped material residue on the reactor wall and large material residue on the reactor bottom, and to use the overall distribution as the decision input for the reactor cleaning operation.

[0052] The targeted cleaning control unit is used to match the corresponding cleaning execution parameters based on the global distribution state, and to control the spray angle, spray pressure and travel trajectory of the high-pressure water gun to perform targeted cleaning operations on the residual areas on the inner surface of the reactor.

[0053] The discharge adaptive stirring unit is used to collect the discharge flow rate data at the discharge port in real time during the discharge operation of the reactor. Using the discharge flow rate data as feedback, the unit dynamically matches the operating frequency of the stirring mechanism through an adaptive stirring frequency algorithm to maintain a stable flow rate during the discharge process.

[0054] The vessel entry safety interlock unit is used to execute power-off interlock, vessel gas purging, and harmful gas concentration detection in a preset safety sequence when the need for personnel to enter the vessel is triggered. Once all safety indicators meet the standards, entry permission is granted to control the vessel entry operation.

[0055] The present invention has the following advantages:

[0056] This invention employs a multi-sensor fusion detection method combining vision and ultrasound to simultaneously acquire information on the location distribution and thickness of residues inside the vessel. This overcomes the limitations of single-sensor detection, which is limited by its limited scope and susceptibility to environmental interference, providing a complete and reliable basis for cleaning operations. Based on the residue distribution, the invention adaptively adjusts the spray angle, pressure, and trajectory of the high-pressure water gun to achieve targeted cleaning. This ensures cleanliness while avoiding the waste of water and energy caused by constant-pressure rinsing throughout the entire process. Furthermore, the entire process eliminates the need for personnel to enter the vessel, completely replacing the high-risk manual cleaning procedure and effectively reducing the risk of injury or death in confined spaces.

[0057] This invention, through real-time feedback of discharge flow rate and linkage with an adaptive stirring frequency algorithm, can dynamically adjust the stirring speed according to the material discharge status in the reactor. It actively optimizes the stirring intensity for working conditions such as high-viscosity materials and large agglomerated materials, effectively alleviating the problems of material bridging and blockage at the discharge port, smoothing fluctuations in discharge flow rate, avoiding production anomalies such as overload and overflow in subsequent processes caused by abnormal flow rate, ensuring the continuous and stable operation of the discharge process and downstream conveying process, and optimizing the overall production cycle.

[0058] This invention integrates power outage interlocking, gas purging, concentration detection, and work permits into a unified interlocking control logic. It automatically executes each step of the process according to a preset safety sequence, avoiding problems such as process reversal, missed detections, misjudgments, and inadequate accountability that are prone to occur under manual control. It adopts a multi-indicator joint judgment mechanism to verify the conditions for entering the reactor. If any indicator fails to meet the standard, the entry permission is blocked, thus eliminating safety hazards such as residual harmful gases and accidental equipment start-up from a mechanism perspective, and meeting the strict safety control requirements for chemical confined space operations. Attached Figure Description

[0059] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0060] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0061] Figure 1 This is a schematic diagram of the reactor cleaning and discharge control method based on multi-sensor fusion provided in an embodiment of the present invention;

[0062] Figure 2 This is a technical roadmap for the reactor cleaning and discharge control method based on multi-sensor fusion provided in this embodiment of the invention;

[0063] Figure 3 This is a comparison chart of the cleaning performance of the control group and the present invention;

[0064] Figure 4 This is a comparison chart of the material feeding performance of the control group and the present invention;

[0065] Figure 5 This is a comparison chart of the safety interlocking performance of the control group and the present invention;

[0066] Figure 6 This is a diagram of the reactor cleaning and discharge control system based on multi-sensor fusion provided in an embodiment of the present invention. Detailed Implementation

[0067] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Example 1

[0069] See Figure 1 and Figure 2 This invention provides a method for controlling the cleaning and discharge of a reaction vessel based on multi-sensor fusion, comprising the following steps:

[0070] Reactor interior status perception: Multi-source data fusion detection is performed by visual sensors and ultrasonic sensors that can extend into the reactor interior to obtain the overall distribution status of material scraping residue on the reactor wall and large material residue on the reactor bottom, and the overall distribution status is used as the decision input for reactor cleaning operation;

[0071] Targeted cleaning control: Based on the global distribution state, the corresponding cleaning execution parameters are matched and the spray angle, spray pressure and travel trajectory of the high-pressure water gun are controlled to perform targeted cleaning operations on the residual areas on the inner surface of the reactor.

[0072] Adaptive stirring during material discharge: During the material discharge operation of the reactor, the material discharge flow rate data at the discharge port is collected in real time. Using the material discharge flow rate data as feedback, the operating frequency of the stirring mechanism is dynamically matched through an adaptive stirring frequency algorithm to maintain a stable flow rate during the material discharge process.

[0073] Safety interlock for entering the vessel: When the need for personnel to enter the vessel is triggered, the power-off interlock, gas purging inside the vessel, and harmful gas concentration detection are executed in sequence according to the preset safety sequence. Once all safety indicators meet the standards, the permission to enter the vessel is granted, so as to control the interlock for entering the vessel.

[0074] Among these features, the internal vessel status perception system utilizes a complementary fusion of visual and ultrasonic sensors to overcome the limitations of single-dimensional detection, providing a status-based basis for cleaning control. Targeted vessel cleaning regulation enables on-demand cleaning based on perception results, differing from the traditional extensive approach of constant pressure rinsing across the entire area. This system simultaneously reduces energy and water consumption while ensuring cleaning effectiveness. Adaptive stirring during material discharge serves as a process control element, using a feedback mechanism to smooth out fluctuations in discharge flow, addressing the industry pain point of unstable discharge of high-viscosity, easily agglomerated materials. The vessel entry safety interlock system, through mandatory time-sequenced interlock logic, technically prevents oversights in manual operations, adhering to the inherent safety design principles of confined space operations.

[0075] In one possible embodiment, the visual sensor and the ultrasonic sensor are integrated at the end of a height-adjustable and rotatable detection bracket, and the spatial coverage of the full-area scanning satisfies the cylindrical coordinate model:

[0076]

[0077] In the formula, The scannable spatial region of the sensor; These are the radial, circumferential, and axial coordinates in the cylindrical coordinate system inside the vessel, respectively. The inner radius of the reactor; The height is the internal height of the reactor; the detection support achieves full-area scanning inside the reactor through a combination of circumferential rotation and axial lifting.

[0078] Specifically, a cylindrical coordinate scanning model is adopted, which highly matches the geometric features of the cylindrical cavity of the reactor. Compared with the conventional rectangular coordinate scanning method, this significantly simplifies the motion control logic while improving scanning efficiency and coverage integrity. After the detection support extends into the reactor from the manhole at the top, its axial lifting motion covers the entire height range of the reactor body, and its circumferential rotational motion covers the entire circumferential area of ​​the reactor wall. The combined motion of these two movements forms a spiral scanning path over the cylindrical space inside the reactor. This eliminates the need for an additional radial motion mechanism to achieve comprehensive coverage of the reactor wall and bottom, simplifying the mechanical structure of the insertion mechanism and reducing equipment complexity and operational failure rate. The radial range of the cylindrical coordinate model is naturally constrained by the inner radius of the reactor, ensuring that the sensor detection range always corresponds to the boundary of the inner wall of the reactor, avoiding invalid scans beyond the reactor body's range, and further improving the spatial accuracy of the detection data.

[0079] In one possible embodiment, the multi-source data fusion is performed through spatial coordinate registration and a weighted fusion model, specifically including:

[0080] Establish the projection mapping relationship between visual pixel coordinates and three-dimensional spatial coordinates within the vessel. The coordinate transformation formula is as follows:

[0081]

[0082] In the formula, The depth scale factor is determined jointly by the geometric constraints of the reactor body and ultrasonic ranging data. The projection matrix of the 3×4D vision sensor includes sensor intrinsic parameters, installation pose extrinsic parameters, and vessel coordinate system transformation parameters; u and v are the pixel coordinates of the residual points in the visual image; The three-dimensional spatial coordinates of the residual points inside the reactor in the reactor body coordinate system;

[0083] The residual thickness data detected by the ultrasonic sensor and the position data recognized by vision are spatially registered and then weighted and fused to obtain the fused estimate of the residual thickness:

[0084]

[0085] In the formula, Spatial coordinates The estimated value of the residual thickness of the material at the location; The residual thickness at this location as identified by the visual sensor; The residual thickness at this location, as detected by the ultrasonic sensor; These are the weighting coefficients for visual inspection and ultrasonic inspection, respectively, and they satisfy... .

[0086] Specifically, the multi-source data fusion in this embodiment is divided into two stages: spatial coordinate registration and data weighted fusion. The ultimate goal is to achieve spatial alignment and complementary advantages between the two types of sensor data. The coordinate transformation formula is based on the imaging principle of a pinhole camera. After being projected by an optical lens, an object point in three-dimensional space will form a two-dimensional pixel on the image sensor plane. This projection process can be mathematically described by a 3×4 projection matrix. Since monocular vision cannot deduce the unique three-dimensional depth information solely from the two-dimensional pixel coordinates, a depth scale factor is introduced. The projection matrix is ​​determined by combining the known geometric constraints of the reactor's inner wall with the direct ranging results of the ultrasonic sensor, ensuring the uniqueness of the pixel coordinate to spatial coordinate mapping and fully reusing the detection data from both types of sensors. The included intrinsic parameters correspond to the camera's own optical parameters such as focal length and principal point, while the extrinsic parameters correspond to the sensor's installation position and orientation on the detection bracket. These parameters can be determined through pre-calibration, thereby achieving the unification of the image coordinate system and the physical coordinate system of the vessel.

[0087] After spatial registration, the data is fused using a weighted fusion model. This model is designed based on the performance characteristics of the two types of sensors: visual sensors offer high accuracy in recognizing the planar contours of residual areas, but thickness measurement is easily affected by material color and surface reflection, resulting in relatively large errors; ultrasonic sensors offer high accuracy in thickness measurement, but have low planar position resolution and blurred boundary recognition. By mapping the two types of data to the same coordinate system and assigning different weights to the two detection results, the reliability of thickness detection can be improved by leveraging ultrasonic data while retaining the positional accuracy advantage of visual detection. The weighting coefficients can be flexibly adjusted according to material characteristics and on-site conditions. The resulting fused thickness data combines both positional and thickness accuracy, providing a reliable decision-making basis for targeted cleaning.

[0088] In one possible embodiment, based on the cleaning execution parameters corresponding to the global distribution state, during the adjustment of the high-pressure water gun's spray angle, spray pressure, and travel trajectory, the high-pressure water gun's spray pressure and travel feed speed are adaptively matched according to the residual characteristics in the global distribution state, satisfying the following control model:

[0089] Injection pressure control formula:

[0090]

[0091] In the formula, Spatial coordinates The high-pressure water jet pressure at the location; Based on the cleaning pressure; This is the pressure gain coefficient; This represents the residual fusion thickness of the material at that location;

[0092] Formula for controlling the feed rate of the injection trajectory:

[0093]

[0094] In the formula, Spatial coordinates The travel speed of the high-pressure water gun; The reference feed rate; This is the velocity attenuation coefficient; This represents the residual coverage rate within the neighborhood of that location.

[0095] Specifically, the adaptive cleaning control model is designed based on the working mechanism of water jet cleaning. It dynamically matches the cleaning energy input according to the local characteristics of residual materials to achieve targeted control effects such as strong cleaning for thick materials, weak cleaning for thin materials, and rapid sweeping for materials without residue. Among them, the jet pressure is positively correlated with the residue thickness, and the impact peeling ability of high-pressure water jet is directly related to the jet pressure. The higher the pressure, the greater the impact kinetic energy per unit area, and the stronger the peeling effect on thick and hard residues. The formula sets a basic cleaning pressure to ensure that even areas without obvious residue can maintain a basic rinsing and cleaning effect. The pressure gain coefficient is used to match the adhesion strength of different materials. The stronger the adhesion of the material, the larger the value is, so that the jet pressure increases linearly with the residue thickness, ensuring that thick material areas receive sufficient peeling energy and avoiding incomplete cleaning.

[0096] The feed rate control logic is related to the cleaning duration. The cleaning effect is not only related to pressure but also positively correlated with the time the water jet spends per unit area. A slower feed rate results in a longer cleaning time per unit area, leading to more thorough cleaning. The formula shows that the feed rate decreases as the residue coverage increases, allowing for longer rinsing times in areas with dense residue, thus enhancing the cleaning effect. Conversely, a faster feed rate is used in areas with sparse or no residue, reducing ineffective rinsing time. The coordinated control of both jet pressure and feed rate parameters minimizes water and energy consumption while maintaining consistent cleanliness across the entire area, improving the economy and efficiency of the cleaning operation.

[0097] In one possible embodiment, the position of the end nozzle of the high-pressure water gun is controlled by a multi-degree-of-freedom robotic arm, and the spatial position transformation of the nozzle satisfies a homogeneous transformation relationship:

[0098]

[0099] In the formula, This is the homogeneous transformation matrix of the nozzle relative to the robotic arm base, used to describe the spatial position and spraying attitude of the nozzle; For the first robotic arm The homogeneous transformation matrix corresponding to each joint; For the first The rotation angle of each joint; This represents the total number of degrees of freedom of the robotic arm.

[0100] Specifically, this embodiment is based on the DH parameter method in robot kinematics. A homogeneous transformation matrix is ​​used to describe the end effector pose of a multi-joint robotic arm, thereby controlling the nozzle's spray angle and position. The advantage of the homogeneous transformation matrix is ​​that it can simultaneously describe the three-dimensional translation and three-dimensional rotation of a rigid body, decomposing complex multi-joint motion into successive transformations between adjacent joints, significantly simplifying the kinematic solution process. Each rotating joint of the robotic arm corresponds to a homogeneous transformation matrix, used to describe the position and attitude changes between adjacent links when that joint rotates. By sequentially multiplying the transformation matrices of all joints from the base to the end nozzle, the total pose of the end nozzle relative to the base can be obtained. Using this mathematical model, the control system can reverse-engineer the required rotation angle of each joint based on the spatial coordinates of the target cleaning point, achieving coordinated control of the nozzle position and spray angle. This ensures that the water jet always impacts the residual surface inside the vessel perpendicularly or at the optimal angle, maximizing the effective utilization of cleaning energy, while flexibly adapting to the cleaning posture requirements of different locations such as the vessel wall, bottom, and corners.

[0101] In one possible embodiment, the adaptive stirring frequency algorithm employs an incremental closed-loop adjustment model, and the real-time adjustment increment of the stirring frequency satisfies the following formula:

[0102]

[0103]

[0104] In the formula, For the first Time relative to the first Adjust the stirring frequency increment at any given time; The first The deviation value of the material discharge flow rate at any given time; For flow deviation, Set the target feed rate; For the first The actual discharge flow rate detected at all times; This is the proportional adjustment coefficient; This is the integral adjustment coefficient; This is the differential adjustment coefficient.

[0105] Specifically, this embodiment employs an incremental PID control algorithm. Through a feedback mechanism, the stirring frequency is dynamically adjusted based on the deviation between the discharge flow rate and the target value, ultimately stabilizing the actual flow rate within the target range. The flow rate deviation reflects the difference between the current discharge flow rate and the set value: when the actual flow rate is lower than the target value, the deviation is positive, and the control system increases the stirring frequency to enhance material agitation and flowability, promoting material descent; when the actual flow rate is higher than the target value, the deviation is negative, and the control system decreases the stirring frequency to prevent excessively rapid material descent and flow overload.

[0106] The incremental algorithm employed outputs the adjustment of the stirring frequency between adjacent time points, rather than the absolute frequency value. This approach offers several advantages: firstly, it requires less computation, needing only the deviation data between the current and two previous time points, making it suitable for real-time operation of embedded controllers; secondly, it exhibits strong resistance to integral saturation, preventing control input from exceeding limits due to long-term deviation accumulation, thus adapting to the output range constraints of variable frequency speed control for stirring motors; and thirdly, the control process is smooth, with no impact during start-up, shutdown, and switching, protecting the stirring equipment and transmission mechanism. The proportional term in the formula rapidly responds to the current deviation, the integral term eliminates steady-state errors to ensure the flow rate ultimately matches the target value, and the derivative term predicts the deviation change trend to suppress flow fluctuations. The combination of these three factors enables rapid and stable control of the discharge flow rate.

[0107] In one possible embodiment, the safety qualification determination of the reactor entry operation adopts a multi-index joint determination model, and the determination logic satisfies the following relation:

[0108]

[0109] In the formula, The result of the safety permit determination for entering the vessel. This indicates that the operation has been deemed successful and the work authorization has been granted. This indicates that the product is deemed unqualified and is prohibited from entering the reactor. This represents the actual concentration of harmful gases inside the vessel. The safe concentration threshold for harmful gases; This represents the actual oxygen concentration inside the reactor. These are the lower and upper limits of safe oxygen concentration, respectively. This is the power-off status flag. This indicates that the power circuit of the reactor has been disconnected.

[0110] Specifically, this multi-indicator joint judgment model follows the failure-safe and full-condition logic principles in safety system design. Work access is granted only when all safety conditions are simultaneously met; failure to meet any condition forcibly blocks the entry channel, thus preventing safety oversights. Specifically, the power-off status flag corresponds to the risk of mechanical injury, ensuring that power equipment such as reactor agitators and feeders cannot be accidentally started during entry, preventing injuries from mechanical operation; the concentration of harmful gases corresponds to the risk of poisoning and explosion, ensuring that the concentration of toxic, flammable, and explosive gases inside the reactor is below safe limits, preventing poisoning and explosion accidents; and the oxygen concentration corresponds to the risk of asphyxiation and oxygen overload. Too low an oxygen concentration can cause asphyxiation, while too high a concentration can exacerbate the risk of explosion, therefore it must be controlled within a reasonable range. Using an AND logic to connect the three conditions fully complies with the safety regulations for confined space operations. Replacing the traditional manual item-by-item verification control mode, it completely avoids omissions and misjudgments caused by human negligence or reversed processes, achieving rigid constraints on safety conditions and significantly improving the inherent safety level of reactor entry operations.

[0111] In one possible embodiment, the process further includes a performance review after the cleaning operation is completed, which is determined using quantitative indicators of cleanliness, calculated as follows:

[0112] Residual coverage inside the vessel:

[0113]

[0114] Maximum residual thickness inside the vessel:

[0115]

[0116] The criteria for judging cleanliness are as follows:

[0117] and

[0118] In the formula, This refers to the coverage rate of residual materials inside the vessel; The entire inner surface area of ​​the reactor; coordinates The residual region indicator function takes a value of 1 when there is a residual and a value of 0 when there is no residual. This represents the total surface area of ​​the reactor's inner surface. This represents the maximum residual thickness across the entire region. The threshold for acceptable residual coverage; This is the maximum acceptable threshold for residual thickness.

[0119] Specifically, this embodiment uses two indicators—coverage rate and maximum thickness—to quantitatively determine the cleaning effect. This allows for a comprehensive evaluation of cleaning quality from two dimensions: the overall residual ratio and localized extreme residues, avoiding the one-sidedness of judging by a single indicator. The residual coverage rate is calculated by integrating the entire domain of an indicator function to determine the ratio of the total residual area to the total surface area inside the vessel, reflecting the overall proportion of residues within the vessel and providing a direct representation of the overall cleaning effect. The binarization of the indicator function simplifies the calculation process and facilitates rapid statistical analysis using sensor data. The maximum residual thickness is calculated by extracting the value of the thickest residue within the entire domain. This is used to control stubborn localized residues, preventing situations where the overall coverage rate meets the standard but some areas still have thick residues that have not been thoroughly cleaned, thus preventing any impact on the quality of the next batch of products.

[0120] The judgment logic of simultaneous achievement of dual indicators ensures that the overall cleanliness of the vessel meets the process requirements, strictly controls the risk of local residues, and can flexibly adjust the threshold according to the cleanliness requirements of different production processes. It is compatible with the cleaning acceptance standards of different industries such as pharmaceuticals, fine chemicals, and food, and realizes objective and quantitative evaluation of the cleaning effect, replacing the subjective mode of traditional manual visual inspection.

[0121] In one possible embodiment, the determination of the completion of the material feeding operation also includes using a continuous flow zero-value time-series detection model, satisfying the determination logic:

[0122]

[0123] In the formula, The material feeding is complete. This indicates that the material feeding operation is complete; for Instantaneous flow rate of material discharge at any given moment; This is the moment when the flow rate first drops to zero; The preset duration for continuous judgment.

[0124] Specifically, this continuous zero-flow timing judgment logic is used to filter out instantaneous flow fluctuations in actual operating conditions, improving the accuracy of discharging completion judgment and avoiding misjudgments. During actual discharging, material agglomeration, air bubbles passing through the flow meter, and sensor signal jitter can all cause the instantaneous flow to briefly drop to zero. If discharging completion is judged solely based on an instantaneous flow of zero, misjudgments are highly likely, leading to premature initiation of subsequent cleaning processes, causing material leakage and ineffective cleaning. By setting a continuous judgment duration, the flow rate must remain zero for a continuous period to confirm that the material in the vessel has been completely emptied, effectively filtering various instantaneous interference signals. The duration can be flexibly set according to material characteristics and discharging pipe diameter. For materials with high viscosity, poor flowability, and long dripping duration, the judgment duration can be appropriately extended to ensure reliable judgment results.

[0125] To verify the practical application effect of the present invention, comparative tests were conducted on the efficiency and cleanliness of the cleaning operation, the stability of the material discharge process, and the reliability of the safety interlock for entering the vessel, so as to quantify the technical advantages and degree of improvement of the present invention compared with the traditional operation mode.

[0126] I. Test Platform

[0127] A standard stainless steel reactor with a nominal volume of 500L was used as the test vessel. The reactor has an inner diameter of 800mm and a height of 1000mm, and is equipped with an anchor-type stirring mechanism. Three parallel platforms were set up simultaneously for the test to ensure that the reactor specifications and initial operating conditions were completely consistent.

[0128] The experimental platform integrates a liftable and rotatable detection bracket (equipped with an industrial vision sensor and an ultrasonic ranging sensor), a six-degree-of-freedom robotic arm high-pressure water gun mechanism, a variable frequency stirring drive module, a reactor inlet safety interlock unit (power failure interlock module, inert gas purging module, multi-component gas detection module), and a central control cabinet, and runs the full-process control method described in this invention.

[0129] Control Group 1 Platform: The same specifications of reaction vessel, equipped with a handheld high-pressure water gun and a fixed speed stirring motor, adopts the traditional operation mode of manual entry into the vessel for cleaning and manual verification of safety conditions item by item.

[0130] Control group 2 platform: The same specifications of reaction vessel, equipped with a fixed trajectory automatic high pressure cleaning device (only equipped with a single vision sensor), a fixed speed stirring motor, and a semi-automatic vessel cleaning mode with a preset trajectory. The safety verification process is still completed manually.

[0131] II. Test Materials

[0132] High-viscosity acrylic resin emulsion, which is prone to adhering to reactor walls and agglomerating in chemical production, was selected as the test material. With a solid content of 45% and a dynamic viscosity of 1200 mPa·s at 25°C, it can realistically simulate the typical working conditions of material adhering to the reactor walls and large residues at the bottom during actual production. Before each test, equal amounts of material were added to each group of reactors, stirred evenly, discharged, and allowed to stand for 12 hours to form a uniformly thick scraped layer on the reactor walls and accumulated residues at the bottom. Sampling tests before the test confirmed that the initial residue state of each group was consistent, with an initial residue thickness of 2–5 mm on the reactor walls and a large residue thickness of 30–50 mm at the bottom.

[0133] III. Experimental Grouping and Evaluation Indicators

[0134] The cleaning performance and discharge performance tests were repeated 5 times per group, and the arithmetic mean was taken as the final test data; the safety interlock test was repeated 20 times per group, including 5 preset abnormal conditions such as power failure and 5 conditions of excessive harmful gases. The evaluation indicators were divided into three categories:

[0135] Performance indicators for cleaning vessel: total cleaning time, water consumption, cleanliness pass rate, and maximum residual thickness inside the vessel.

[0136] Feeding performance indicators: total feeding time, feeding flow rate fluctuation coefficient, number of times of material blockage during feeding process

[0137] Safety interlock indicators: total time spent on the preparation process before entering the reactor, false alarm rate for safety conditions, and number of missed interlock anomalies.

[0138] IV. Test Items and Methods

[0139] (41) Comparison test of cleaning performance

[0140] Each group completed the full-capsule cleaning according to the corresponding operation mode. After cleaning, the same calibrated multi-sensor detection system was used to perform full-area scanning detection, record the total cleaning time and total water consumption, and calculate the maximum residual thickness inside the capsule.

[0141] The cleanliness qualification standard is uniformly set as follows: residual coverage rate ≤1% and maximum residual thickness ≤0.2mm. Both indicators must be met simultaneously to be considered qualified. The cleanliness qualification rate of each group is calculated accordingly.

[0142] (42) Comparative test of material feeding stability

[0143] In each experiment, 300 kg of test material was added to the reactor, stirred evenly at room temperature, and then discharged by gravity through the discharge valve. Control groups 1 and 2 used a fixed stirring speed of 60 rpm, while the experimental group used the adaptive stirring frequency algorithm of this invention, with the target discharge flow rate set at 15 L / min.

[0144] The flow rate data is collected in real time at a frequency of 1Hz using an electromagnetic flow meter at the discharge port. The total discharge time is recorded, the flow fluctuation coefficient (the ratio of the standard deviation of the flow rate to the average flow rate) is calculated, and the number of flow interruptions and blockages caused by material bridging during the discharge process is counted.

[0145] (43) Comparative test of the reliability of safety interlocks

[0146] Simulating a scenario where personnel enter the reactor for maintenance, each group completes the entire process of power outage, purging, gas detection, and work permit issuance before entering the reactor according to the corresponding procedures. Each group conducts a total of 20 tests, including 5 random insertions of abnormal conditions such as power outage failure and 5 insertions of abnormal conditions with excessive harmful gases. The average time taken for the entire process is recorded, and the number of misjudgments of safety conditions and the number of missed interlocking abnormalities are statistically analyzed (entry permits are granted even if the power outage / purging steps are not completed).

[0147] V. Experimental Results and Analysis

[0148] (51) Results of the cleaning performance test

[0149] The comparison of the cleaning performance data of the three groups of experiments is shown in Table 1:

[0150] Table 1 Comparison of cleaning performance data of the three groups of experiments

[0151]

[0152] See Figure 3 As shown in Table 1, the cleaning time of the experimental group of this invention was only 39% of that of the manual mode, and 42.5% shorter than that of the semi-automatic mode; the water consumption for cleaning was reduced by 53.8% compared to the manual mode and by 39.6% compared to the semi-automatic mode; the cleanliness pass rate reached 100%, and the maximum residual thickness was significantly lower than that of the two control groups. This is because the present invention uses a fusion of visual and ultrasonic multi-sensor technology to identify the location distribution and thickness differences of the residue, and targets and matches the spray pressure and feed speed accordingly. Thick material areas are cleaned with high pressure and slow speed, while thin material areas are cleaned with low pressure and fast speed. This avoids the blind spots and randomness of manual cleaning, and solves the energy and water waste caused by the constant pressure rinsing of the entire area in the semi-automatic mode, while simultaneously balancing cleaning efficiency and cleaning quality. The manual mode is limited by operator skill and field of vision, and thick material is easily left in corner areas, resulting in the lowest pass rate; the semi-automatic mode has a fixed trajectory and insufficient cleaning energy for thick material areas, resulting in the next lowest pass rate; the present invention can ensure full compliance through effect verification and supplementary cleaning.

[0153] (52) Results of material discharge performance test

[0154] Table 2 shows a comparison of the material discharge performance data from the three groups of tests.

[0155] Table 2 Comparison of material discharge performance data in the three groups of tests

[0156]

[0157] See Figure 3 As shown in Table 2, the experimental results indicate that the discharge flow rate fluctuation coefficient of the experimental group of this invention is only 0.09, far lower than the 0.35 or higher of the control group. There was no material blockage during the entire discharge process, and the total discharge time was shortened by approximately 22% compared to the control group. This effect is attributed to the adaptive stirring frequency algorithm: when the flow rate decreases, the system automatically increases the stirring frequency to enhance material disturbance, disrupt the material bridging structure at the bottom of the vessel, and improve material flowability; when the flow rate is too high, the system automatically reduces the stirring speed to prevent concentrated material spillage. This dynamic adjustment achieves stable control of the discharge flow rate, effectively solving the common industry problem of easy blockage and large fluctuations in the discharge of high-viscosity materials.

[0158] (53) Results of safety interlock test

[0159] Table 3 compares the safety interlocking performance data of the three sets of tests.

[0160] Table 3 Comparison of safety interlock performance data from the three sets of tests

[0161]

[0162] See Figure 5 As shown in Table 3, the test results indicate that the fully automated interlocking process of this invention takes only 47% of the time of the manual mode, and the false judgment rate and the number of missed interlocks were both 0 in 20 tests. Under preset abnormal conditions such as power outage failure and excessive gas levels, it can accurately identify and block the entry permission. Traditional manual verification methods heavily rely on personnel responsibility and operational standards, and are prone to problems such as process omissions and misinterpretation of test data; semi-automatic modes still retain manual intervention, posing a certain risk of oversight. This invention, through fully automated time-series interlocking and multi-indicator joint judgment, eliminates the space for human error from a mechanism perspective, significantly improving the inherent safety level of entry operations, while also greatly shortening the safety preparation cycle.

[0163] VI. Experimental Conclusions

[0164] The aforementioned comparative experiments fully verified the technical effects of this invention. Compared with traditional operating modes, this invention can significantly improve the efficiency of reactor cleaning operations, reduce cleaning energy and water consumption, and ensure higher cleaning quality; it can significantly smooth out fluctuations in discharge flow, eliminate the risk of material blockage, and improve the stability of the discharge process; it can realize fully automated interlocked control of the entire reactor entry process, completely avoid human safety oversights, and shorten the safety preparation cycle. This invention comprehensively solves the problems of high risk and low efficiency, extensive control, and insufficient safety assurance in existing reactor cleaning and discharge operations, and meets the technical needs of intelligent manufacturing upgrades and safety production rectification in the chemical industry.

[0165] Example 2

[0166] See Figure 6 Embodiment 2 of the present invention also provides a reactor cleaning and discharge control system based on multi-sensor fusion, which adopts the reactor cleaning and discharge control method based on multi-sensor fusion of the above embodiments, including:

[0167] The reactor interior state sensing unit 100 is used to perform multi-source data fusion detection through a visual sensor and an ultrasonic sensor that can extend into the reactor interior, to obtain the overall distribution status of scraped material residue on the reactor wall and large material residue on the reactor bottom, and to use the overall distribution status as the decision input for the reactor cleaning operation.

[0168] The targeted cleaning control unit 200 is used to match the corresponding cleaning execution parameters based on the global distribution state, and to control the spray angle, spray pressure and travel trajectory of the high-pressure water gun to perform targeted cleaning operations on the residual areas on the inner surface of the reactor.

[0169] The discharge adaptive stirring unit 300 is used to collect the discharge flow rate data of the discharge port in real time during the discharge operation of the reactor. Using the discharge flow rate data as feedback, the operating frequency of the stirring mechanism is dynamically matched through the stirring frequency adaptive algorithm to maintain the flow rate stability during the discharge process.

[0170] The safety interlock unit 400 for entering the vessel is used to execute the power-off interlock, gas purging inside the vessel, and harmful gas concentration detection in a preset safety sequence when the need for personnel to enter the vessel is triggered. Once all safety indicators are met, the permission to enter the vessel is granted, so as to control the interlock of the entry operation.

[0171] As a preferred solution for a reactor cleaning and discharge control system based on multi-sensor fusion, in the reactor state sensing unit 100, the visual sensor and the ultrasonic sensor are integrated at the end of a liftable and rotatable detection bracket, and the spatial coverage of the full-area scanning satisfies the cylindrical coordinate model:

[0172]

[0173] In the formula, The scannable spatial region of the sensor; These are the radial, circumferential, and axial coordinates in the cylindrical coordinate system inside the vessel, respectively. The inner radius of the reactor; The height is the internal height of the reactor; the detection support achieves full-area scanning inside the reactor through a combination of circumferential rotation and axial lifting.

[0174] As a preferred solution for a reactor cleaning and discharge control system based on multi-sensor fusion, the reactor state sensing unit 100 performs multi-source data fusion through spatial coordinate registration and a weighted fusion model, specifically including:

[0175] Establish the projection mapping relationship between visual pixel coordinates and three-dimensional spatial coordinates within the vessel. The coordinate transformation formula is as follows:

[0176]

[0177] In the formula, The depth scale factor is determined jointly by the geometric constraints of the reactor body and ultrasonic ranging data. The projection matrix of the 3×4D vision sensor includes sensor intrinsic parameters, installation pose extrinsic parameters, and vessel coordinate system transformation parameters; u and v are the pixel coordinates of the residual points in the visual image; The three-dimensional spatial coordinates of the residual points inside the reactor in the reactor body coordinate system;

[0178] The residual thickness data detected by the ultrasonic sensor and the position data recognized by vision are spatially registered and then weighted and fused to obtain the fused estimate of the residual thickness:

[0179]

[0180] In the formula, Spatial coordinates The estimated value of the residual thickness of the material at the location; The residual thickness at this location as identified by the visual sensor; The residual thickness at this location, as detected by the ultrasonic sensor; These are the weighting coefficients for visual inspection and ultrasonic inspection, respectively, and they satisfy... .

[0181] As a preferred solution for a reactor cleaning and discharge control system based on multi-sensor fusion, in the targeted cleaning control unit 200, the high-pressure water gun's spray angle, spray pressure, and travel trajectory are adjusted based on the cleaning execution parameters matched to the corresponding global distribution state. During this process, the high-pressure water gun's spray pressure and travel feed speed are adaptively matched according to the residual characteristics in the global distribution state, satisfying the following control model:

[0182] Injection pressure control formula:

[0183]

[0184] In the formula, Spatial coordinates The high-pressure water jet pressure at the location; Based on the cleaning pressure; This is the pressure gain coefficient; This represents the residual fusion thickness of the material at that location;

[0185] Formula for controlling the feed rate of the injection trajectory:

[0186]

[0187] In the formula, Spatial coordinates The travel speed of the high-pressure water gun; The reference feed rate; This is the velocity attenuation coefficient; This represents the residual coverage rate within the neighborhood of that location.

[0188] As a preferred solution for a reactor cleaning and discharge control system based on multi-sensor fusion, in the targeted cleaning control unit 200, the end nozzle pose of the high-pressure water gun is controlled by a multi-degree-of-freedom robotic arm, and the spatial pose transformation of the nozzle satisfies a homogeneous transformation relationship:

[0189]

[0190] In the formula, This is the homogeneous transformation matrix of the nozzle relative to the robotic arm base, used to describe the spatial position and spraying attitude of the nozzle; For the first robotic arm The homogeneous transformation matrix corresponding to each joint; For the first The rotation angle of each joint; This represents the total number of degrees of freedom of the robotic arm.

[0191] As a preferred solution for the reactor cleaning and discharge control system based on multi-sensor fusion, the discharge adaptive stirring unit 300 employs an incremental closed-loop adjustment model for its adaptive stirring frequency algorithm, and the real-time adjustment increment of the stirring frequency satisfies the following formula:

[0192]

[0193] ;

[0194] In the formula, For the first Time relative to the first Adjust the stirring frequency increment at any given time; The first The deviation value of the material discharge flow rate at any given time; For flow deviation, Set the target feed rate; For the first The actual discharge flow rate detected at all times; This is the proportional adjustment coefficient; This is the integral adjustment coefficient; This is the differential adjustment coefficient.

[0195] As a preferred solution for a reactor cleaning and discharge control system based on multi-sensor fusion, the reactor entry safety interlock unit 400 uses a multi-index joint judgment model to determine the safety qualification of the reactor entry operation, and the judgment logic satisfies the following relation:

[0196]

[0197] In the formula, The result of the safety permit determination for entering the vessel. This indicates that the operation has been deemed successful and the work authorization has been granted. This indicates that the product is deemed unqualified and is prohibited from entering the reactor. This represents the actual concentration of harmful gases inside the vessel. The safe concentration threshold for harmful gases; This represents the actual oxygen concentration inside the reactor. These are the lower and upper limits of safe oxygen concentration, respectively. This is the power-off status flag. This indicates that the power circuit of the reactor has been disconnected.

[0198] As a preferred embodiment of a reactor cleaning and discharge control system based on multi-sensor fusion, the targeted cleaning control unit 200 is also used to verify the effect after the cleaning operation is completed, and the determination is made by a quantitative indicator of cleanliness, calculated as follows:

[0199] Residual coverage inside the vessel:

[0200]

[0201] Maximum residual thickness inside the vessel:

[0202]

[0203] The criteria for judging cleanliness are as follows:

[0204] and

[0205] In the formula, This refers to the coverage rate of residual materials inside the vessel; The entire inner surface area of ​​the reactor; coordinates The residual region indicator function takes a value of 1 when there is a residual and a value of 0 when there is no residual. This represents the total surface area of ​​the reactor's inner surface. This represents the maximum residual thickness across the entire region. The threshold for acceptable residual coverage; This is the maximum acceptable threshold for residual thickness.

[0206] As a preferred solution for a reactor cleaning and discharge control system based on multi-sensor fusion, the discharge adaptive stirring unit 300 uses a continuous flow zero-value time-series detection model to determine the completion of the discharge operation, satisfying the following judgment logic:

[0207]

[0208] In the formula, The material feeding is complete. This indicates that the material feeding operation is complete; for Instantaneous flow rate of material discharge at any given moment; This is the moment when the flow rate first drops to zero; The preset duration for continuous judgment.

[0209] It should be noted that the information interaction and execution process between the various units of the above system are based on the same concept as the method embodiment in Embodiment 1 of this application, and the resulting technical effects are the same as those in the method embodiment of this application. For details, please refer to the description in the method embodiment shown above in this application, and it will not be repeated here.

[0210] Example 3

[0211] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium storing program code for a reactor cleaning and discharge control method based on multi-sensor fusion. The program code includes instructions for executing the reactor cleaning and discharge control method based on multi-sensor fusion as described in Embodiment 1 or any possible implementation thereof.

[0212] Computer-readable storage media can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0213] Example 4

[0214] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0215] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor can call the program instructions to execute the reactor cleaning and discharge control method based on multi-sensor fusion in Embodiment 1 or any possible implementation thereof.

[0216] Specifically, a processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. This memory can be integrated into the processor or located outside the processor and exist independently.

[0217] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0218] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0219] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for controlling reactor cleaning and discharge based on multi-sensor fusion, characterized in that, Includes the following steps: Reactor interior status perception: Multi-source data fusion detection is performed by visual sensors and ultrasonic sensors that can extend into the reactor interior to obtain the overall distribution status of material scraping residue on the reactor wall and large material residue on the reactor bottom, and the overall distribution status is used as the decision input for reactor cleaning operation; Targeted cleaning control: Based on the global distribution state, the corresponding cleaning execution parameters are matched and the spray angle, spray pressure and travel trajectory of the high-pressure water gun are controlled to perform targeted cleaning operations on the residual areas on the inner surface of the reactor. Adaptive stirring during material discharge: During the material discharge operation of the reactor, the material discharge flow rate data at the discharge port is collected in real time. Using the material discharge flow rate data as feedback, the operating frequency of the stirring mechanism is dynamically matched through an adaptive stirring frequency algorithm to maintain a stable flow rate during the material discharge process. Safety interlock for entering the vessel: When the need for personnel to enter the vessel is triggered, the power-off interlock, gas purging inside the vessel, and harmful gas concentration detection are executed in sequence according to the preset safety sequence. Once all safety indicators meet the standards, the permission to enter the vessel is granted, so as to control the interlock for entering the vessel.

2. The reactor cleaning and discharge control method based on multi-sensor fusion according to claim 1, characterized in that, The visual sensor and the ultrasonic sensor are integrated at the end of a height-adjustable and rotatable detection bracket, and the spatial coverage of the full-area scanning satisfies the cylindrical coordinate model: ， In the formula, The scannable spatial region of the sensor; These are the radial, circumferential, and axial coordinates in the cylindrical coordinate system inside the vessel, respectively. The inner radius of the reactor; The height is the internal height of the reactor; the detection support achieves full-area scanning inside the reactor through a combination of circumferential rotation and axial lifting.

3. The reactor cleaning and discharge control method based on multi-sensor fusion according to claim 1, characterized in that, The multi-source data fusion is performed through spatial coordinate registration and a weighted fusion model, specifically including: Establish the projection mapping relationship between visual pixel coordinates and three-dimensional spatial coordinates within the vessel. The coordinate transformation formula is as follows: ， In the formula, The depth scale factor is determined jointly by the geometric constraints of the reactor body and ultrasonic ranging data. The projection matrix of the 3×4D vision sensor includes sensor intrinsic parameters, installation pose extrinsic parameters, and vessel coordinate system transformation parameters; u and v are the pixel coordinates of the residual points in the visual image; The three-dimensional spatial coordinates of the residual points inside the reactor in the reactor body coordinate system; The residual thickness data detected by the ultrasonic sensor and the position data recognized by vision are spatially registered and then weighted and fused to obtain the fused estimate of the residual thickness: ， In the formula, Spatial coordinates The estimated value of the residual thickness of the material at the location; The residual thickness at this location as identified by the visual sensor; The residual thickness at this location, as detected by the ultrasonic sensor; These are the weighting coefficients for visual inspection and ultrasonic inspection, respectively, and they satisfy... .

4. The reactor cleaning and discharge control method based on multi-sensor fusion according to claim 1, characterized in that, Based on the cleaning execution parameters corresponding to the global distribution state, during the adjustment of the high-pressure water gun's spray angle, spray pressure, and travel trajectory, the high-pressure water gun's spray pressure and travel feed speed are adaptively matched according to the residual characteristics in the global distribution state, satisfying the following control model: Injection pressure control formula: ， In the formula, Spatial coordinates The high-pressure water jet pressure at the location; Based on the cleaning pressure; This is the pressure gain coefficient; This represents the residual fusion thickness of the material at that location; Formula for controlling the feed rate of the injection trajectory: ， In the formula, Spatial coordinates The travel speed of the high-pressure water gun; The reference feed rate; This is the velocity attenuation coefficient; This represents the residual coverage rate within the neighborhood of that location.

5. The reactor cleaning and discharge control method based on multi-sensor fusion according to claim 4, characterized in that, The position and orientation of the high-pressure water gun's end nozzle are controlled by a multi-degree-of-freedom robotic arm, and the spatial orientation transformation of the nozzle satisfies a homogeneous transformation relationship: ， In the formula, This is the homogeneous transformation matrix of the nozzle relative to the robotic arm base, used to describe the spatial position and spraying attitude of the nozzle; For the first robotic arm The homogeneous transformation matrix corresponding to each joint; For the first The rotation angle of each joint; This represents the total number of degrees of freedom of the robotic arm.

6. The reactor cleaning and discharge control method based on multi-sensor fusion according to claim 1, characterized in that, The adaptive stirring frequency algorithm adopts an incremental closed-loop adjustment model, and the real-time adjustment increment of the stirring frequency satisfies the following formula: ， ； In the formula, For the first Time relative to the first Adjust the stirring frequency increment at any given time; The first The deviation value of the material discharge flow rate at any given time; For flow deviation, Set the target feed rate; For the first The actual discharge flow rate detected at all times; This is the proportional adjustment coefficient; This is the integral adjustment coefficient; This is the differential adjustment coefficient.

7. The reactor cleaning and discharge control method based on multi-sensor fusion according to claim 1, characterized in that, The safety qualification determination of the reactor entry operation adopts a multi-index joint determination model, and the determination logic satisfies the following relation: ， In the formula, The result of the safety permit determination for entering the vessel. This indicates that the operation has been deemed successful and the work authorization has been granted. This indicates that the product is deemed unqualified and is prohibited from entering the reactor. This represents the actual concentration of harmful gases inside the vessel. The safe concentration threshold for harmful gases; This represents the actual oxygen concentration inside the reactor. These are the lower and upper limits of safe oxygen concentration, respectively. This is the power-off status flag. This indicates that the power circuit of the reactor has been disconnected.

8. The reactor cleaning and discharge control method based on multi-sensor fusion according to claim 1, characterized in that, This also includes a review of the effectiveness of the cleaning operation after completion, which is judged by quantitative indicators of cleanliness, calculated using the following formula: Residual coverage inside the vessel: ， Maximum residual thickness inside the vessel: ， The criteria for judging cleanliness are as follows: and , In the formula, This refers to the coverage rate of residual materials inside the vessel; The entire inner surface area of ​​the reactor; coordinates The residual region indicator function takes a value of 1 when there is a residual and a value of 0 when there is no residual. This represents the total surface area of ​​the reactor's inner surface. This represents the maximum residual thickness across the entire region. The threshold for acceptable residual coverage; This is the maximum acceptable threshold for residual thickness.

9. The reactor cleaning and discharge control method based on multi-sensor fusion according to claim 1, characterized in that, This also includes the use of a continuous flow zero-value time-series detection model to determine the completion of the material feeding operation, satisfying the following judgment logic: ， In the formula, The material feeding is complete. This indicates that the material feeding operation is complete; for Instantaneous flow rate of material discharge at any given moment; This is the moment when the flow rate first drops to zero; The preset duration for continuous judgment.

10. A reactor cleaning and discharge control system based on multi-sensor fusion, employing the reactor cleaning and discharge control method based on multi-sensor fusion as described in any one of claims 1 to 9, characterized in that, include: The reactor interior state sensing unit is used to perform multi-source data fusion detection through visual sensors and ultrasonic sensors that can extend into the reactor interior, to obtain the overall distribution of scraped material residue on the reactor wall and large material residue on the reactor bottom, and to use the overall distribution as the decision input for the reactor cleaning operation. The targeted cleaning control unit is used to match the corresponding cleaning execution parameters based on the global distribution state, and to control the spray angle, spray pressure and travel trajectory of the high-pressure water gun to perform targeted cleaning operations on the residual areas on the inner surface of the reactor. The discharge adaptive stirring unit is used to collect the discharge flow rate data at the discharge port in real time during the discharge operation of the reactor. Using the discharge flow rate data as feedback, the unit dynamically matches the operating frequency of the stirring mechanism through an adaptive stirring frequency algorithm to maintain a stable flow rate during the discharge process. The vessel entry safety interlock unit is used to execute power-off interlock, vessel gas purging, and harmful gas concentration detection in a preset safety sequence when the need for personnel to enter the vessel is triggered. Once all safety indicators meet the standards, entry permission is granted to control the vessel entry operation.

Images