Printing and dyeing process parameter optimization control method and control system

Abstract

The invention belongs to the technical field of printing and dyeing process parameter optimization control, and discloses a printing and dyeing process parameter optimization control method and system, and the method comprises the steps: collecting a temperature sequence, a steam parameter, a dyeing process and other first parameters, extracting short-term abnormality and long-term dependence characteristics through a multilayer LSTM, and obtaining a first parameter; and generating an implicit vector containing the temperature state, the temperature rise trend and the equipment state. And taking the implicit vector as state input, combining a discrete action space and a multi-target reward function, calculating a Q value through Q-Learning, and selecting an optimal temperature rise action. The genetic algorithm is used for optimizing Q-Learning hyper-parameters, a parameter freezing and self-adaptive fine tuning mechanism is set, the device working condition change is adapted, and the problems that traditional control is low in precision, high in energy consumption, poor in adaptability and the like are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of printing and dyeing process parameter optimization control, in particular to a printing and dyeing process parameter optimization control method, a control system, an electronic device, a medium and a computer program product. BACKGROUND

[0002] In the textile dyeing and finishing industry, the temperature control of the intermittent dyeing machine is crucial, but the traditional control method has many bottlenecks. Manual or fixed program control is difficult to adapt to the nonlinear change of the equipment, the temperature control precision is insufficient, the equipment adaptability is poor, the energy waste is serious, the equipment exception can only be responded passively, and the production efficiency is low.

[0003] Therefore, there is an urgent need for a printing and dyeing process parameter optimization control method for optimizing the temperature control of the intermittent dyeing machine. SUMMARY

[0004] To this end, the present application provides a printing and dyeing process parameter optimization control method, a control system, an electronic device, a medium and a computer program product to at least partially solve the above technical problems.

[0005] The present application provides a printing and dyeing process parameter optimization control method, comprising the following method steps: obtaining the first parameter of the intermittent dyeing machine, the first parameter including temperature sequence data, steam flow and pressure parameter and dyeing parameter in the current preset time, the dyeing parameter including fabric type, dye type, dye liquor pH value and target dyeing temperature; based on the time sequence analysis of the first parameter by the multi-layer LSTM unit, the correlation mode of temperature abnormal fluctuation and equipment parameter change is identified, and the hidden vector representation is output, the hidden vector representation including the current temperature state, the temperature rising trend and the equipment running state; the hidden vector representation is taken as the state input, combined with the discretized action space and the preset reward function, the Q value of each action is calculated by the Q-Learning algorithm, the optimal temperature control action is selected, the control action including the temperature rising rate adjustment and the steam valve opening degree adjustment; based on the control action, the printing and dyeing of the intermittent dyeing machine is controlled.

[0006] In another aspect, the application provides a printing and dyeing process parameter optimization control system, comprising: a first acquisition module configured to acquire first parameters of an intermittent dyeing machine, the first parameters comprising temperature sequence data, steam flow and pressure parameters, and dyeing parameters in a preset time, the dyeing parameters comprising fabric type, dye type, dye solution pH value, and target dyeing temperature; a first output module configured to perform time sequence analysis on the first parameters based on a multi-layer LSTM unit, identify an association mode of temperature abnormal fluctuation and equipment parameter change, and output a hidden vector representation, the hidden vector representation comprising a current temperature state, a temperature rising trend, and an equipment running state; a second output module configured to take the hidden vector representation as a state input, combine a discretized action space and a preset reward function, calculate Q values of each action by a Q-Learning algorithm, and select an optimal temperature rising control action, the control action comprising temperature rising rate adjustment and steam valve opening degree adjustment; and a control module configured to control the intermittent dyeing machine to perform printing and dyeing based on the control action.

[0007] In another aspect, the application provides an electronic device, comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the printing and dyeing process parameter optimization control method as described above.

[0008] In another aspect, the application provides a computer-readable storage medium having stored thereon computer program instructions executable by a processor to implement the printing and dyeing process parameter optimization control method as described above.

[0009] In another aspect, the application provides a computer program product comprising a computer program executable by a processor to implement the printing and dyeing process parameter optimization control method as described above.

[0010] Through the scheme provided by the embodiments of the application, the hierarchical feature extraction of the temperature sequence by the multi-layer LSTM and the dynamic decision of the Q-Learning greatly reduce the temperature control deviation, significantly improve the dyeing uniformity, and greatly improve the dyeing quality stability; the optimized temperature rising strategy reduces the unit steam consumption and shortens the production cycle, effectively solving the problems of low control precision, high energy consumption, and poor adaptability of the traditional control. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor by reading the following detailed description of the non-limiting embodiments with reference to the following drawings.

[0012] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the drawings.

[0013] Figure 1 A printing process parameter optimization control method provided by the embodiments of the present application is shown in the schematic diagram.

[0014] Figure 2 A structure schematic diagram of a printing process parameter optimization control system provided by the embodiments of the present application is shown in the schematic diagram.

[0015] Figure 3 A structure schematic diagram of an electronic device provided by the embodiments of the present application is shown in the schematic diagram. DETAILED DESCRIPTION

[0016] In order to make the purposes, technical solutions and advantages of the embodiments of the present application more clear, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0017] In a typical configuration of the present application, the devices of the terminal and the service network each include one or more processors (CPU), input / output interfaces, network interfaces and memories.

[0018] The memory can include a non-permanent memory in a computer readable medium, a random access memory (RAM) and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory (flash RAM). The memory is an example of the computer readable medium.

[0019] Computer-readable media includes permanent and non-permanent, removable and non-removable media, which can be implemented by any method or technology to store information. Information can be computer program instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage device, or any other non-transmission medium that can be used to store information accessible by a computing device.

[0020] As shown in Figure 1 The embodiment of the present application discloses a printing and dyeing process parameter optimization control method schematic diagram, including the following method steps: step S101, acquiring the first parameter of the intermittent dyeing machine, the first parameter includes temperature sequence data in the current preset time, steam flow and pressure parameter and dyeing parameter, the dyeing parameter includes fabric type, dye type, dye liquor pH value and target dyeing temperature.

[0021] In some embodiments, the current preset time can be 10-30 minutes in the past (which can be adjusted according to the device response delay), for example, the preset time is 30 minutes, the temperature data of the current time and the previous 29 minutes is collected, the sampling frequency is ≥1 times / minute, the time dependence of temperature change (such as linear heating, rate mutation) is captured, exemplarily, temperature collection can be through temperature sensor, multiple points are arranged in the dyeing vat (such as top, middle and bottom), local overheating or temperature blind area is avoided. At the same time, the steam volume or mass flow entering the dyeing vat (unit: m³ / h or kg / h) is monitored in real time, the real-time pressure value (unit: MPa or kPa) in the steam pipeline is detected, exemplarily, the steam flowmeter is installed in the steam inlet pipeline, the vortex flowmeter or orifice flowmeter is selected, the accuracy is ≥±1%; exemplarily, the pipeline pressure gauge selects diffusion silicon pressure transmitter, the range covers 0-1.6MPa, the response time is ≤50ms.

[0022] Specifically, the dyeing parameters include fabric type, dye type, dye liquor pH value and target dyeing temperature, the fabric type includes cotton, polyester, wool, blended fabric, etc., which affects the heat transfer efficiency and temperature resistance limit (for example, wool can resist ≤95℃, polyester can resist 130℃ high temperature); the dye type includes, for example, reactive dye, disperse dye, acid dye, etc., different dyes have different dyeing temperature windows (for example, disperse dye needs 120-130℃, reactive dye is usually at 60-90℃); the dye liquor pH value reflects the acidity or alkalinity of the dye liquor, which affects the dye solubility and dyeing rate (for example, reactive dye is easy to fix in alkaline conditions); the target dyeing temperature represents the final constant temperature set by the process (for example, 110℃), which directly affects the end point control of the temperature rising curve.

[0023] In step S102, time series analysis is performed on the first parameters based on a multi-layer LSTM unit, an associated mode of temperature abnormal fluctuation and equipment parameter change is identified, and a hidden vector representation is output, the hidden vector representation including a current temperature state, a temperature rising trend and an equipment running state.

[0024] Preferably, the multi-layer LSTM unit includes a first layer LSTM unit and a second layer LSTM unit, the first layer LSTM unit receives the first parameters, processes the first parameters through an input gate, a forgetting gate and an output gate, and outputs a first hidden state containing short-term features; the second layer LSTM unit receives the first hidden state and the first parameters, and outputs a second hidden state containing high-level state abstraction.

[0025] In some embodiments, the multi-layer LSTM extracts features by layers, mines dynamic patterns of different time scales from the first parameters (temperature sequence, steam parameters, dyeing process, etc.), and the specific process can include: the first layer LSTM processes the original first parameters to capture short-term (minute-level) temperature changes and local anomalies; the second layer LSTM combines the first layer output and the original parameters to analyze long-term (more than ten minutes) dependencies and multivariate correlations; the hidden vector generation integrates the processing results of the two layers into an abstract representation containing temperature state, temperature rising trend and equipment state.

[0026] The first layer LSTM is a basic feature extraction layer, including an input gating system for receiving real-time data such as original temperature sequence, steam pressure, flow rate, etc. as input . The Input Gate selectively filters the valid information in the current data, such as the linear trend segment of "2℃ per minute temperature rise". The Cell State, as the storage unit, discards irrelevant historical information (e.g., minor fluctuations in stable segments) through the Forget Gate and retains short-term trend characteristics (e.g., the slope of continuous temperature rise or fall). For example: when the temperature suddenly drops by 0.5℃, the Forget Gate suppresses the memory of the regular trend, the Input Gate strengthens the record of the abnormal point, and the memory unit focuses on the local fluctuation. The Output Gate controls the output of the hidden state at the current time according to the content of the memory unit , which contains short-term characteristics (such as the instantaneous temperature rise rate and abnormal point identification). Both as the output of the first layer to the second layer and participating in the state update of itself at the next time step.

[0027] The second layer LSTM is a high-level semantic abstraction layer, which is used to integrate the local features of the first layer, and to mine long-term dependencies and multi-variable associations. For example, the composite input includes two parts: the original data (e.g., pressure, temperature sensor values) at the current time and the hidden state (encoded short-term features) of the first layer. For example: receiving the "3-minute continuous temperature rise rate drop" feature passed by the first layer, combining the current pressure data, and analyzing the long-term trend. The long-time dependency processing unit: the memory unit, which is used to filter short-term noise and retain cross-time scale dependencies through the cooperation of the Forget Gate and the Input Gate. For example, integrating the multi-level feature of "10-minute continuous temperature rise rate drop" to infer the risk of "final temperature deficiency", and the memory unit continuously accumulates the weight of this trend. Multi-variable association modeling: the output gate generates the second layer hidden state according to the content of the memory unit, which encodes the delay association between variables (e.g., the causal pattern of "temperature drop 5 minutes after 20% pressure drop"). It can be understood that, The abstract representation of short-term features (from the first layer) and long-term association features are fused, and the results of the two layers are integrated into an abstract representation containing temperature state, temperature rise trend, and device state.

[0028] As an example, the cotton fabric active dyeing process is taken as an example to illustrate the workflow of multi-layer LSTM, where the target temperature is 90℃, and the first parameter within 30 minutes is as follows: temperature sequence: from 60℃ to 80℃, the temperature rise rate is 2℃ / min for the first 20 minutes, and the temperature drops to 79.5℃ at the 21st minute.

[0029] Steam pressure: stable at 0.4MPa for the first 20 minutes, and drops to 0.35MPa at the 21st minute.

[0030] Fabric type: cotton; dye type: reactive dye; pH value: 10 (alkaline).

[0031] First-layer LSTM processing: input gate identifies the "2℃ / min linear ramp-up" in the first 20 minutes as a valid feature and allows it to enter the memory unit; when the temperature drops 0.5℃ at the 21st minute, the input gate strengthens the record of this abnormal point (marked as "temperature mutation").

[0032] Forget gate ignores minor fluctuations (±0.2℃) in the stable section (e.g., 65-75℃) and retains key information such as the ramp-up slope; when the steam pressure drops, the forget gate suppresses the historical memory of regular pressure fluctuations and focuses on the current anomaly.

[0033] Output first hidden state: contains short-term features such as "current temperature 80℃, average ramp-up rate 2℃ / min in the last 20 minutes, temperature drop 0.5℃ at the 21st minute, and steam pressure drop 5%".

[0034] Second-layer LSTM processing: composite input integration: the first-layer hidden state includes short-term anomalies such as "temperature sudden drop + pressure drop"; original parameters: cotton fabric temperature resistance (≤100℃), reactive dye dyeing temperature window (60-90℃), target temperature 90℃ (current 80℃, 10℃ away from target).

[0035] Long-term dependency mining: analyzing the data from the first 21 minutes, it finds a pattern that "after a 5% continuous pressure drop, the temperature increase slows down", predicting that if the pressure continues to drop, the final temperature may not reach 90℃; combined with the process characteristics of reactive dyes that dyeing rate increases above 80℃, it judges that the current temperature drop may affect dye activation.

[0036] Multivariate correlation learning: establishes a causal chain of "steam pressure drop → heat supply insufficient → ramp-up rate reduction", calculates that for every 10% drop in pressure, the temperature increase rate decreases by approximately 0.3℃ / min; considering the alkaline condition of pH=10, the heat capacity of the dye liquor is higher, and the temperature recovery speed may be slower than in acidic conditions.

[0037] Output second hidden state: high-level abstract representation is "current temperature 80℃, 10℃ away from target, recent ramp-up rate affected by steam pressure drop reduced to 1.7℃ / min, estimated to reach target temperature in 6 minutes following this trend, but there is a risk of pressure continuous insufficient leading to overtime, equipment status marked as 'pressure anomaly'".

[0038] According to the above example, the hidden vector representation can include the current temperature state: absolute temperature 80℃; target deviation +10℃ (from 90℃); temperature stability near 5-minute fluctuation range ±0.5℃ (normal) / ±1℃ (abnormal); temperature rising trend: instantaneous rate 1.7℃ / min (decreased under pressure); historical average rate 2℃ / min in the past 20 minutes, 1.5℃ / min in the past 1 minute; predicted arrival time 6 minutes (current rate) or 8 minutes (pressure continues to decrease).

[0039] Device operating state: steam pressure 0.35MPa (normal range 0.38-0.42MPa, marked as "mild deficiency"); flow matching current flow 150m³ / h, with a gap from the theoretical requirement of 180m³ / h for the target temperature rising rate; process compliance pH=10 within the suitable range of active dyes (8-11), fabric temperature resistance not exceeding the upper limit.

[0040] Step S103, input the hidden vector representation as the state, combine the discretized action space with the preset reward function, calculate the Q value of each action through the Q-Learning algorithm, and select the optimal temperature rising control action, which includes temperature rising rate adjustment and steam valve opening degree adjustment.

[0041] In some embodiments, Q-Learning is a reinforcement learning algorithm that optimizes the control strategy through closed-loop iteration of "state-action-reward". Referring to the above example, the hidden vector representation is input as the state: state (State): hidden vector output by multi-layer LSTM (including temperature state, temperature rising trend, device state).

[0042] Action (Action): discretized options for temperature rising rate adjustment and steam valve opening degree adjustment.

[0043] Reward (Reward): quantitative feedback based on temperature accuracy, energy consumption, process compliance, etc.

[0044] The ultimate goal is to determine the optimal temperature rising control strategy by maximizing the cumulative reward.

[0045] As an example, consider the hidden vector of the aforementioned cotton fabric dyeing example (assuming current temperature 80°C, distance to target 10°C, temperature ramp rate 1.7°C / min, steam pressure 0.35 MPa), the workflow includes: mapping the continuous hidden vector to a discrete state space, where the temperature deviation: 10°C -> classified as “medium deviation (5-10°C)”; temperature ramp rate: 1.7°C / min -> classified as “medium rate (1.5-2°C / min)”; steam pressure: 0.35 MPa -> classified as “slightly under (normal range ±10%)”. The discrete state is represented as a triple: (temperature deviation = medium, temperature ramp rate = medium, pressure state = slightly under).

[0046] The Q-value table stores the “expected cumulative reward of performing action A in state S”, for example Q[(medium deviation, medium rate, slightly under), increase steam opening] = 7.5, indicating that when in the state of “medium deviation + medium rate + slightly under”, performing the action of “increasing steam opening”, the average reward that can be obtained in the future is 7.5 points.

[0047] Exemplarily, as shown in the following table, the Q-value table shows the expected reward values of performing each action in different states. The states in the table are composed of temperature deviation, temperature ramp rate, and steam pressure state, and the actions include temperature ramp rate adjustment and steam valve opening adjustment, and the Q-values are generated by historical iteration optimization.

[0048]

[0049] Exemplarily, the action combination can include: temperature ramp rate adjustment: -2°C / min (decrease 2), -1°C / min (decrease 1), 0 (maintain), +1°C / min (increase 1), +2°C / min (increase 2); steam valve opening: 20%, 40%, 60%, 80%, 100%.

[0050] Preferably, after each action is performed, the corresponding Q-value is updated according to the actual reward value. For example: initial Q(medium deviation, medium rate, slightly under, 80% opening) = 7.0; after performing the action, the temperature deviation is reduced to 5°C, and a reward of +2 points is obtained, the future reward discount value is 9.5, and the updated Q = 7.9.

[0051] Preferably, the hyperparameters of Q-Learning, including learning rate a and discount factor g, are optimized using a genetic algorithm, specifically, the learning rate a and discount factor g of Q-Learning are encoded as chromosomes; a population of second parameter combinations is initialized, which covers the reasonable value range of a and g; a fitness function is constructed, which is a weighted function for evaluating each second parameter combination in the population; the population is iteratively evolved through roulette selection, crossover and mutation operations, where the crossover operation includes single-point crossover and arithmetic crossover, and the mutation operation is to take the bits of the chromosome complement or add Gaussian noise; in response to the termination of the iterative evolution, the second parameter combination with the highest fitness is taken as the optimal hyperparameters of Q-Learning.

[0052] In some embodiments, the learning rate a (determining the weight of new information update) and the discount factor g (determining the importance of future rewards) of Q-Learning are automatically adjusted by a genetic algorithm. Specifically, first, the learning rate a and the discount factor g of Q-Learning are encoded as chromosomes, which can be represented as floating-point numbers, for example, the chromosome is [a=0.2, g=0.95], or, binary encoding is performed, and a (0-1) and g (0-1) are encoded as 10-bit binary numbers, for example, a=0.2→0011001100, g=0.95→1110111011, and the total length can be adjusted according to actual conditions.

[0053] In this embodiment, the population is initialized, and the second parameter combination, the parameter range can be set as a∈[0.01, 0.99], g∈[0.8, 0.99], covering the full scene of "a small (conservative learning) -a large (aggressive learning)" and "g small (focus on immediate rewards) -g large (focus on long-term rewards)"; the population size is usually set to 50-100 individuals, for example, 50 different a and g combinations, to ensure coverage of the search space.

[0054] In some embodiments, the fitness function is constructed in a multi-objective weighted manner to evaluate the comprehensive performance of each (a, g) in the temperature control, specifically, as an example, the fitness function can be simplified as the formula: Fitness(a, g)=0.4×(1 / MAE)+0.3×(1 / energy consumption)+0.2×(1 / convergence iteration number)-0.1×violation times.

[0055] Wherein, the reciprocal of the average temperature deviation (1 / MAE), the smaller the MAE, the higher the score, the weight 40%, used to represent the temperature control accuracy; the reciprocal of the steam consumption of the unit heating amount (1 / energy consumption), the lower the consumption, the higher the score, the weight 30%, used to represent the energy consumption efficiency; the reciprocal of the number of iterations of Q-Learning to reach the stable strategy (1 / number of iterations), the weight 20%, used to represent the convergence speed; the number of times of violating the fabric temperature limit penalty (10 points deducted for each violation), the weight 10%, used to represent the process compliance, the weight here can be set according to experience, this embodiment is not limited.

[0056] Roulette selection, select parents according to fitness ratio, the higher the fitness of the parameter combination, the higher the probability of being selected, and the top 5% of individuals in each generation (such as the top 3 optimal parameter combinations) are forced to be retained to avoid loss of optimal solution. For example: individual A: Fitness=0.87, selection probability=0.87 / ΣFitness×100%; individual B: Fitness=0.75, selection probability=0.75 / ΣFitness×100%.

[0057] Randomly select the crossover point (for example, binary encoding), exchange part of the genes of the two parent chromosomes, for example: parent 1: [a=0.2(0011001100) g=0.95(1110111011)].

[0058] Parent 2: [a=0.3(0100110011) g=0.90(1101101100)].

[0059] After crossover: [a=0.2(0011001100) g=0.90(1101101100)].

[0060] Arithmetic crossover (for example, real number encoding): offspring parameter = l*parent 1 + (1-l)*parent 2, l is random in [0,1], for example: parent 1: a=0.2, g=0.95; parent 2: a=0.3, g=0.90; l=0.6.

[0061] Offspring: a=0.2*0.6+0.3*0.4=0.24, g=0.95*0.6+0.90*0.4=0.93.

[0062] Mutation operation, for example, binary encoding, inverting a bit of the chromosome with a probability of 5%, for example, the third bit of the binary string of a changes from 0 to 1, corresponding to a from 0.2 to 0.2+Δ (Δ is determined by the coding precision, the more the number of binary bits, the smaller the Δ, the higher the precision).

[0063] For example, add Gaussian noise to the real number coding, such as α = 0.2 mutation is 0.2 + 0.05 × N (0, 1), prevent the algorithm from falling into local optimum. Among them, N (0, 1) is the standard normal distribution with mean 0, variance 1, 0.05 is the "noise intensity" (adjustable); After mutation, α becomes 0.2 + a small random perturbation, which not only preserves the original parameter information but also introduces new changes.

[0064] In response to the termination of the iterative evolution, the second parameter combination with the highest fitness is taken as the optimal hyperparameter of Q-Learning. Specifically, the evolution process can include: the 0th generation (initial population): individual 1: α = 0.5, γ = 0.8 → temperature deviation ± 2℃, energy consumption 1.2t, slow convergence (200 iterations); Individual 2: α = 0.1, γ = 0.99 → temperature stable but high energy consumption (1.5t).

[0065] The 50th generation: through crossover and mutation, a combination of α = 0.25, γ = 0.95 is generated, the temperature deviation is reduced to ± 1℃, the energy consumption is 1.1t, and the fitness is improved by 30%.

[0066] The 100th generation: elite preservation enables the parameter cluster of α = 0.2-0.25, γ = 0.95-0.97 to continue optimization, and the energy consumption is reduced to 1.05t.

[0067] The 150th generation: mutation generates a combination of α = 0.2, γ = 0.95, the temperature deviation is ± 0.8℃, the energy consumption is 1.0t, and the production indicators are met.

[0068] The termination conditions include: the fitness improvement of 10 consecutive generations is less than 1%; the preset number of iterations (e.g. 200 generations) is reached; the parameter combination meets the industrial indicators (e.g. MAE <0.5℃, energy consumption ≤1.0t / ton cloth).

[0069] Preferably, the running state of the equipment is monitored, as well as the replacement record and running data of key components of the equipment; when the key components of the equipment are not replaced for a continuous preset time, and the steam pressure fluctuation, temperature control accuracy and unit energy consumption in the running data all meet the preset stable threshold, the parameter freezing mechanism is triggered; under the parameter freezing mechanism, the iterative optimization of the learning rate α and the discount factor γ of Q-Learning is suspended, and the current optimal parameter combination is fixedly used; in response to detecting the replacement of key components or the running data exceeding the stable threshold, the parameter freezing is released.

[0070] In some embodiments, in the temperature rise control of the intermittent dyeing machine, the hyperparameters (learning rate α, discount factor γ) of Q-Learning are optimized by genetic algorithm, and when the equipment operating condition is stable, continuous iteration may cause parameter drift due to random fluctuations in data, deviating from the optimal solution. The parameter freezing mechanism selectively suspends parameter updating by judging the stability of the equipment state, achieving the dual goals of saving computing resources and ensuring policy stability.

[0071] Specifically, a period of observation (e.g., 3 consecutive months) is set, and if no replacement of key components of the equipment (e.g., steam pipes, heat exchangers) occurs in this period, it is considered that the physical state is stable. Alternatively, the operation data of the equipment (e.g., steam pressure fluctuation range, temperature control accuracy) are collected, and for example, if the following conditions are met, it is determined that the working condition is stable: steam pressure fluctuation ≤ ±5% of the historical average; temperature control MAE ≤ 0.8°C and no significant upward trend; unit energy consumption fluctuation ≤ ±10% of the historical average. It can be understood that the specific parameter threshold can be set according to the actual situation, and the present embodiment is not limited.

[0072] When the time dimension (e.g., 3 months) and data dimension (e.g., pressure, temperature, and energy consumption stability) standards are met at the same time, the parameter freezing is automatically triggered; during the freezing period, the iterative optimization of the genetic algorithm for α and γ is suspended, and the current optimal parameters (e.g., α = 0.2, γ = 0.95) are fixed, Q-Learning continues to perform temperature control based on the frozen parameters, but the Q value table is still updated normally (only optimize the action strategy, do not change the learning rule).

[0073] When the key components of the equipment are replaced (e.g., steam pipe repair), or the data dimension index breaks through the threshold (e.g., temperature MAE > 1.2°C for 3 consecutive days), the freezing is released; or in specific cases, such as process adjustment, the parameters are manually optimized to trigger the unfreezing.

[0074] Preferably, a sliding window is used to filter the dyeing data of the last N batches to identify abnormal data; in response to the proportion of abnormal data in the sliding window being less than a preset threshold, the parameter freezing is not released.

[0075] In some embodiments, accidental sensor failure or material fluctuation (e.g., abnormal dye concentration in a batch) may occur in the production process, and if the abnormal data is directly used to release the freezing and update the parameters, it will cause the strategy to deteriorate. The sliding window filtering can identify and filter sudden abnormalities.

[0076] Specifically, the dyeing data of the last N batches (e.g., N = 50) are selected as the analysis window and updated in real time; the fitness value (e.g., temperature MAE, energy consumption, etc.) of each batch of data is calculated; the extreme value exceeding 3 times the standard deviation is identified as abnormal data; if the proportion of abnormal data is <10% (e.g., ≤5 times in 50 batches), it is determined that it is accidental interference, and the parameter update is not triggered; if the proportion is ≥10%, it is considered that the equipment state changes, and the freezing is released. Only when necessary (high proportion of abnormal data or significant device aging) is the parameter adjustment triggered, accidental abnormal data is filtered, parameter misupdate caused by single steam pressure fluctuation or sensor error is avoided, and the calculation resources are optimized. It can be understood that the specific parameter threshold can be set according to the actual situation, and the present embodiment is not limited.

[0077] Figure 2 A printing process parameter optimization control system 200 is shown. The system embodiment corresponds to the method embodiment shown, and the system can be specifically applied to various electronic devices. Figure 1

[0078] As shown in Figure 2 The printing process parameter optimization control system 200 provided by the embodiments of the present application includes: a first acquisition module 201 configured to acquire first parameters of an intermittent dyeing machine, the first parameters including temperature sequence data in a preset time, steam flow and pressure parameters, and dyeing parameters, the dyeing parameters including fabric type, dye type, dye solution pH value, and target dyeing temperature.

[0079] A first output module 202 is configured to perform time sequence analysis on the first parameters based on a multi-layer LSTM unit, identify the correlation mode of temperature abnormal fluctuation and equipment parameter change, and output a hidden vector representation, the hidden vector representation including current temperature state, temperature rising trend, and equipment running state.

[0080] A second output module 203 is configured to input the hidden vector representation as a state, combine a discretized action space and a preset reward function, calculate the Q value of each action through a Q-Learning algorithm, and select an optimal temperature rising control action, the control action including temperature rising rate adjustment and steam valve opening degree adjustment.

[0081] A control module 204 is configured to control the intermittent dyeing machine to perform printing based on the control action.

[0082] Based on the same inventive concept, the embodiments of the present application also provide an electronic device, the method corresponding to the electronic device can be the method in the foregoing embodiments, and the problem solving principle of the electronic device is similar to that of the method. The electronic device provided by the embodiments of the present application includes: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method and / or technical solution of the foregoing embodiments of the present application.

[0083] ​The electronic device can be a user device, or a device integrated by a user device and a network device through a network, or also can be an application program running on the above device, the user device includes but is not limited to computers, mobile phones, tablet computers, smart watches, wristbands and various terminal devices, and the network device includes but is not limited to network hosts, single network servers, multiple network server sets or cloud computing-based computer sets, which can be used to realize part of the processing function when setting an alarm. Here, the cloud is composed of a large number of hosts or network servers based on cloud computing, wherein cloud computing is a kind of distributed computing, which is composed of a virtual computer formed by a group of loosely coupled computer sets.

[0084] Figure 3 The structure of a device suitable for implementing the method and / or technical scheme in the embodiments of the present application is shown, the device 300 includes a central processing unit (CPU, Central Processing Unit) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM, Read Only Memory) 302 or the program loaded from the storage part 308 to the random access memory (RAM, Random Access Memory) 303. In the RAM 303, various programs and data required for system operation are also stored. The CPU 301, the ROM 302 and the RAM 303 are connected to each other through a bus 304. An input / output (I / O, Input / Output) interface 305 is also connected to the bus 304.

[0085] The following components are connected to the I / O interface 305: an input part 306 including a keyboard, a mouse, a touch screen, a microphone, an infrared sensor, etc.; an output part 307 including a cathode ray tube (CRT, Cathode Ray Tube), a liquid crystal display (LCD, Liquid Crystal Display), an LED display, an OLED display, etc., and a speaker, etc.; a storage part 308 including one or more computer readable media such as a hard disk, an optical disk, a magnetic disk, a semiconductor memory, etc.; and a communication part 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication part 309 performs communication processing via a network such as the Internet.

[0086] In particular, the methods and / or embodiments of the present application can be implemented as a computer program. For example, the embodiments disclosed herein include a computer program product comprising a computer program that is carried on a computer readable medium, the computer program comprising program code for executing the methods illustrated in the flowcharts. When the computer program is executed by a central processing unit (CPU) 301, the above-mentioned functions defined in the methods of the present application are executed.

[0087] Another embodiment of the present application also provides a computer readable storage medium having stored thereon computer program instructions, which can be executed by a processor to implement the method and / or technical solutions of any one or more embodiments of the present application.

[0088] In particular, the embodiments can employ any combination of one or more computer readable media. The computer readable media can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this document, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0089] Finally, it should be noted that the above embodiments are merely intended to illustrate the technical solutions of the present application, and not to limit the same; even though the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some of the technical features; and such modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

[0090] In addition, it is clear that the word "comprise" does not exclude other elements or steps, and the singular does not exclude the plural. Multiple elements or devices recited in a device claim can also be implemented by one element or device performing the functions of the multiple elements or devices. The words "first", "second", and the like do not imply any particular order, but are used to name the elements to distinguish them from each other.

Claims

1. A method for optimizing control of printing process parameters, characterized in that, The method comprises the following steps: obtaining first parameters of an intermittent dyeing machine, the first parameters including temperature sequence data, steam flow and pressure parameters and dyeing parameters in a preset time, the dyeing parameters including fabric type, dye type, dyeing liquid pH value and target dyeing temperature; performing time sequence analysis on the first parameters based on a multi-layer LSTM unit, identifying the correlation mode of temperature abnormal fluctuation and equipment parameter change, and outputting a hidden vector representation, the hidden vector representation including current temperature state, temperature rising trend and equipment running state; taking the hidden vector representation as state input, combining a discretized action space and a preset reward function, calculating the Q value of each action through a Q-Learning algorithm, selecting an optimal temperature rising control action, and controlling the intermittent dyeing machine to perform printing and dyeing based on the control action.

2. The method for optimizing and controlling printing and dyeing process parameters according to claim 1, characterized in that: The multi-layer LSTM unit comprises a first layer LSTM unit and a second layer LSTM unit, the first layer LSTM unit receives the first parameters, processes the first parameters through an input gate, a forgetting gate and an output gate, and outputs a first hidden state containing short-term features; the second layer LSTM unit receives the first hidden state and the first parameters, and outputs a second hidden state containing high-level state abstraction.

3. The method for optimizing and controlling printing and dyeing process parameters according to claim 1, characterized in that: The hyperparameters of the Q-Learning are optimized by using a genetic algorithm, the hyperparameters including a learning rate α and a discount factor γ, and the genetic algorithm comprises the following steps: encoding the learning rate α and the discount factor γ of the Q-Learning into a chromosome; initializing a population of second parameter combinations, the second parameter combinations covering the reasonable value range of α and γ; constructing a fitness function, the fitness function being a weighted function used to evaluate each second parameter combination in the population; iteratively evolving the population through roulette selection, crossover and mutation operations, wherein the crossover operation includes single-point crossover or arithmetic crossover, and the mutation operation is bit inversion or adding Gaussian noise to the chromosome; and in response to termination of the iterative evolution, taking the second parameter combination with the highest fitness as the optimal hyperparameters of the Q-Learning.

4. The method for optimizing control of printing and dyeing process parameters according to claim 3, characterized in that, Monitoring the equipment running state, key component replacement record and running data of the equipment; when the equipment has not replaced a key component for a preset time, and the steam pressure fluctuation, temperature control accuracy and unit energy consumption in the running data all meet preset stable threshold values, triggering a parameter freezing mechanism; under the parameter freezing mechanism, suspending the iterative optimization of the learning rate α and the discount factor γ of the Q-Learning, and fixedly using the current optimal parameter combination; in response to detecting key component replacement or running data exceeding the stable threshold values, releasing the parameter freezing.

5. The method for optimizing control of printing and dyeing process parameters according to claim 4, characterized in that, Further comprising: filtering the dyeing data of the last N batches by using a sliding window, and identifying abnormal data therein; in response to the proportion of the abnormal data in the sliding window being less than a preset threshold value, not releasing the parameter freezing.

6. A printing process parameter optimization control system characterized by, The method comprises the following steps: The first acquisition module is configured to acquire first parameters of the intermittent dyeing machine, the first parameters including temperature sequence data, steam flow and pressure parameters and dyeing parameters in a preset time before the current time, the dyeing parameters including fabric type, dye type, dyeing solution pH value and target dyeing temperature; The first output module is configured to perform time sequence analysis on the first parameters based on the multi-layer LSTM unit, identify an association mode of temperature abnormal fluctuation and equipment parameter change, and output a hidden vector representation, the hidden vector representation including a current temperature state, a temperature rising trend and an equipment running state; The second output module is configured to take the hidden vector representation as state input, combine a discretized action space and a preset reward function, calculate Q values of each action through a Q-Learning algorithm, select an optimal temperature rising control action, and output the control action, the control action including temperature rising rate adjustment and steam valve opening degree adjustment.

7. A process parameter optimization control system for textile printing and dyeing as claimed in claim 6 wherein: The multi-layer LSTM unit includes a first layer LSTM unit and a second layer LSTM unit, the first layer LSTM unit receiving the first parameters, processing the first parameters through an input gate, a forgetting gate and an output gate, and outputting a first hidden state containing short-term features; and the second layer LSTM unit receiving the first hidden state and the first parameters, and outputting a second hidden state containing high-level state abstraction.

8. An electronic device, comprising: At least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.

9. A computer readable medium having stored thereon computer program instructions, characterized in that: The computer program instructions can be executed by the processor to implement the method of any one of claims 1-5.

10. A computer program product comprising a computer program, characterized in that: The computer program is executed by the processor to implement the method of any one of claims 1-5. The computer program is executed by the processor to implement the method of any one of claims 1-5.

Images