Drug production process scheduling optimization system and method based on artificial intelligence

Abstract

The invention discloses a drug production process scheduling optimization system and method based on artificial intelligence, and relates to the technical field of artificial intelligence, and the system comprises a data management module, a data analysis module and a scheduling optimization module. Through cooperative work of all the modules, key data in the production process can be monitored and obtained in real time, key factors influencing the drug production quality can be accurately recognized, the production process is deeply mined and analyzed through a factor analysis algorithm, a scheduling scheme is optimized, the production efficiency is effectively improved, and the production cost is reduced. By dynamically adjusting key operation, resource waste is avoided, the production cycle is shortened, in addition, a scheduling optimization module continuously optimizes a production scheduling strategy, the high efficiency and stability of the medicine production process are ensured, the production cost is reduced, the medicine quality is improved, and the high requirement of the market for medicine is met; and powerful technical support is provided for sustainable development of enterprises.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based drug production process scheduling optimization system and method. Background Art

[0002] Artificial intelligence (AI) is an emerging technical science that studies and develops theories, methods, technologies and application systems to simulate, extend and expand human intelligence. With the continuous advancement of artificial intelligence technology, AI has gradually penetrated into all walks of life and has become a core tool for improving efficiency and optimizing resource allocation. In the field of pharmaceutical production, especially in the scheduling optimization of pharmaceutical production processes, the application of AI is of great significance. The pharmaceutical production process involves many links, such as raw material procurement, production line scheduling, quality control, etc. These links have strict requirements for the optimization of time, resources and costs. AI can help monitor the production process in real time, analyze data and predict potential problems to ensure a smooth and efficient production process.

[0003] In the existing technology, it is not easy to accurately identify the key factors affecting the quality of drugs, resulting in lax quality control in the production process, fluctuations in product quality, and affecting the stability and compliance of the final product. In addition, when the production environment changes, the traditional scheduling system lacks flexibility and cannot respond to changes quickly, resulting in delays in production plans and inability to adjust production strategies in time to cope with emergencies. The lack of real-time dynamic adjustment capabilities leads to irrational resource allocation in the production process, long production cycles, and low efficiency.

[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention proposes an artificial intelligence-based drug production process scheduling optimization system and method, which solves the problems mentioned in the above background technology, such as the inconvenience in accurately identifying the key factors affecting the quality of drugs, resulting in lax quality control in the production process, fluctuations in product quality, and affecting the stability and compliance of the final product. In addition, when the production environment changes, the traditional scheduling system lacks flexibility and cannot respond quickly to changes, resulting in delays in production plans, inability to adjust production strategies in time to cope with emergencies, and lacks the ability to make real-time dynamic adjustments, resulting in unreasonable resource allocation in the production process, long production cycles, and low efficiency.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: According to one aspect of the present invention, there is provided a drug production process scheduling optimization system based on artificial intelligence, comprising: Data management module, used to obtain real-time production data during the drug production process and extract key production feature data during the production process; The data analysis module is used to analyze the key production characteristic data obtained by using the factor analysis algorithm to identify the key factors affecting the quality of drug production; The scheduling optimization module is used to establish a scheduling optimization model based on key factors, use the scheduling optimization model to dynamically adjust key operations in the production process, and optimize the production scheduling strategy.

[0007] Furthermore, the data management module includes: The data acquisition module is used to acquire real-time production data in the drug production process through sensors, and to perform denoising, filtering and smoothing preprocessing on duplicate data, missing values ​​and abnormal values ​​of the real-time production data to obtain processed real-time production data; The feature extraction module uses the Fourier transform method to extract features from the processed real-time production data to obtain key production feature data in the production process.

[0008] Furthermore, the key production characteristic data were analyzed using factor analysis algorithms to identify the key factors affecting drug production quality, including: Define the parameters of the factor analysis algorithm and the quality assessment model; randomly generate several groups of initial production feature combinations; Use historical production data to train the quality assessment model, calculate the quality impact coefficient of each feature combination, record the optimal production feature combination according to different production parameter configurations, and evaluate its impact on drug production quality; Based on changes in the production environment, the sensitivity parameters of the factor analysis algorithm are updated in real time according to the dynamic adjustment function, and the parameters in the factor analysis model are dynamically adjusted by sensing environmental changes during the production process; The evaluation algorithm is used to evaluate the contribution of each combination of production features to the quality of drug production. In each round of optimization, the production environment changes and the quality evaluation model are combined to select the production feature combination with the greatest contribution. The selected feature combinations with the greatest contribution are input into the quality assessment model to identify the key factors affecting the quality of drug production.

[0009] Furthermore, the evaluation algorithm is used to evaluate the contribution of each group of production feature combinations to the quality of drug production. In each round of optimization, the production feature combinations with the greatest contribution are screened out by combining the production environment changes and the quality evaluation model, including: Initialize the evaluation algorithm, set the maximum number of iterations of the evaluation algorithm, and set the search space; In the set search space, a set of drug production feature combinations is randomly generated, and the fitness value of each feature combination for drug quality is calculated based on the current production environment and quality assessment model; according to the size of the fitness value, the initial optimal production feature combination is sorted and determined; In each round of optimization, the quality evaluation value of each production feature combination is calculated, and a random number is generated. According to the preset switch probability, whether to perform a global search or a local search is selected. If the random number is less than the preset switch probability, a global search is performed to update the production feature combination; otherwise, a local search is performed to optimize some attributes of the production feature combination. After the production feature combination is updated, its corresponding quality fitness value is evaluated again, and all production feature combinations are sorted according to the fitness value to update the current optimal production feature combination; Optimize the current optimal production feature combination dimension by dimension, adjust the optimal production feature combination using the Cauchy mutation method, and perform boundary control on the result after mutation; Using the greedy algorithm, the fitness values ​​of the production feature combinations before and after mutation are compared, and the best fitness value after mutation is retained; Determine whether the evaluation algorithm has reached the maximum number of iterations. If so, output the optimal production feature combination and the best fitness value as the production feature combination with the greatest contribution.

[0010] Furthermore, the formula for adjusting the optimal production feature combination using the Cauchy mutation method is: ; In the formula, represents the first j The value of a combination of production characteristics; Indicates the current optimal j The value of a combination of production characteristics; d represents the number of production feature combinations; Cauchy (0, 1) represents a random number generated from a Cauchy distribution; rand Represents a random number.

[0011] Furthermore, using the greedy algorithm, the fitness values ​​of the production feature combinations before and after mutation are compared, and the best fitness values ​​after mutation are retained, including: Randomly select an initial production feature combination as the first set of candidate solutions; Based on the current production feature combination, randomly select new features for combination, calculate the fitness value of the combination, and select the feature combination with the highest fitness value as the new production feature combination; Verify whether the new production feature combination meets the preset constraints. If not, cancel the new production feature combination, select the suboptimal feature combination and re-verify the constraints until a production feature combination that meets the preset constraints is obtained; Continuously optimize the feature combination until the combination reaches the local optimum, and determine the current production feature combination as the best production feature combination; On the basis of the current best combination of production features, random factors are introduced to randomly mutate a combination of production features and find the next potential best combination of production features; Compare the fitness value of the mutated production feature combination with the current best combination. If the fitness value after mutation is greater than the fitness value before mutation, the mutated production feature combination is used as the new best production feature combination. Otherwise, the current best production feature combination is kept unchanged and the current greedy search process ends. When this round of greedy search ends, a new initial production feature combination is selected and the greedy algorithm optimization is performed again. Multiple rounds of search are repeated, and finally the optimal solution is selected from all the best production feature combinations found, and the best fitness value after mutation is retained.

[0012] Furthermore, the scheduling optimization module includes: The model building module is used to obtain the data set of key factors, divide the data set into training set and test set, and establish a scheduling optimization model; The scheduling optimization module is used to optimize the established scheduling optimization model using the scheduling algorithm, and to dynamically adjust the key operations in the production process using the optimized scheduling optimization model to optimize the production scheduling strategy.

[0013] Furthermore, the scheduling algorithm is used to optimize the established scheduling optimization model, and the optimized scheduling optimization model is used to dynamically adjust the key operations in the production process. The optimized production scheduling strategy includes: Randomly generate several groups of initial production scheduling plans and set the maximum number of iterations of the scheduling algorithm; According to the production efficiency, the fitness value of each scheduling scheme is calculated, the number of new scheduling schemes generated by each scheduling scheme is calculated according to the quantity formula, and the standard deviation of each new scheduling scheme is calculated according to the standard deviation formula; Determine whether the number of current scheduling schemes reaches the threshold. If not, increase the number of iterations of the scheduling algorithm and continue to optimize. If it reaches the threshold, sort the current scheduling schemes from high to low according to the fitness value, select the best scheduling scheme as the best scheduling scheme for the current iteration, and select several scheduling schemes before reaching the threshold as the next generation candidate schemes. If the scheduling algorithm reaches the maximum number of iterations, the scheduling optimization process of the scheduling algorithm ends, otherwise, the scheduling optimization continues; Perform chaos search around the best scheduling solution, generate several new scheduling solutions, and calculate the fitness value of the new scheduling solution; if a new scheduling solution with the best fitness value is found, replace the original best scheduling solution as the new best scheduling solution, otherwise, continue the scheduling optimization process; The scheduling scheme with the best fitness value is output as the optimal scheduling strategy.

[0014] Furthermore, the formula for calculating the fitness value of each scheduling scheme is: ; The quantity formula is: ; The standard deviation formula is: ; In the formula, F Represents the fitness value of the scheduling scheme; Q Indicates production efficiency; e represents the penalty term; Number _ Seeds [ i ] indicates the i The number of new scheduling solutions generated by the scheduling solution; F i Indicates i The fitness value of each scheduling scheme; F g represents the optimal fitness value; F w Indicates the worst fitness value; S max Indicates the maximum number of new scheduling solutions generated; S min Indicates the minimum number of new scheduling solutions generated; s iter Indicates that it is iter The standard deviation of the iterations; iter max Indicates the maximum number of iterations; s i represents the starting standard deviation; s f represents the final standard deviation; n represents the nonlinear harmonic parameter.

[0015] According to another aspect of the present invention, there is also provided a method for optimizing drug production process scheduling based on artificial intelligence, the method comprising the following steps: S1. Obtain real-time production data during the drug production process and extract key production feature data during the production process; S2. Analyze the key production characteristic data obtained using factor analysis algorithms to identify key factors affecting drug production quality; S3. Based on key factors, a scheduling optimization model is established, and the scheduling optimization model is used to dynamically adjust key operations in the production process and optimize the production scheduling strategy.

[0016] The beneficial effects of the present invention are: 1. Through the collaborative work between modules, the present invention can monitor and obtain key data in the production process in real time, accurately identify key factors affecting the quality of drug production, and use factor analysis algorithms to deeply explore and analyze the production process, optimize the scheduling plan, and effectively improve production efficiency. By dynamically adjusting key operations, the system can flexibly respond to various changes in production, avoid waste of resources and shorten the production cycle. In addition, the scheduling optimization module continuously optimizes the production scheduling strategy, ensures the efficiency and stability of the drug production process, reduces production costs, improves drug quality, meets the high market demand for drugs, and provides strong technical support for the sustainable development of enterprises.

[0017] 2. The present invention effectively improves the optimization capability of the drug production process by combining factor analysis algorithms and evaluation algorithms. The system acquires production feature data in real time and dynamically adjusts model parameters to accurately identify and optimize key factors affecting drug production quality. It uses Cauchy mutation and greedy algorithms and other technologies to perform global and local searches for production feature combinations, continuously improve the quality adaptability of the production process, and optimize production scheduling strategies, thereby ensuring the efficiency, stability and quality controllability of the production process, reducing human intervention, lowering production costs, and improving production efficiency.

[0018] 3. The present invention improves the intelligence of the production process by constructing a scheduling optimization model and optimizing the production scheduling plan in combination with a scheduling algorithm. The system can dynamically adjust key production operations, optimize production scheduling strategies, ensure the rational allocation of production resources, and improve production efficiency. It uses methods such as fitness calculation, chaos search, and standard deviation adjustment to continuously optimize the scheduling plan and reduce resource waste and production delays caused by unreasonable scheduling, thereby ensuring the efficiency and stability of the drug production process and improving the overall intelligence level of scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1It is a principle block diagram of a drug production process scheduling optimization system based on artificial intelligence according to an embodiment of the present invention; Figure 2 It is a flowchart of a drug production process scheduling optimization method based on artificial intelligence according to an embodiment of the present invention.

[0021] In the figure: 1. Data management module; 2. Data analysis module; 3. Scheduling optimization module. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0023] In the description of the present invention, unless otherwise specified, the meaning of "plurality" is two or more. In addition, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0024] According to an embodiment of the present invention, a drug production process scheduling optimization system and method based on artificial intelligence are provided.

[0025] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, the drug production process scheduling optimization system based on artificial intelligence according to an embodiment of the present invention includes: Data management module 1, used to obtain real-time production data during the drug production process and extract key production feature data during the production process; Specifically, sensors can be installed on production equipment to monitor the equipment's operating status, temperature, pressure, speed and other parameters in real time. These sensor data can be transmitted in real time via the Industrial Internet of Things (IIoT) or SCADA (Supervisory Control and Data Acquisition System).

[0026] Specifically, real-time production data includes: 1) Equipment status data: whether the equipment is running, operating status, production speed, load, temperature, pressure, etc.

[0027] 2) Production progress data: including current production steps, completion progress of production tasks, expected completion time, etc.

[0028] 3) Production data: real-time production volume, number of production batches, and number of drugs produced per unit time.

[0029] 4) Resource usage data: raw material consumption, energy usage (such as electricity, gas, etc.), labor time utilization, etc.

[0030] 5) Equipment failure and maintenance data: equipment failure report, maintenance time, downtime record, failure type, etc.

[0031] 6) Quality control data: online quality test results, such as drug content, appearance inspection, solubility, pH value, etc.

[0032] 7) Production environment data: such as temperature, humidity, air quality, workshop cleanliness and other environmental data.

[0033] Specifically, key production characteristic data include: 1) Equipment performance data: including equipment operating speed, load, efficiency, maintenance frequency, etc., which directly affect production continuity and production capacity.

[0034] 2) Quality data: quality indicators of drugs, such as the quality of raw materials, the quality of intermediates in the production process, and the quality inspection data of the final product (such as drug content, particle size distribution, dissolution curve, etc.).

[0035] 3) Production environment parameters: such as temperature, humidity, air cleanliness, etc. These environmental parameters have an important impact on the stability and quality of drugs, especially the stability of certain drugs in the pharmaceutical process.

[0036] 4) Material flow data: including real-time data on the supply, storage, and use of materials, such as raw material inventory, material consumption, transportation conditions, etc.

[0037] 5) Production time data: Time records for each production link, such as equipment preparation time, production cycle, waiting time, etc., which helps to evaluate production efficiency.

[0038] 6) Personnel operation data: operator work efficiency, working hours, workload, etc. These data can help analyze the relationship between personnel efficiency and production quality.

[0039] 7) Batches and batch differences: Data on production conditions, quality fluctuations, raw material sources, etc. of different batches of drugs help analyze production differences between batches.

[0040] Data analysis module 2, used to analyze the obtained key production characteristic data using factor analysis algorithm to identify key factors affecting drug production quality; Specifically, the key factors affecting the quality of drug production include: 1) Raw material quality: Source of raw materials: The quality of raw materials directly affects the final quality of the drug, especially whether the standards are met during the procurement, storage and use of raw materials.

[0041] Raw material purity and composition: Raw materials used in drug production must have a certain purity and stable composition ratio. Any impurities or variations may affect the efficacy and safety of the drug.

[0042] The physical and chemical properties of raw materials: such as particle size, solubility, fluidity, etc. These factors have a great influence on the production process of drugs and the performance of the final product.

[0043] 2) Production environment: Temperature: Temperature control is very important in the pharmaceutical production process. The chemical reaction or stability of some drugs is very sensitive to temperature changes. Too high or too low a temperature may result in substandard drug quality.

[0044] Humidity: Humidity has an impact on the stability of many drug products, especially during the storage and processing of powdered, granular and solid drug products.

[0045] Air quality: Dust or harmful substances may be released during the pharmaceutical process, so air cleanliness directly affects the purity and quality of the product.

[0046] Cleanliness and microbial control: In the pharmaceutical production environment, especially in the aseptic production process, a clean environment is essential. Microorganisms or contaminants in the air may cause product contamination.

[0047] 3) Production equipment and process parameters: Equipment performance and status: The operating status, maintenance cycle, and failure rate of production equipment will affect production efficiency and product quality. If the equipment fails or operates unstably, it may cause the production line to stagnate or the product to fail.

[0048] Production process: Every step in the drug production process must be strictly controlled, such as mixing time, reaction temperature, pressure, stirring speed, drying time, etc. Slight changes in process parameters may cause fluctuations in drug quality.

[0049] Automation control and system stability: The reliability and stability of the automation control system in the production process are crucial to the efficiency and quality control of production. Any failure or deviation of the automation system will affect the entire production process.

[0050] 4) Production planning and scheduling: Reasonableness of production plan: unreasonable production scheduling may lead to waste of production resources or production line stoppage. Production plan should be optimized based on market demand, raw material supply, production capacity and other factors.

[0051] Batch-to-batch variation: Each production batch may have slight differences in the process and raw materials, which may affect the quality of the final product. Reasonable batch management and quality control strategies can reduce this fluctuation.

[0052] 5) Operator and personnel management: Operator skills and training: The operator's experience and skills directly affect the quality of operations during production. If the operator is not adequately trained or inexperienced, it may lead to irregular operations, which in turn affects product quality.

[0053] Workload and fatigue of personnel: Operators’ working hours and load will also affect the quality of operation. When they are overly tired or stressed, they are prone to operational errors, resulting in unstable production.

[0054] 6) Quality Control and Testing: Online quality inspection: During the production process, timely quality inspection can identify problems and make adjustments. Quality inspection data such as drug content, solubility, particle size, etc. can reflect whether the production process meets quality standards.

[0055] Accuracy and reliability of quality inspection: The accuracy of inspection equipment and methods is crucial in the quality control process. If the inspection equipment is not accurate or the inspection method is unscientific, the quality of the drug may not be correctly assessed.

[0056] 7) Continuity and stability of production process: Continuity of production process: Every step of drug production must be coordinated and consistent. Any break or instability in the process may affect the quality of the final product. Ensuring a smooth transition of each step in the production process is crucial to quality control.

[0057] Variability in the production process: Variability in the production process, such as fluctuations in temperature, pressure, stirring speed, etc., will directly affect the quality of the drug. Therefore, controlling the stability of the production process is key to ensure consistency and repeatability.

[0058] 8) External environment and logistics: Transportation conditions: The transportation and storage conditions of drugs may also affect the quality, especially for drugs that require cold chain transportation. If the temperature, humidity and other conditions are not suitable during transportation, the quality of the drugs may deteriorate.

[0059] Supply chain stability: Supply chain stability of raw materials is critical in drug production. If the supply of raw materials is unstable or the quality does not meet the standards, it will lead to production interruptions or quality problems.

[0060] The scheduling optimization module 3 is used to establish a scheduling optimization model based on key factors, use the scheduling optimization model to dynamically adjust key operations in the production process, and optimize the production scheduling strategy.

[0061] Specifically, the key operations include: 1) Material preparation and transportation: Ensure that raw materials are prepared on time, in quantity and according to standards, and rationally plan the transportation routes and timing of materials to reduce waiting time and waste.

[0062] 2) Equipment startup and adjustment: When starting production equipment, check the equipment status and adjust equipment parameters (such as temperature, pressure, speed, etc.) to ensure stable operation of the equipment and avoid equipment failure and unplanned downtime.

[0063] 3) Production process operation: including mixing, reaction, filtration, drying, filling, packaging and other process operations, ensuring that each production link is carried out in accordance with the standard operating procedure (SOP) to ensure quality consistency.

[0064] 4) Quality control and testing: Real-time quality testing, such as the determination of drug content, particle size, solubility and other quality indicators. Adjust production parameters based on test results to ensure compliance with drug quality standards.

[0065] 5) Equipment maintenance and overhaul: timely inspection and maintenance of equipment to avoid production interruptions caused by equipment failure. Maintenance operations involve cleaning, overhaul, calibration, etc.

[0066] 6) Personnel operation management: Personnel work arrangements, operation skill training, fatigue management, etc. are also key factors affecting production efficiency and quality. Reasonable arrangements for personnel rotation and work tasks can improve operation efficiency and reduce human errors.

[0067] 7) Environmental monitoring and regulation: Monitor the temperature, humidity, cleanliness, etc. in the production environment. Especially in the aseptic production process, environmental factors are crucial to the quality of drugs.

[0068] 8) Coordination and switching of production processes: In a multi-product production line, coordinate the production processes of different products, switch the production tasks of products, optimize the production line switching time, and avoid excessive production pauses.

[0069] Specifically, the production scheduling strategy includes: 1) Task priority scheduling: Set the priority of different tasks according to factors such as the urgency of drug production, delivery date, order quantity, etc. Prioritize high-priority orders or urgent tasks to ensure the timeliness of production progress.

[0070] 2) Optimal resource scheduling: Rationally schedule production resources, including raw materials, production equipment, personnel, etc., to ensure that resources in each link are optimally allocated and avoid idle or excessive concentration of resources.

[0071] 3) Production batch scheduling: Rationally arrange the production time of different production batches, minimize the switching time between batches, avoid unnecessary downtime and switching, and optimize the production flow line.

[0072] 4) Equipment load balancing scheduling: Reasonable scheduling is carried out according to the status, capacity and load of the production line equipment to avoid equipment overload or idleness and ensure maximum equipment efficiency.

[0073] 5) Production cycle optimization: By analyzing the time nodes in the production process, optimize the time allocation of each production link, minimize the pause, waiting and idle time of each link, and improve the overall efficiency of production.

[0074] 6) Dynamically adjust the scheduling strategy: During the production process, real-time monitoring of production data (such as equipment status, environmental conditions, quality control data, etc.) is performed, and the scheduling strategy is dynamically adjusted based on this data. For example, when a device fails, the load of other devices is automatically adjusted, or when the quality does not meet the standards, the process parameters and production process are adjusted.

[0075] 7) Multi-objective scheduling optimization: In production scheduling, it is often necessary to balance multiple objectives, such as production efficiency, production cost, product quality and delivery time. By setting a multi-objective optimization model, comprehensive consideration of various objectives can be made to develop the optimal scheduling plan.

[0076] 8) Flexible production plan adjustment: Flexibly adjust production plans according to changes in market demand, raw material supply, production capacity and other factors. For example, when faced with sudden orders or production stoppages, the production plan can be quickly replanned and resources adjusted.

[0077] 9) Inventory management and scheduling: Scheduling is carried out according to the inventory status of raw materials and finished products during the production process to ensure the smooth operation of the production line and avoid production risks caused by excessive or low inventory.

[0078] In this optional embodiment, the data management module 1 includes: The data acquisition module is used to acquire real-time production data in the drug production process through sensors, and to perform denoising, filtering and smoothing preprocessing on duplicate data, missing values ​​and abnormal values ​​of the real-time production data to obtain processed real-time production data; The feature extraction module uses the Fourier transform method to extract features from the processed real-time production data to obtain key production feature data in the production process.

[0079] Specifically, first, collect real-time production data and preprocess it to remove noise, outliers, etc. Then, divide the data into small segments and apply a window function to reduce frequency domain leakage. Next, apply Fourier transform (usually using fast Fourier transform FFT) to convert the time domain data into frequency domain signals to obtain the amplitude and phase information of the frequency components. By analyzing the frequency domain signal, extract the main frequency components and spectral features, such as spectral density, bandwidth, etc., to identify the periodic and non-periodic change patterns in the production process. After that, select the key features related to production quality from the frequency domain features.

[0080] In this optional embodiment, the key production characteristic data obtained are analyzed using a factor analysis algorithm to identify key factors affecting drug production quality, including: Define the parameters of the factor analysis algorithm and the quality assessment model; randomly generate several groups of initial production feature combinations; Use historical production data to train a quality assessment model (i.e., extreme learning machine model), calculate the quality impact coefficient of each feature combination, record the optimal production feature combination based on different production parameter configurations, and evaluate its impact on drug production quality; Based on changes in the production environment, the sensitivity parameters of the factor analysis algorithm are updated in real time according to the dynamic adjustment function, and the parameters in the factor analysis model are dynamically adjusted by sensing environmental changes during the production process; The evaluation algorithm is used to evaluate the contribution of each combination of production features to the quality of drug production. In each round of optimization, the production environment changes and the quality evaluation model are combined to select the production feature combination with the greatest contribution. The selected feature combinations with the greatest contribution are input into the quality assessment model to identify the key factors affecting the quality of drug production.

[0081] Specifically, the factor analysis algorithm is a crow search algorithm. The crow algorithm is a natural heuristic optimization algorithm that simulates the foraging behavior of crows and belongs to a swarm intelligence optimization algorithm. Its basic idea is to use the distribution behavior of crows in the process of finding food to find the optimal solution to the problem. The algorithm randomly generates multiple candidate solutions and approaches the global optimal solution through information exchange and adjustment. In the present invention, the crow algorithm is used to analyze and select production feature combinations, that is, to identify the key factors affecting the quality of drug production by optimizing the production feature combination. The extreme learning machine (ELM) is a learning algorithm based on a single hidden layer feedforward neural network. It randomly generates input weights and biases and trains the network by minimizing the output error. Unlike traditional neural network algorithms, ELM does not require a back propagation algorithm to optimize weights, but obtains all parameters through one calculation. This gives ELM a significant advantage in training speed and performance. In the present invention, the extreme learning machine model is used to construct a quality assessment model, that is, to evaluate the contribution of each group of production feature combinations to the quality of drug production.

[0082] In order to facilitate understanding of the above technical solution of the present invention, the present invention will be described in detail below on how to analyze the key production characteristic data obtained by using a factor analysis algorithm to identify key factors affecting the quality of drug production.

[0083] Step 1: Define the parameters of the factor analysis algorithm and quality assessment model: First, the parameters of the factor analysis algorithm need to be set, which help the model identify factors related to drug manufacturing quality.

[0084] 1) Assume the following parameters are selected: Number of factors: Select the number of factors to be analyzed as 3 (such as temperature, humidity, and equipment load).

[0085] Sensitivity parameter: Set the sensitivity parameter to 0.8 to adjust the impact of environmental changes on model updates.

[0086] 2) Parameters of the quality assessment model: Quality score function: The quality score is defined as the proportion of qualified products in the production process, assuming that the quality score ranges from 0 to 100.

[0087] Influence coefficient: Each production feature combination has an influence coefficient, and the initial value is set to 0.

[0088] Step 2: Randomly generate initial production feature combinations: 1) Based on the key characteristics of the drug production process, five groups of initial production feature combinations are randomly generated. Assume that these features include: Temperature (T), unit: °C.

[0089] Humidity (H), unit: %.

[0090] Equipment load (L), unit: %.

[0091] 2) Initial feature combination (assuming the following data is generated): Combination 1: T=30°C, H=60%, L=75%.

[0092] Combination 2: T=32°C, H=62%, L=78%.

[0093] Combination 3: T=28°C, H=65%, L=80%.

[0094] Combination 4: T=35°C, H=58%, L=70%.

[0095] Combination 5: T=33°C, H=61%, L=72%.

[0096] Step 3: Use historical production data to train the quality assessment model: 1) Use historical production data to train the quality assessment model. Historical data includes data such as temperature, humidity, equipment load, etc. in previous production and their corresponding drug quality scores. Assume that through training, the following calculation formula is obtained to evaluate the impact of each set of feature combinations on quality: Quality scoring function: Q=80+(T-30)×0.5+(H-60)×0.3+(L-75)×0.2.

[0097] 2) For each set of feature combinations, substitute the quality score function for calculation: Combination 1: Q1=80+(3030)×0.5+(6060)×0.3+(7575)×0.2=80.

[0098] Combination 2: Q2=80+(3230)×0.5+(6260)×0.3+(7875)×0.2=81.3.

[0099] Combination 3: Q3=80+(2830)×0.5+(6560)×0.3+(8075)×0.2=80.6.

[0100] Combination 4: Q4=80+(3530)×0.5+(5860)×0.3+(7075)×0.2=79.4.

[0101] Combination 5: Q5=80+(3330)×0.5+(6160)×0.3+(7275)×0.2=80.1.

[0102] Step 4: Dynamically adjust parameters based on changes in the production environment: 1) During the production process, temperature, humidity and equipment load are affected by external environmental factors. Assume that the temperature changes from 30°C to 32°C, and use the dynamic adjustment function to update the sensitivity parameter. According to the dynamic adjustment function: Adjusted sensitivity: D new = S current × (1+ T change / T max )=0.8×(1+2 / 5)=0.96.

[0103] Step 5: Use the evaluation algorithm to select the production feature combination with the greatest contribution: 1) Continue to use the quality assessment model to calculate the contribution of each feature combination to the quality of drug production. According to the above quality score calculation method, assume that the feature combination with the greatest contribution is screened out. Assume that the combination with the best contribution is combination 2 (T=32°C, H=62%, L=78%), and its quality score is 81.3.

[0104] Step 6: Identify the key factors affecting drug production quality: Based on the above analysis, the key factors affecting the quality of drug production are obtained: 1) Temperature (T): affects the quality score. Changes in temperature have a significant impact on drug quality, especially when the temperature rises, the quality score will increase.

[0105] 2) Humidity (H): Changes in humidity also have a certain impact on quality, although the magnitude of the change is not as significant as temperature.

[0106] 3) Equipment load (L): Equipment load has an impact on quality within a certain range. Too high or too low a load will affect production efficiency and quality.

[0107] 4) Finally, the key factors identified were: temperature, humidity and equipment load.

[0108] In this optional embodiment, an evaluation algorithm is used to evaluate the contribution of each group of production feature combinations to the drug production quality. In each round of optimization, the production feature combinations with the greatest contribution are screened out in combination with production environment changes and quality evaluation models, including: Initialize the evaluation algorithm, set the maximum number of iterations of the evaluation algorithm, and set the search space; In the set search space, a set of drug production feature combinations is randomly generated, and the fitness value of each feature combination for drug quality is calculated based on the current production environment and quality assessment model; according to the size of the fitness value, the initial optimal production feature combination is sorted and determined; In each round of optimization, the quality evaluation value of each production feature combination is calculated, and a random number is generated. According to the preset switch probability, whether to perform a global search or a local search is selected. If the random number is less than the preset switch probability, a global search is performed to update the production feature combination; otherwise, a local search is performed to optimize some attributes of the production feature combination. After the production feature combination is updated, its corresponding quality fitness value is evaluated again, and all production feature combinations are sorted according to the fitness value to update the current optimal production feature combination; Optimize the current optimal production feature combination dimension by dimension, adjust the optimal production feature combination using the Cauchy mutation method, and perform boundary control on the result after mutation; Using the greedy algorithm, the fitness values ​​of the production feature combinations before and after mutation are compared, and the best fitness value after mutation is retained; Determine whether the evaluation algorithm has reached the maximum number of iterations. If so, output the optimal production feature combination and the best fitness value as the production feature combination with the greatest contribution.

[0109] Specifically, the evaluation algorithm is an improved butterfly optimization algorithm. The core idea of ​​the basic butterfly optimization algorithm is to simulate the butterfly population to determine the direction of travel by secreting, sensing and analyzing the "fragrance". Compared with other population optimization algorithms, the basic butterfly optimization algorithm does not need to set other parameters through experience and multiple tests, and is easy to program. On the basis of not increasing the complexity of the basic butterfly optimization algorithm, the present invention adopts dynamic switching probability to expand the global search space in the early stage of the algorithm, improve the global exploration ability, accelerate the convergence speed, and introduce nonlinear adaptive weight factors when updating individual positions to ensure the local development ability of the algorithm in the later stage. At the same time, the optimal position of the current iteration is subjected to dimension-by-dimensional Cauchy mutation, and the greedy algorithm is used to determine whether to retain the updated mutated position, thereby ensuring that the population approaches the global optimal position and improving the solution accuracy.

[0110] It needs to be explained that, first, the parameters of the evaluation algorithm are set: Maximum number of iterations: Assume that the maximum number of iterations is set to 100 to limit the calculation time of the algorithm.

[0111] Search space: Set the search space, which includes all possible combinations of drug production features. The value range of each feature (such as temperature, humidity, equipment load, etc.) will be determined according to the actual production environment.

[0112] Within the set search space, the algorithm randomly generates a set of drug production feature combinations. These combinations include different values ​​of parameters such as temperature, humidity, and equipment load. Then, based on the current production environment and quality assessment model, the fitness value of each feature combination for drug production quality is calculated. Through the fitness function, we can obtain the quality fitness score of each production feature combination and evaluate its contribution to drug production quality. All generated feature combinations will be sorted according to the fitness value to determine the initial optimal production feature combination. In each round of optimization, the algorithm calculates the quality assessment value of each production feature combination and generates a random number based on the assessment value: Global search: If the random number is less than the preset switch probability (such as 0.3), a global search is performed. This means that the algorithm will jump out of the current local optimal solution, explore possible better solutions in the search space, and update the production feature combination.

[0113] Local search: If the random number is greater than the switch probability, a local search is performed. Local search will optimize some attributes of the current feature combination, such as adjusting parameters such as temperature and humidity, to try to find a more suitable feature combination.

[0114] Whether it is a global search or a local search, the new feature combination generated after optimization will recalculate its quality fitness value. All production feature combinations are reordered according to the fitness value, and the current optimal production feature combination is updated. At the end of each round of optimization, the algorithm will optimize the current optimal production feature combination dimension by dimension, and try to adjust the value of each feature to further improve the fitness. At this time, the algorithm will use the Cauchy mutation method to adjust the optimal feature combination: Cauchy mutation: According to the set mutation range, some parameters of the feature combination are randomly adjusted to generate new feature combinations. The results after mutation need to be subject to boundary control to ensure that the adjusted feature values ​​are still within the set reasonable range. Apply the greedy algorithm to compare the fitness values ​​of the production feature combinations before and after mutation: if the mutated feature combination has a higher fitness value, the mutated combination is retained. If the fitness value of the mutated feature combination is lower, the combination before mutation is retained. This process ensures that the production feature combination that contributes most to the quality of drug production is always retained. After each round of optimization, the algorithm checks whether the maximum number of iterations has been reached. If not, continue to optimize; if reached, output the current optimal production feature combination and its corresponding optimal fitness value as the production feature combination with the greatest contribution.

[0115] In this optional embodiment, the formula for adjusting the optimal production feature combination by using the Cauchy mutation method is: ; In the formula, represents the firstj The value of a combination of production characteristics; Indicates the current optimal j The value of a combination of production characteristics; d represents the number of production feature combinations; Cauchy (0, 1) represents a random number generated from a Cauchy distribution; rand Represents a random number, which is a randomly generated number in the interval [0, 1].

[0116] Specifically, rand is a randomly generated number in the interval [0, 1], with an expected value of 0.5. rand ,calculate π × rand The results will be distributed in [0, π ], and tan (·) in [0, π ] The range of values ​​in the interval is not symmetrical, which may easily lead to deviation. Therefore, rand -0.5 makes the random number range become [-0.5, 0.5], and then multiply by π , so that the angle range becomes [-0.5 π , 0.5 π ],so, tan The distribution of (·) is symmetrical, which is conducive to maintaining the balance of variation. It can change in both positive and negative directions.

[0117] In this optional embodiment, the greedy algorithm is used to compare the fitness values ​​of the production feature combination before and after the mutation, and the best fitness value after the mutation is retained, including: Randomly select an initial production feature combination as the first set of candidate solutions; Based on the current production feature combination, randomly select new features for combination, calculate the fitness value of the combination, and select the feature combination with the highest fitness value as the new production feature combination; Verify whether the new production feature combination meets the preset constraints. If not, cancel the new production feature combination, select the suboptimal feature combination and re-verify the constraints until a production feature combination that meets the preset constraints is obtained; Continuously optimize the feature combination until the combination reaches the local optimum, and determine the current production feature combination as the best production feature combination; On the basis of the current best combination of production features, a random factor is introduced to randomly mutate a combination of production features and find the next potential best combination of production features; Compare the fitness value of the mutated production feature combination with the current best combination. If the fitness value after mutation is greater than the fitness value before mutation, the mutated production feature combination is used as the new best production feature combination. Otherwise, the current best production feature combination is kept unchanged and the current greedy search process ends. When this round of greedy search ends, a new initial production feature combination is selected and the greedy algorithm optimization is performed again. Multiple rounds of search are repeated, and finally the optimal solution is selected from all the best production feature combinations found, and the best fitness value after mutation is retained.

[0118] Specifically, the greedy algorithm introduces random factors on the basis of the basic greedy algorithm to improve the greedy algorithm, solve the defects of the algorithm in initial estimation and optimization effect judgment, and improve the quality and accuracy of the algorithm solution. In the traditional greedy algorithm, since the current optimal solution is selected every time, the algorithm may stagnate in a local optimal area and no longer continue to search for the global optimal solution. By introducing random factors, the algorithm can jump out of the current optimal solution in some cases, thus having the opportunity to explore other potential better solutions. For example, the algorithm may choose a feature combination that is not the current optimal, but through random adjustments, it can find a better solution in subsequent steps.

[0119] It should be explained that, first, a production feature combination is randomly selected as the initial candidate solution. Assume that the feature combination includes production parameters such as temperature, humidity, and equipment load, and the initial values ​​of these parameters are randomly determined. This initial feature combination will serve as the starting point of the search process. Based on the current production feature combination, a production feature is randomly selected for adjustment. For example, one of the parameters such as temperature, humidity, or equipment load may be adjusted to generate a new feature combination. Subsequently, the fitness value of the new combination is calculated to evaluate its contribution to the quality of drug production. The fitness values ​​of the current feature combination and the new feature combination are compared, and the feature combination with a higher fitness value is selected as the new production feature combination. If the fitness value of the new combination is better, the combination is accepted and the optimization continues. After selecting a new feature combination, it is necessary to verify whether the combination meets the preset constraints. The preset constraints may include production environment requirements, equipment restrictions, quality standards, etc. If the new production feature combination does not meet these constraints, the combination is cancelled and the suboptimal feature combination is selected. Verification is continued until a production feature combination that meets all constraints is found. By continuously selecting and optimizing feature combinations, the algorithm will eventually find a local optimal solution. In this process, the production feature combination is gradually adjusted to the optimal state, meeting all constraints and achieving the optimal production quality under the current environment. Based on the current optimal production feature combination, random factors are introduced for mutation, and one or more parameters in the feature combination are randomly adjusted. Through this mutation operation, a possible better solution is found. The new combination after mutation will be used as one of the potential optimal production feature combinations. The fitness value of the mutated production feature combination is compared with the current best production feature combination. If the fitness value after mutation is higher than the fitness value before mutation, the mutated combination is accepted as the new best production feature combination. On the contrary, if the fitness value after mutation is lower, the current best production feature combination is kept unchanged, and the current greedy search process ends. When this round of greedy search ends, a new initial production feature combination is reselected and the greedy algorithm is executed again for optimization. This process will be carried out in multiple rounds of search, starting from a new initial combination each time, and constantly looking for a better solution. After multiple rounds of search, the one with the highest fitness value is selected from all the best production feature combinations found, and it is used as the final optimal solution. Ultimately, the algorithm retains the best fitness value obtained after mutation as the production feature combination that contributes most to the quality of drug production.

[0120] In this optional embodiment, the scheduling optimization module 3 includes: The model building module is used to obtain the data set of key factors, divide the data set into training set and test set, and establish a scheduling optimization model; Specifically, first, obtain a data set of key factors related to the drug production process. The data set should contain features that affect production quality, such as temperature, humidity, equipment load, etc. Next, divide the data set into a training set and a test set, usually in a ratio of 7:3 or 8:2, to ensure that the training and evaluation of the model are sufficiently representative. Then, use the training set data to train the scheduling optimization model, and select a suitable algorithm (such as regression analysis, neural network, or support vector machine) to predict production quality or scheduling efficiency. During the training process, adjust the model parameters to improve the accuracy and generalization ability of the model. After the training is completed, use the test set to verify the model, evaluate the model's predictive performance, and ensure that it works well on unseen data. Finally, establish a production scheduling optimization strategy based on the trained model, dynamically adjust the production schedule through the results of model prediction, optimize the allocation of production resources, and improve production efficiency and product quality.

[0121] The scheduling optimization module is used to optimize the established scheduling optimization model using the scheduling algorithm, and to dynamically adjust the key operations in the production process using the optimized scheduling optimization model to optimize the production scheduling strategy.

[0122] In this optional embodiment, the established scheduling optimization model is optimized by using a scheduling algorithm, and the optimized scheduling optimization model is used to dynamically adjust key operations in the production process. The optimized production scheduling strategy includes: Randomly generate several groups of initial production scheduling plans and set the maximum number of iterations of the scheduling algorithm; According to the production efficiency, the fitness value of each scheduling scheme is calculated, the number of new scheduling schemes generated by each scheduling scheme is calculated according to the quantity formula, and the standard deviation of each new scheduling scheme is calculated according to the standard deviation formula; Determine whether the number of current scheduling schemes reaches the threshold. If not, increase the number of iterations of the scheduling algorithm and continue to optimize. If it reaches the threshold, sort the current scheduling schemes from high to low according to the fitness value, select the best scheduling scheme as the best scheduling scheme for the current iteration, and select several scheduling schemes before reaching the threshold as the next generation candidate schemes. If the scheduling algorithm reaches the maximum number of iterations, the scheduling optimization process of the scheduling algorithm ends, otherwise, the scheduling optimization continues; Perform chaos search around the best scheduling solution, generate several new scheduling solutions, and calculate the fitness value of the new scheduling solution; if a new scheduling solution with the best fitness value is found, replace the original best scheduling solution as the new best scheduling solution, otherwise, continue the scheduling optimization process; The scheduling scheme with the best fitness value is output as the optimal scheduling strategy.

[0123] Specifically, the scheduling algorithm is an improved weed algorithm. The basic weed algorithm is a swarm intelligence algorithm that simulates the weed invasion process. It mainly has four steps: initializing the population; growth and reproduction; spatial distribution; and competition for survival. However, the basic weed algorithm is prone to lack of weed diversity and local optimal solutions in the later stage of the search. In view of the shortcomings of the basic weed algorithm, the present invention adopts an embedded chaotic local search strategy to perform chaotic search near the local optimal value, generate several sets of new solutions, and improve the local search ability of the algorithm.

[0124] It should be explained that, first, the algorithm will randomly generate several groups of initial production scheduling schemes. These schemes include the scheduling order and time arrangement of each link in the production process. To ensure the breadth and diversity of the search, the initial scheme has a certain degree of randomness. Set the maximum number of iterations of the scheduling algorithm to ensure sufficient search and optimization time during the calculation process. According to the production efficiency of each scheduling scheme, calculate the fitness value of the scheme. The fitness value reflects factors such as production efficiency and product quality. Then, use the quantity formula to calculate the number of new scheduling schemes generated by each scheduling scheme, and use the standard deviation formula to calculate the standard deviation of each new scheduling scheme. This helps to evaluate the stability and feasibility of the scheduling scheme. Determine whether the number of scheduling schemes currently generated has reached the set threshold. If the threshold has not been reached, the scheduling algorithm will continue to optimize, increase the number of iterations, and generate more scheduling schemes. If the threshold is reached, the algorithm will sort the currently generated scheduling schemes according to the fitness value, select the best scheduling scheme from high to low as the best scheduling scheme for the current iteration, and retain several scheduling schemes that have not reached the threshold as the next generation of candidate schemes. If the number of iterations of the scheduling algorithm has not reached the maximum value, the algorithm will continue to optimize the scheduling. If the maximum number of iterations has been reached, the optimization process of the scheduling algorithm ends. After finding the current best scheduling solution, the algorithm will perform a chaotic search around it and generate several new scheduling solutions. After these new solutions are calculated through fitness value, if there is a new solution with a better fitness value, it will replace the original best scheduling solution and become the new best solution. If there is no better solution, the scheduling optimization will continue. When the scheduling algorithm completes all iterations, it will output the scheduling solution with the best fitness value as the final optimal production scheduling strategy. This strategy will be applied in the production process to dynamically adjust the production schedule to ensure maximum production efficiency and optimized product quality.

[0125] In this optional embodiment, the formula for calculating the fitness value of each scheduling scheme is: ; The quantity formula is: ; The standard deviation formula is: ; In the formula, F Represents the fitness value of the scheduling scheme; Q represents production efficiency (i.e., objective function); e represents the penalty term; Number _ Seeds [ i ] indicates the i The number of new scheduling solutions generated by the scheduling solution; F i Indicates i The fitness value of each scheduling scheme; F g represents the optimal fitness value; F w Indicates the worst fitness value; S max Indicates the maximum number of new scheduling solutions generated; S min Indicates the minimum number of new scheduling solutions generated; s iter Indicates that it is iter The standard deviation of the iterations; iter max Indicates the maximum number of iterations; s i represents the starting standard deviation; s f represents the final standard deviation; n represents the nonlinear harmonic parameter.

[0126] According to another embodiment of the present invention, Figure 2 As shown, a drug production process scheduling optimization method based on artificial intelligence is also provided, and the method includes the following steps: S1. Obtain real-time production data during the drug production process and extract key production feature data during the production process; S2. Analyze the key production characteristic data obtained using factor analysis algorithms to identify key factors affecting drug production quality; S3. Based on key factors, a scheduling optimization model is established, and the scheduling optimization model is used to dynamically adjust key operations in the production process and optimize the production scheduling strategy.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A drug production process scheduling optimization system based on artificial intelligence, characterized in that: include: Data management module, used to obtain real-time production data during the drug production process and extract key production feature data during the production process; The data analysis module is used to analyze the key production characteristic data obtained by using the factor analysis algorithm to identify the key factors affecting the quality of drug production; The scheduling optimization module is used to establish a scheduling optimization model based on key factors, dynamically adjust key operations in the production process using the scheduling optimization model, and optimize the production scheduling strategy; The key production characteristic data obtained by using the factor analysis algorithm are analyzed to identify the key factors affecting the quality of drug production, including: Define the parameters of the factor analysis algorithm and the quality assessment model; randomly generate several groups of initial production feature combinations; Use historical production data to train the quality assessment model, calculate the quality impact coefficient of each feature combination, record the optimal production feature combination according to different production parameter configurations, and evaluate its impact on drug production quality; Based on changes in the production environment, the sensitivity parameters of the factor analysis algorithm are updated in real time according to the dynamic adjustment function, and the parameters in the factor analysis model are dynamically adjusted by sensing environmental changes during the production process; The evaluation algorithm is used to evaluate the contribution of each combination of production features to the quality of drug production. In each round of optimization, the production environment changes and the quality evaluation model are combined to select the production feature combination with the greatest contribution. The selected feature combinations with the greatest contribution are input into the quality assessment model to identify the key factors affecting the quality of drug production.

2. According to claim 1, a drug production process scheduling optimization system based on artificial intelligence is characterized in that: The data management module comprises: The data acquisition module is used to acquire real-time production data in the drug production process through sensors, and to perform denoising, filtering and smoothing preprocessing on duplicate data, missing values ​​and abnormal values ​​of the real-time production data to obtain processed real-time production data; The feature extraction module uses the Fourier transform method to extract features from the processed real-time production data to obtain key production feature data in the production process.

3. The drug production process scheduling optimization system based on artificial intelligence according to claim 1 is characterized in that: The evaluation algorithm is used to evaluate the contribution of each group of production feature combinations to the drug production quality. In each round of optimization, the production feature combinations with the greatest contribution are screened out by combining the production environment changes and the quality evaluation model, including: Initialize the evaluation algorithm, set the maximum number of iterations of the evaluation algorithm, and set the search space; In the set search space, a set of drug production feature combinations is randomly generated, and the fitness value of each feature combination for drug quality is calculated based on the current production environment and quality assessment model; according to the size of the fitness value, the initial optimal production feature combination is sorted and determined; In each round of optimization, the quality evaluation value of each production feature combination is calculated, and a random number is generated. According to the preset switch probability, whether to perform a global search or a local search is selected. If the random number is less than the preset switch probability, a global search is performed to update the production feature combination; otherwise, a local search is performed to optimize some attributes of the production feature combination. After the production feature combination is updated, its corresponding quality fitness value is evaluated again, and all production feature combinations are sorted according to the fitness value to update the current optimal production feature combination; Optimize the current optimal production feature combination dimension by dimension, adjust the optimal production feature combination using the Cauchy mutation method, and perform boundary control on the result after mutation; Using the greedy algorithm, the fitness values ​​of the production feature combinations before and after mutation are compared, and the best fitness value after mutation is retained; Determine whether the evaluation algorithm has reached the maximum number of iterations. If so, output the optimal production feature combination and the best fitness value as the production feature combination with the greatest contribution.

4. The drug production process scheduling optimization system based on artificial intelligence according to claim 3 is characterized in that: The formula for adjusting the optimal production feature combination by using the Cauchy mutation method is: ； In the formula, represents the first j The value of a combination of production characteristics; Indicates the current optimal j The value of a combination of production characteristics; d represents the number of production feature combinations; Cauchy (0, 1) represents a random number generated from a Cauchy distribution; rand Represents a random number.

5. The drug production process scheduling optimization system based on artificial intelligence according to claim 3 is characterized in that: The greedy algorithm is used to compare the fitness values ​​of the production feature combination before and after the mutation, and the best fitness value after the mutation is retained, including: Randomly select an initial production feature combination as the first set of candidate solutions; Based on the current production feature combination, randomly select new features for combination, calculate the fitness value of the combination, and select the feature combination with the highest fitness value as the new production feature combination; Verify whether the new production feature combination meets the preset constraints. If not, cancel the new production feature combination, select the suboptimal feature combination and re-verify the constraints until a production feature combination that meets the preset constraints is obtained; Continuously optimize the feature combination until the combination reaches the local optimum, and determine the current production feature combination as the best production feature combination; On the basis of the current best combination of production features, random factors are introduced to randomly mutate a combination of production features and find the next potential best combination of production features; Compare the fitness value of the mutated production feature combination with the current best combination. If the fitness value after mutation is greater than the fitness value before mutation, the mutated production feature combination is used as the new best production feature combination. Otherwise, the current best production feature combination is kept unchanged and the current greedy search process ends. When this round of greedy search ends, a new initial production feature combination is selected and the greedy algorithm optimization is performed again. Multiple rounds of search are repeated, and finally the optimal solution is selected from all the best production feature combinations found, and the best fitness value after mutation is retained.

6. The drug production process scheduling optimization system based on artificial intelligence according to claim 1 is characterized in that: The scheduling optimization module includes: The model building module is used to obtain the data set of key factors, divide the data set into training set and test set, and establish a scheduling optimization model; The scheduling optimization module is used to optimize the established scheduling optimization model using the scheduling algorithm, and to dynamically adjust the key operations in the production process using the optimized scheduling optimization model to optimize the production scheduling strategy.

7. The drug production process scheduling optimization system based on artificial intelligence according to claim 6 is characterized in that: The scheduling algorithm is used to perform scheduling optimization on the established scheduling optimization model, and the optimized scheduling optimization model is used to dynamically adjust key operations in the production process. The optimized production scheduling strategy includes: Randomly generate several groups of initial production scheduling plans and set the maximum number of iterations of the scheduling algorithm; According to the production efficiency, the fitness value of each scheduling scheme is calculated, the number of new scheduling schemes generated by each scheduling scheme is calculated according to the quantity formula, and the standard deviation of each new scheduling scheme is calculated according to the standard deviation formula; Determine whether the number of current scheduling schemes reaches the threshold. If not, increase the number of iterations of the scheduling algorithm and continue to optimize. If it reaches the threshold, sort the current scheduling schemes from high to low according to the fitness value, select the best scheduling scheme as the best scheduling scheme for the current iteration, and select several scheduling schemes before reaching the threshold as the next generation candidate schemes. If the scheduling algorithm reaches the maximum number of iterations, the scheduling optimization process of the scheduling algorithm ends, otherwise, the scheduling optimization continues; Perform chaos search around the best scheduling solution, generate several new scheduling solutions, and calculate the fitness value of the new scheduling solution; if a new scheduling solution with the best fitness value is found, replace the original best scheduling solution as the new best scheduling solution, otherwise, continue the scheduling optimization process; The scheduling scheme with the optimal fitness value is output as the optimal scheduling strategy.

8. The drug production process scheduling optimization system based on artificial intelligence according to claim 7 is characterized in that: The formula for calculating the fitness value of each scheduling scheme is: ； The quantity formula is: ； The standard deviation formula is: ； In the formula, F Represents the fitness value of the scheduling scheme; Q Indicates production efficiency; ε represents the penalty term; Num _ Seeds [ i ] indicates the i The number of new scheduling solutions generated by the scheduling solution; F i Indicates i The fitness value of each scheduling scheme; F g represents the optimal fitness value; F w Indicates the worst fitness value; S max Indicates the maximum number of new scheduling solutions generated; S min Indicates the minimum number of new scheduling solutions generated; σ iter Indicates that it is iter The standard deviation of the iterations; iter max Indicates the maximum number of iterations; σ i represents the starting standard deviation; σ f represents the final standard deviation; n represents the nonlinear harmonic parameter.

9. A drug production process scheduling optimization method based on artificial intelligence, using the drug production process scheduling optimization system based on artificial intelligence as described in any one of claims 1 to 8, characterized in that: The method comprises the following steps: S1. Obtain real-time production data during the drug production process and extract key production feature data during the production process; S2. Analyze the key production characteristic data obtained using factor analysis algorithms to identify key factors affecting drug production quality; S3. Based on key factors, a scheduling optimization model is established, and the scheduling optimization model is used to dynamically adjust key operations in the production process and optimize the production scheduling strategy.

Images