Intelligent temperature control self-adaptive reaction kettle cleaning method

Abstract

The intelligent temperature control self-adaptive reaction kettle cleaning method comprises the following steps: firstly, carrying out system preparation and initialization, and setting cleaning parameters and reinforcement learning algorithm parameters; positioning and detecting the reaction kettle, and obtaining the pollution degree of the inner wall as an initial state input; then the intelligent control system determines the temperature, the reinforcement learning algorithm adjusts the temperature according to the state, the controller is used for controlling the heating coil to heat, and the temperature is accurately controlled through the PID algorithm; then starting a washing mechanism to clean the reaction kettle, adjusting the rotating speed of the washing mechanism according to the cleaning state through a reinforcement learning algorithm, spraying heated cleaning liquid by a water pump to cooperate with cleaning, monitoring the cleaning effect, temperature and pressure in real time in the cleaning process, and adjusting a cleaning strategy through the reinforcement learning algorithm according to feedback information; and finally, the intelligent control system judges that cleaning is finished, equipment is reset, the reaction kettle is cleaned, checked and dried, and cleaning data is recorded for subsequent analysis and optimization. The self-adaptive cleaning is realized through a reinforcement learning algorithm, and the cleaning quality and the safe operation of the system can be ensured.

Description

Technical Field

[0001] The invention relates to the technical field of chemical equipment, and in particular to an intelligent temperature-controlled self-adaptive reactor cleaning method. Background Art

[0002] Reactors are widely used key equipment in many industries such as chemical, pharmaceutical, and food, and are used for various chemical reactions, material mixing, processing, and other operations. However, during the use of the reactor, various materials such as reactants, products, and catalysts will inevitably remain on the inner wall. If these residual materials are not cleaned thoroughly and in time, they will cause a series of serious problems.

[0003] On the one hand, the residual materials will accumulate, dry up, and deteriorate in the reactor, which will not only reduce the effective volume of the reactor, affecting subsequent production efficiency and product quality, but may also trigger chemical reactions, produce harmful substances, and pollute the products. In serious cases, the entire batch of products may be scrapped, causing huge economic losses to the company. On the other hand, the long-term accumulation of residues will corrode the inner wall of the reactor, shorten the service life of the reactor, and increase the cost of equipment maintenance and replacement. At the same time, if the residual materials are cross-contaminated between different batches of production, it may cause safety accidents and threaten the life and health of operators.

[0004] Traditional reactor cleaning methods mainly include manual cleaning, chemical cleaning and simple mechanical cleaning, etc. Manual cleaning is inefficient, labor-intensive, and difficult to ensure uniformity and thoroughness of cleaning; although chemical cleaning can dissolve residual materials to a certain extent, the use of chemical reagents may damage the reactor material, and improper treatment of waste liquid after cleaning will pollute the environment; ordinary mechanical cleaning often cannot be intelligently adjusted according to the actual situation of the reactor, and has poor adaptability. Summary of the invention

[0005] To solve the above problems, the present invention proposes an intelligent temperature-controlled adaptive reactor cleaning method, which achieves efficient cleaning, energy saving and consumption reduction, and data optimization application through system preparation, intelligent temperature control, dynamic cleaning and monitoring adjustment. The specific steps are as follows, which are characterized by:

[0006] Step 1: System preparation and initialization: install the cleaning device, spraying mechanism, scrubbing mechanism and temperature adjustment mechanism, initialize the system, set the cleaning parameters, and initialize the reinforcement learning algorithm;

[0007] Step 2: Positioning and detecting the reactor: placing the reactor in the center and detecting the degree of contamination on the inner wall as the initial state input of the reinforcement learning algorithm;

[0008] Step 3: The temperature of the cleaning fluid is adjusted. The intelligent control system determines the temperature, the reinforcement learning algorithm adjusts the temperature according to the state, and the controller controls the heating coil to heat and ensure the temperature is stable.

[0009] Step 4: Clean the reactor, start the scrubbing mechanism, and rotate the cleaning brush against the inner wall of the reactor to clean it. The reinforcement learning algorithm adjusts the speed of the scrubbing mechanism, and the water pump sprays the heated cleaning liquid to clean it with the cleaning brush.

[0010] Step 5: Cleaning process monitoring and adjustment, real-time monitoring of cleaning effect, temperature and pressure, and reinforcement learning algorithm to adjust the cleaning strategy based on feedback information;

[0011] Step 6: After cleaning is completed and subsequent processing, the intelligent control system determines that cleaning is complete, resets the equipment, cleans and inspects the reactor, performs drying, and records cleaning data for subsequent analysis and optimization strategies.

[0012] Furthermore, the system preparation and initialization in step 1 is expressed as:

[0013] Step 1.1, install the fixed frame, plug and connect the fixed frame with the positioning column on the mobile seat through its bottom, and insert the fixing holes on the side wall of the positioning column into the fixing pins and other parts to ensure that the fixed frame is firmly installed. The universal wheels under the mobile seat facilitate the movement of the entire device in workshops and other places, and are convenient for adjusting the cleaning position;

[0014] Step 1.2, install the cleaning seat plate inside the fixed frame, so that it is movably connected with the inside of the frame, to ensure that the cleaning seat plate can be flexibly lifted or rotated in the frame; install telescopic electric cylinders on both sides of the fixed frame, the telescopic electric cylinders are connected to the frame through the fixed seat, the pin shaft at the connection ensures the reliability and rotatability of the connection, and the output end of the telescopic electric cylinder is rotatably connected with the cleaning seat plate to realize the lifting and lowering control of the cleaning seat plate;

[0015] Step 1.3, install the spray mechanism, fix the water tank in a suitable position, install a water pump inside the water tank, open a drain port at the connection between the water pump and the side wall of the water tank and connect a valve, the valve is used to control the outflow of the cleaning liquid; the drain port is connected to the temperature adjustment mechanism through a hose to ensure that the cleaning liquid can be smoothly transported to the temperature adjustment mechanism;

[0016] Step 1.4, install the washing mechanism, fix the washing motor on the upper surface of the washing seat plate, the central shaft penetrates the washing seat plate and is connected to the output end of the washing motor through a chain to realize power transmission; fix the cleaning brush on the surface of the central shaft, and the cleaning brush and the central shaft are movably connected through a connecting arm with a spring. The spring provides a certain elastic force, so that the cleaning brush can fit tightly on the inner wall of the reactor with different inner diameters to ensure the cleaning effect; the central shaft is provided with a flow channel and a spray hole for the subsequent spraying of the washing liquid;

[0017] Step 1.5, install the temperature regulating mechanism, cover the surface of the heating tank with the insulation barrel to reduce heat loss; install the heating coil between the heating tank and the insulation barrel close to the surface of the heating tank, the heating coil generates heat when powered on to heat the cleaning liquid in the heating tank; fix the controller on the outer wall of the insulation barrel, the controller is connected to the heating coil to realize the control of the heating coil; the lower end of the heating tank is provided with a water inlet connected to the spraying mechanism hose, and the upper end of the water outlet is connected to the washing mechanism, so that the cleaning liquid can circulate in the system; a spiral lining is fixed inside the heating tank to extend the residence time of the cleaning liquid in the heating tank and improve the heating efficiency; a number of temperature sensors are connected to the side wall to monitor the temperature of the cleaning liquid in real time;

[0018] Step 1.6, initialize the cleaning parameters and set the cleaning time t according to the material characteristics of the reactor. 清 :

[0019] t 清 =45+15×(HH 阈值1 )

[0020] Among them, H is the hardness of the reactor material, H 阈值1 The material hardness threshold of the cleaning time, the speed n of the scrubbing mechanism is set to:

[0021] n=100-10×(H 阈值2 -H)

[0022] Among them, H 阈值2 Set the cleaning temperature T as the material hardness threshold of the rotation speed 清

[0023] T 清 =T 软 +ΔT 安全

[0024] Among them, the softening temperature of the residual material is T 软 , ΔT 安全 Temperature margin to consider safety and cleaning effect;

[0025] Step 1.7, initialize the reinforcement learning algorithm parameters, set the learning rate α = 0.05, the number of algorithm iterations to N, the discount factor to γ ​​= 0.95, and the algorithm space state vector s can be expressed as:

[0026] s=(s 1 ,s 2 ,s 3 )

[0027] Among them, s 1 is the pollution degree of the inner wall of the reactor, s 2 is the cleaning temperature, s 3 is the system pressure, and the spatial action vector a can be expressed as:

[0028] a=(a 1 ,a 2 )

[0029] Among them, a 1 is the speed adjustment value of the scrubbing mechanism, a 2 is the heating coil power adjustment value, and the reward function r can be expressed as:

[0030] r=p 1 ×Δs 1 -p 2 ×E

[0031] Among them, p 1 and p 2 is the weight coefficient of the reward function, Δs 1 is the reduction in pollution level after cleaning, and E is the energy consumption.

[0032] Further, the process of adjusting the temperature of the cleaning fluid in step 3 is expressed as follows:

[0033] Step 3.1, set the initial temperature, and set the upper and lower limits of the initial temperature according to the cleaning liquid temperature and the degree of inner wall contamination;

[0034] Step 3.2, in the policy network of reinforcement learning, each time an action a is taken in state s, a reward function r is obtained. The network parameters are optimized by minimizing the loss function. The loss function is defined as the difference between the predicted reward and the actual reward. The loss function L can be expressed as:

[0035]

[0036] Where n is the number of samples, r pred,i is the predicted reward, r real,i The actual reward is calculated according to the reward function during the training process to guide the algorithm to learn the optimal temperature adjustment strategy. When the algorithm outputs the heating coil power adjustment value action, the adjusted temperature T is calculated according to the power-temperature relationship model of the heating coil. next , and use it as the new temperature setting value;

[0037] Step 3.3, during the heating process, the temperature sensor on the side wall of the heating tank monitors the cleaning temperature T in real time sen And feed the temperature data back to the controller. Assume that the measurement error of the temperature sensor is ΔT sen , the actual temperature received by the controller is T real :

[0038] T real =T sen -ΔT sen

[0039] The controller compares the actual temperature with the set temperature. If the actual temperature exceeds the set temperature range, the controller calculates the power value that needs to be adjusted again according to the preset PID control algorithm to make the cleaning temperature more accurate.

[0040] Further, the process of monitoring and adjusting the cleaning process in step 5 is expressed as follows:

[0041] Step 5.1, convert the image into a grayscale image, and set the grayscale value of each pixel in the image to g(x,y), where x and y are the horizontal and vertical coordinates of the pixel, respectively. By setting the grayscale thresholds g of the residual material area, the cleaning liquid coverage area, and the inner wall of the reactor t1 , g t2 , g t3 , respectively identifying the residual material area, the cleaning liquid covered area, and the original surface area of ​​the reactor inner wall;

[0042] Step 5.2, calculate the area ratio S of the residual material p1

[0043]

[0044] Among them, n p1 is the number of pixels in the residual material area, R is the resolution of the camera, r is the inner radius of the reactor, and L is h is the length that can be photographed in the horizontal direction, L v The vertical shooting height;

[0045] Step 5.3, the cleaning effect of the reactor is represented by the change in the area ratio of the residual material, and the state vector s of the reinforcement learning is updated 1 ;

[0046] Step 5.4, update the state vector s of reinforcement learning through the real-time detection values ​​of temperature sensor and pressure sensor respectively. 2 、s 3 .

[0047] The intelligent temperature-controlled self-adaptive reactor cleaning method of the present invention has beneficial effects: the technical effects of the present invention are:

[0048] 1. The present invention integrates a reinforcement learning algorithm and can dynamically adjust the cleaning strategy according to real-time status information such as the degree of contamination of the inner wall of the reactor, the cleaning temperature, and the system pressure.

[0049] 2. The present invention adopts an intelligent temperature control system, which monitors the temperature of the cleaning liquid in real time through a temperature sensor and performs precise control according to the set temperature.

[0050] 3. The reinforcement learning algorithm of the present invention can also intelligently adjust the power of the heating coil according to the current cleaning status and temperature information, thereby achieving the goals of precise temperature control and energy saving and high efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flow chart of the present invention;

[0052] Figure 2 It is the structural diagram of the reaction kettle of the present invention;

[0053] Figure 3 Schematic diagram of actual working situation. DETAILED DESCRIPTION

[0054] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0055] The present invention proposes an intelligent temperature-controlled self-adaptive reactor cleaning method, the invention flow chart is as follows: Figure 1-3 As shown, the steps of the present invention are described in detail below.

[0056] Step 1: System preparation and initialization: install the cleaning device, spraying mechanism, scrubbing mechanism and temperature adjustment mechanism, initialize the system, set the cleaning parameters, and initialize the reinforcement learning algorithm;

[0057] Step 1.1, install the fixed frame, plug the fixed frame into the positioning column on the mobile seat through its bottom, and insert the fixing holes on the side wall of the positioning column into the fixing pins and other parts to ensure that the fixed frame is firmly installed. The universal wheels under the mobile seat facilitate the movement of the entire device in workshops and other places, and are convenient for adjusting the cleaning position.

[0058] Reactor structure diagram Figure 2 As shown;

[0059] Step 1.2, install the cleaning seat plate inside the fixed frame, so that it is movably connected with the inside of the frame, to ensure that the cleaning seat plate can be flexibly lifted or rotated in the frame; install telescopic electric cylinders on both sides of the fixed frame, the telescopic electric cylinders are connected to the frame through the fixed seat, the pin shaft at the connection ensures the reliability and rotatability of the connection, and the output end of the telescopic electric cylinder is rotatably connected with the cleaning seat plate to realize the lifting and lowering control of the cleaning seat plate;

[0060] Step 1.3, install the spray mechanism, fix the water tank in a suitable position, install a water pump inside the water tank, open a drain port at the connection between the water pump and the side wall of the water tank and connect a valve, the valve is used to control the outflow of the cleaning liquid; the drain port is connected to the temperature adjustment mechanism through a hose to ensure that the cleaning liquid can be smoothly transported to the temperature adjustment mechanism;

[0061] Step 1.4, install the washing mechanism, fix the washing motor on the upper surface of the washing seat plate, the central shaft penetrates the washing seat plate and is connected to the output end of the washing motor through a chain to realize power transmission; fix the cleaning brush on the surface of the central shaft, and the cleaning brush and the central shaft are movably connected through a connecting arm with a spring. The spring provides a certain elastic force, so that the cleaning brush can fit tightly on the inner wall of the reactor with different inner diameters to ensure the cleaning effect; the central shaft is provided with a flow channel and a spray hole for the subsequent spraying of the washing liquid;

[0062] Step 1.5, install the temperature regulating mechanism, cover the surface of the heating tank with the insulation barrel to reduce heat loss; install the heating coil between the heating tank and the insulation barrel close to the surface of the heating tank, the heating coil generates heat when powered on to heat the cleaning liquid in the heating tank; fix the controller on the outer wall of the insulation barrel, the controller is connected to the heating coil to realize the control of the heating coil; the lower end of the heating tank is provided with a water inlet connected to the spraying mechanism hose, and the upper end of the water outlet is connected to the washing mechanism, so that the cleaning liquid can circulate in the system; a spiral lining is fixed inside the heating tank to extend the residence time of the cleaning liquid in the heating tank and improve the heating efficiency; a number of temperature sensors are connected to the side wall to monitor the temperature of the cleaning liquid in real time;

[0063] Step 1.6, initialize the cleaning parameters and set the cleaning time t according to the material characteristics of the reactor. 清 :

[0064] t 清 =45+15×(HH 阈值1 )

[0065] Among them, H is the hardness of the reactor material, H 阈值1 The material hardness threshold of the cleaning time, the speed n of the scrubbing mechanism is set to:

[0066] n=100-10×(H 阈值2 -H)

[0067] Among them, H 阈值2 Set the cleaning temperature T as the material hardness threshold of the rotation speed 清

[0068] T 清 =T 软 +ΔT 安全

[0069] Among them, the softening temperature of the residual material is T 软 , ΔT 安全 Temperature margin to consider safety and cleaning effect;

[0070] Step 1.7, initialize the reinforcement learning algorithm parameters, set the learning rate α = 0.05, the number of algorithm iterations to N, the discount factor to γ ​​= 0.95, and the algorithm space state vector s can be expressed as:

[0071] s=(s 1 ,s 2 ,s 3 )

[0072] Among them, s 1 is the pollution degree of the inner wall of the reactor, s 2 is the cleaning temperature, s 3 is the system pressure, and the spatial action vector a can be expressed as:

[0073] a=(a 1 ,a 2 )

[0074] Among them, a 1 is the speed adjustment value of the scrubbing mechanism, a 2 is the heating coil power adjustment value, and the reward function r can be expressed as:

[0075] r=p 1 ×Δs 1 -p 2 ×E

[0076] Among them, p 1 and p 2 is the weight coefficient of the reward function, Δs 1 is the reduction in pollution level after cleaning, and E is the energy consumption.

[0077] Step 2: Positioning and detecting the reactor: placing the reactor in the center and detecting the degree of contamination on the inner wall as the initial state input of the reinforcement learning algorithm;

[0078] Step 3: The temperature of the cleaning fluid is adjusted. The intelligent control system determines the temperature, the reinforcement learning algorithm adjusts the temperature according to the state, and the controller controls the heating coil to heat and ensure the temperature is stable.

[0079] Step 3.1, set the initial temperature, and set the upper and lower limits of the initial temperature according to the cleaning liquid temperature and the degree of inner wall contamination;

[0080] Step 3.2, in the policy network of reinforcement learning, each time an action a is taken in state s, a reward function r is obtained. The network parameters are optimized by minimizing the loss function. The loss function is defined as the difference between the predicted reward and the actual reward. The loss function L can be expressed as:

[0081]

[0082] Where n is the number of samples, r pred,i is the predicted reward, r real,iThe actual reward is calculated according to the reward function during the training process to guide the algorithm to learn the optimal temperature adjustment strategy. When the algorithm outputs the heating coil power adjustment value action, the adjusted temperature T is calculated according to the power-temperature relationship model of the heating coil. next , and use it as the new temperature setting value;

[0083] Step 3.3, during the heating process, the temperature sensor on the side wall of the heating tank monitors the cleaning temperature T in real time sen And feed the temperature data back to the controller. Assume that the measurement error of the temperature sensor is ΔT sen , the actual temperature received by the controller is T real :

[0084] T real =T sen -ΔT sen

[0085] The controller compares the actual temperature with the set temperature. If the actual temperature exceeds the set temperature range, the controller calculates the power value that needs to be adjusted again according to the preset PID control algorithm to make the cleaning temperature more accurate.

[0086] Step 4: Clean the reactor, start the scrubbing mechanism, and rotate the cleaning brush against the inner wall of the reactor to clean it. The reinforcement learning algorithm adjusts the speed of the scrubbing mechanism, and the water pump sprays the heated cleaning liquid to clean it with the cleaning brush.

[0087] Start the cleaning motor, and the cleaning brush fits tightly against the inner wall of the reactor under the action of the spring and the connecting arm. The reinforcement learning algorithm adjusts the speed of the scrubbing mechanism according to the current cleaning status. The algorithm outputs a speed adjustment value Δn based on the pre-trained strategy network. If the degree of contamination is high and the temperature of the cleaning liquid is within the appropriate range, the strategy network outputs a positive value Δn to increase the speed of the scrubbing mechanism to enhance the cleaning effect. On the contrary, if the inner wall of the reactor is relatively fragile or the cleaning effect is already good, the strategy network outputs a negative value Δn to reduce the speed to avoid excessive cleaning.

[0088] Step 5: Cleaning process monitoring and adjustment, real-time monitoring of cleaning effect, temperature and pressure, and reinforcement learning algorithm to adjust the cleaning strategy based on feedback information;

[0089] During the cleaning process, the inner wall of the reactor is photographed at a high frequency to timely capture the changes in the inner wall state during the cleaning process. For each frame of the image, the cleaning degree is analyzed;

[0090] Step 5.1, convert the image into a grayscale image, and set the grayscale value of each pixel in the image to g(x,y), where x and y are the horizontal and vertical coordinates of the pixel, respectively. By setting the grayscale thresholds g of the residual material area, the cleaning liquid coverage area, and the inner wall of the reactor t1, g t2 , g t3 , respectively identifying the residual material area, the cleaning liquid covered area, and the original surface area of ​​the reactor inner wall;

[0091] Step 5.2, calculate the area ratio S of the residual material p1

[0092]

[0093] Among them, n p1 is the number of pixels in the residual material area, R is the resolution of the camera, r is the inner radius of the reactor, and L is h is the length that can be photographed in the horizontal direction, L v The vertical shooting height;

[0094] Step 5.3, the cleaning effect of the reactor is represented by the change in the area ratio of the residual material, and the state vector s of the reinforcement learning is updated 1 ;

[0095] Step 5.4, update the state vector s of reinforcement learning through the real-time detection values ​​of temperature sensor and pressure sensor respectively. 2 、s 3 ;

[0096] After the algorithm executes the action, the system enters a new state and continues to monitor and provide feedback. By continuously repeating this process, the reinforcement learning algorithm can dynamically adjust the cleaning strategy according to actual conditions to adapt to different cleaning needs and system operating conditions, thereby achieving an efficient and stable cleaning process, ensuring the best cleaning quality, and ensuring the safe operation of the system.

[0097] Step 6: After cleaning is completed and subsequent processing, the intelligent control system determines that cleaning is complete, resets the equipment, cleans and inspects the reactor, performs drying, and records cleaning data for subsequent analysis and optimization strategies.

[0098] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. Intelligent temperature control and adaptive reactor cleaning method, the specific steps are as follows, characterized in that: Step 1: System preparation and initialization: install the cleaning device, spraying mechanism, scrubbing mechanism and temperature adjustment mechanism, initialize the system, set the cleaning parameters, and initialize the reinforcement learning algorithm; Step 2: Positioning and detecting the reactor: placing the reactor in the center and detecting the degree of contamination on the inner wall as the initial state input of the reinforcement learning algorithm; Step 3: The temperature of the cleaning fluid is adjusted. The intelligent control system determines the temperature, the reinforcement learning algorithm adjusts the temperature according to the state, and the controller controls the heating coil to heat and ensure the temperature is stable. Step 4: Clean the reactor, start the scrubbing mechanism, and rotate the cleaning brush against the inner wall of the reactor to clean it. The reinforcement learning algorithm adjusts the speed of the scrubbing mechanism, and the water pump sprays the heated cleaning liquid to clean it with the cleaning brush. Step 5: Cleaning process monitoring and adjustment, real-time monitoring of cleaning effect, temperature and pressure, and reinforcement learning algorithm to adjust the cleaning strategy based on feedback information; Step 6: After cleaning is completed and subsequent processing, the intelligent control system determines that cleaning is complete, resets the equipment, cleans and inspects the reactor, performs drying, and records cleaning data for subsequent analysis and optimization strategies.

2. The intelligent temperature-controlled self-adaptive reactor cleaning method according to claim 1 is characterized in that: The system preparation and initialization in step 1 is expressed as: Step 1.1, install the fixed frame, plug and connect the fixed frame with the positioning column on the mobile seat through its bottom, and insert the fixing holes on the side wall of the positioning column into the fixing pins and other parts to ensure that the fixed frame is firmly installed. The universal wheels under the mobile seat facilitate the movement of the entire device in workshops and other places, and are convenient for adjusting the cleaning position; Step 1.2, install the cleaning seat plate inside the fixed frame, so that it is movably connected with the inside of the frame, to ensure that the cleaning seat plate can be flexibly lifted or rotated in the frame; install telescopic electric cylinders on both sides of the fixed frame, the telescopic electric cylinders are connected to the frame through the fixed seat, the pin shaft at the connection ensures the reliability and rotatability of the connection, and the output end of the telescopic electric cylinder is rotatably connected with the cleaning seat plate to realize the lifting and lowering control of the cleaning seat plate; Step 1.3, install the spray mechanism, fix the water tank in a suitable position, install a water pump inside the water tank, open a drain port at the connection between the water pump and the side wall of the water tank and connect a valve, the valve is used to control the outflow of the cleaning liquid; the drain port is connected to the temperature adjustment mechanism through a hose to ensure that the cleaning liquid can be smoothly transported to the temperature adjustment mechanism; Step 1.4, install the washing mechanism, fix the washing motor on the upper surface of the washing seat plate, the central shaft penetrates the washing seat plate and is connected to the output end of the washing motor through a chain to realize power transmission; fix the cleaning brush on the surface of the central shaft, and the cleaning brush and the central shaft are movably connected through a connecting arm with a spring. The spring provides a certain elastic force, so that the cleaning brush can fit tightly on the inner wall of the reactor with different inner diameters to ensure the cleaning effect; the central shaft is provided with a flow channel and a spray hole for the subsequent spraying of the washing liquid; Step 1.5, install the temperature regulating mechanism, cover the surface of the heating tank with the insulation barrel to reduce heat loss; install the heating coil between the heating tank and the insulation barrel close to the surface of the heating tank, the heating coil generates heat when powered on to heat the cleaning liquid in the heating tank; fix the controller on the outer wall of the insulation barrel, the controller is connected to the heating coil to realize the control of the heating coil; the lower end of the heating tank is provided with a water inlet connected to the spraying mechanism hose, and the upper end of the water outlet is connected to the washing mechanism, so that the cleaning liquid can circulate in the system; a spiral lining is fixed inside the heating tank to extend the residence time of the cleaning liquid in the heating tank and improve the heating efficiency; a number of temperature sensors are connected to the side wall to monitor the temperature of the cleaning liquid in real time; Step 1.6, initialize the cleaning parameters and set the cleaning time t according to the material characteristics of the reactor. 清 : t 清 =45+15×(H-H 阈值1 ) Among them, H is the hardness of the reactor material, H 阈值1 The material hardness threshold of the cleaning time, the speed n of the scrubbing mechanism is set to: n=100-10×(H 阈值2 -H) Among them, H 阈值2 Set the cleaning temperature T as the material hardness threshold of the rotation speed 清 T 清 =T 软 +ΔT 安全 Among them, the softening temperature of the residual material is T 软 , ΔT 安全 Temperature margin to consider safety and cleaning effect; Step 1.7, initialize the reinforcement learning algorithm parameters, set the learning rate α = 0.05, the number of algorithm iterations to N, the discount factor to γ ​​= 0.95, and the algorithm space state vector s can be expressed as: s=(s1,s2,s3) Among them, s1 is the degree of contamination of the inner wall of the reactor, s2 is the cleaning temperature, s3 is the system pressure, and the spatial action vector a can be expressed as: a=(a1,a2) Among them, a1 is the speed adjustment value of the scrubbing mechanism, a2 is the power adjustment value of the heating coil, and the reward function r can be expressed as: r=p1×Δs1-p2×E Among them, p1 and p2 are the weight coefficients of the reward function, Δs1 is the reduction in pollution level after cleaning, and E is the energy consumption.

3. The intelligent temperature-controlled self-adaptive reactor cleaning method according to claim 1 is characterized in that: The process of adjusting the temperature of the cleaning fluid in step 3 is as follows: Step 3.1, set the initial temperature, and set the upper and lower limits of the initial temperature according to the cleaning fluid temperature and the degree of inner wall contamination; Step 3.2, in the policy network of reinforcement learning, each time an action a is taken in state s, a reward function r is obtained. The network parameters are optimized by minimizing the loss function. The loss function is defined as the difference between the predicted reward and the actual reward. The loss function L can be expressed as: Where n is the number of samples, r pred,i is the predicted reward, r real,i The actual reward is calculated according to the reward function during the training process to guide the algorithm to learn the optimal temperature adjustment strategy. When the algorithm outputs the heating coil power adjustment value action, the adjusted temperature T is calculated according to the power-temperature relationship model of the heating coil. next , and use it as the new temperature setting value; Step 3.3, during the heating process, the temperature sensor on the side wall of the heating tank monitors the cleaning temperature T in real time sen And feed the temperature data back to the controller. Assume that the measurement error of the temperature sensor is ΔT sen , the actual temperature received by the controller is T real : T real =T sen -ΔT sen The controller compares the actual temperature with the set temperature. If the actual temperature exceeds the set temperature range, the controller calculates the power value that needs to be adjusted again according to the preset PID control algorithm to make the cleaning temperature more accurate.

4. The intelligent temperature-controlled self-adaptive reactor cleaning method according to claim 1 is characterized in that: The process of monitoring and adjusting the cleaning process in step 5 is shown as follows: Step 5.1, convert the image into a grayscale image, and set the grayscale value of each pixel in the image to g(x,y), where x and y are the horizontal and vertical coordinates of the pixel, respectively. By setting the grayscale thresholds g of the residual material area, the cleaning liquid coverage area, and the inner wall of the reactor t1 , g t2 , g t3 , respectively identifying the residual material area, the cleaning liquid covered area, and the original surface area of ​​the reactor inner wall; Step 5.2, calculate the area ratio S of the residual material p1 Among them, n p1 is the number of pixels in the residual material area, R is the resolution of the camera, r is the inner radius of the reactor, and L is h is the length that can be photographed in the horizontal direction, L v The vertical shooting height; Step 5.3, the cleaning effect of the reactor is represented by the change in the area ratio of the residual material, and the state vector s1 of the reinforcement learning is updated; Step 5.4, update the state vectors s2 and s3 of the reinforcement learning through the real-time detection values ​​of the temperature sensor and the pressure sensor, respectively.

Images