A real-time monitoring and optimization system for smart factory production processes

Abstract

This invention discloses a real-time monitoring and optimization system for intelligent factory production processes, belonging to the field of intelligent factory production management. Specifically, it includes: acquiring process parameters and production action sequence data of the pipe fitting forming process using the Industrial Internet of Things (IIoT) and transmitting them to a data processing center; constructing a motion waste analysis model based on Markov chain theory and the motion sequence data; identifying motion waste by calculating state transition probabilities and steady-state distributions and comparing them with standard motion flows; and feeding the results back to the production management system; combining process parameters and motion waste identification results, the production management system uses queuing theory and operations research to construct a dynamic bottleneck station identification model, calculating the probability of each station becoming a bottleneck in real time, and determining the dynamic bottleneck station based on the amount of work-in-process inventory and production efficiency; and adjusting the production cycle based on the dynamic bottleneck station identification results to achieve real-time monitoring and optimization of the production process, thereby improving production efficiency and quality.

Description

Technical Field

[0001] This invention belongs to the field of intelligent factory production management, specifically a real-time monitoring and optimization system for intelligent factory production processes. Background Technology

[0002] Numerous problems exist in the production process of smart factories, such as space waste due to unreasonable material stacking, time waste due to numerous unnecessary steps in production, difficulty in accurately identifying bottleneck workstations on the production line in real time, and the inability to adaptively adjust production takt time according to actual conditions. These problems affect production efficiency and enterprise benefits. Traditional production monitoring and optimization methods are insufficient to comprehensively, in real time, and accurately solve these problems, and cannot meet the needs of efficient and lean production in smart factories.

[0003] For example, Chinese patent application CN109360118A discloses a factory status monitoring method, apparatus, system, equipment, and storage medium, including: acquiring the status of factory change items, setting scores for the status of the factory change items to obtain factory change item status scores, wherein the factory change items include at least production changes, inventory changes, and production-sales balance changes, and the status of the factory change items includes at least four states: normal, watchful, abnormal, and warning; calculating a factory status score based on the factory change item status scores, and determining the factory status based on the factory status scores. This invention monitors the status of various factory change items in the production process in real time, and by digitizing the status of each factory change item, monitors whether the factory status is normal, thereby achieving the purpose of real-time monitoring of production, timely detection of production errors, and making factory operations more efficient.

[0004] The existing technologies mentioned above have the following problems: they mainly focus on factory change items such as output changes, inventory changes, and production-sales balance changes, emphasizing macro-level status monitoring at the overall factory operation level; they simply set scores for the status of factory change items and then calculate the factory status score to judge the factory status, the analysis method is relatively simple, and lacks in-depth exploration of the inherent laws of the production process; and there is no detailed feedback and adjustment mechanism for specific problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a real-time monitoring and optimization system for intelligent factory production processes. It utilizes the Industrial Internet of Things (IIoT) to acquire process parameters and production motion sequence data for pipe fitting forming, and transmits this data to a data processing center. Based on Markov chain theory, a motion waste analysis model is constructed using the motion sequence data. By calculating state transition probabilities and steady-state distributions, motion waste is identified by comparing with standard motion flows, and the results are fed back to the production management system. The production management system combines the process parameters of the pipe fitting forming process with the motion waste identification results, and uses queuing theory and operations research to construct a dynamic bottleneck station identification model. It calculates the probability of each station becoming a bottleneck in real time, and determines the dynamic bottleneck station by combining the work-in-process inventory and production efficiency. Based on the dynamic bottleneck station identification results, the production cycle is adjusted to achieve real-time monitoring and optimization of the production process, thereby improving production efficiency and quality.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A real-time monitoring and optimization system for intelligent factory production processes includes: a data acquisition module, a motion waste analysis module, a bottleneck station identification module, and a production cycle adjustment module;

[0008] The motion waste analysis module constructs a motion waste analysis model based on Markov chain theory according to the motion sequence data collected by the data acquisition module. By calculating the state transition probability and the steady-state distribution of the Markov chain, it identifies motion waste situations.

[0009] The bottleneck station identification module, based on the received action waste identification results and combined with the pipe forming process parameters collected by the data acquisition module, constructs a dynamic bottleneck station identification model to quantitatively evaluate the probability value of each station evolving into a bottleneck station at different time points.

[0010] The production cycle adjustment module adjusts the production cycle based on the identification results of the dynamic bottleneck station.

[0011] Specifically, the motion waste analysis module constructs a motion waste analysis model based on Markov chain theory according to the motion sequence data, and identifies motion waste situations, including:

[0012] A1: Obtain the action sequence data during the pipe fitting production process and perform preprocessing;

[0013] A2: Based on the pipe fitting production process, a unique discrete state code is defined for each action, and a mapping rule from action to discrete state is formulated. The mapping is performed according to the name of the action. By traversing the preprocessed action sequence data, each action is converted into the corresponding discrete state code according to the mapping rule.

[0014] A3: Analyze the discrete action sequence data, classify the actions into different categories according to their type, and combine K actions into an action state based on the execution order of the actions; the action state includes raw material classification and screening, processing technology, quality inspection, and finished product classification and handling;

[0015] A4: After removing duplicate action states from all the action states obtained in A3, summarize them and form the state space of the Markov chain. ,in, This represents the nth action state, where n represents the number of action states.

[0016] A5: Starting from the first action state in the preprocessed action sequence data, check adjacent action states sequentially. and ;

[0017] If two adjacent action states are different, then a state transition has occurred, that is, from one action state... Transition to another action state ;

[0018] A6: The transition frequency is obtained by counting the number of state transitions in the preprocessed action sequence data using an initialized counter. ;

[0019] A7: By calculating the transfer frequency With Transfer to all non The ratio of the total number of times the action state is used to obtain the state transition probability. ;

[0020] A8: Create The matrix, with the state transition probabilities calculated in A7. Assigning values ​​to the matrix yields the action transition matrix. .

[0021] Specifically, the motion waste analysis module, based on the motion sequence data and Markov chain theory, constructs a motion waste analysis model to identify motion waste situations, and further includes:

[0022] A9: Based on the properties of steady-state distribution, establish the Markov chain equilibrium equations, and solve the Markov chain equilibrium equations using a system of linear equations to obtain the steady-state distribution vector. ;

[0023] A10: Based on the steady-state distribution vector Determine the frequency of occurrence of each action state in the actual production process. The frequency of occurrence of each action state is the steady-state probability of each action state. ,and ;

[0024] A11: Based on the process requirements for pipe fitting production, a standard frequency range is set for each operating state. At the same time, a preset standard threshold is set for each action state. ,in, and Let represent the minimum and maximum values ​​of the standard frequency of the i-th action state, respectively;

[0025] A12: For each action state , action state frequency of occurrence With standard frequency range Compare;

[0026] like ,and Then determine the action state. There is wasted action;

[0027] A13: Feed back the results of motion waste identification to the production management system through the data interface.

[0028] Specifically, the Markov chain equilibrium equations simultaneously satisfy:

[0029] The system transitions from all previous action states to the action state. The sum of the weighted state transition probabilities equals the system's state of action. The steady-state probability;

[0030] The sum of the steady-state probabilities of the system in all action states is 1.

[0031] Specifically, the specific steps of A10 include:

[0032] A10.1: Obtaining the steady-state distribution vector The state space of Markov chains ;

[0033] A10.2: Traversal The index of each action state, from the steady-state distribution vector. Extract the element value at the corresponding position, and the element value is the steady-state probability of the action state.

[0034] Specifically, the bottleneck station identification module, based on the received action waste identification results and combined with the pipe forming process parameters collected by the data acquisition module, constructs a dynamic bottleneck station identification model to quantitatively evaluate the probability of each station evolving into a bottleneck station at different time points, including:

[0035] B1: Acquire the results of motion waste identification and the process parameters of the pipe forming process. At the same time, monitor the amount of work-in-process and production efficiency at each station in real time. The results of motion waste identification include wasted motions and the degree of waste.

[0036] B2: Assign a unique identifier to each workstation, treat each workstation as a queuing system, and treat work-in-process as a customer. Based on the process parameters of the pipe forming process and the results of motion waste identification, determine the arrival rate and service rate of each queuing system; the arrival rate is the rate at which work-in-process arrives at the workstation; the service rate is the rate at which the workstation processes work-in-process.

[0037] B3: Combining the M / M / 1 model, a dynamic bottleneck workstation identification model is established using dynamic programming methods;

[0038] B4: Integrate the acquired motion waste identification results, pipe forming process parameters, and real-time monitoring of the work-in-process accumulation quantity and production efficiency of each station to form pipe data. Input the pipe data into the dynamic bottleneck station identification model and use the dynamic bottleneck station identification model to calculate the probability value of each station evolving into a bottleneck station at different time points in real time.

[0039] Specifically, the bottleneck station identification module, based on the received action waste identification results and combined with the pipe forming process parameters collected by the data acquisition module, constructs a dynamic bottleneck station identification model to quantitatively evaluate the probability value of each station evolving into a bottleneck station at different time points. It also includes:

[0040] B5: Real-time monitoring of the work-in-process inventory and production efficiency at each workstation, and setting a threshold for the work-in-process inventory.

[0041] If the amount of work-in-process at any workstation exceeds the work-in-process quantity threshold and its production efficiency is lower than the demand efficiency of the downstream workstation, the workstation is determined to be the current dynamic bottleneck workstation by combining the probability value calculated by the dynamic bottleneck workstation identification model.

[0042] B6: Output the identification results of dynamic bottleneck workstations in chart form and feed them back to the production management system.

[0043] Specifically, the steps of B2 include:

[0044] B2.1: Identifier rules shall be formulated according to the characteristics of the workstations; the identifier rules shall adopt a combination of letters and numbers, where letters represent work sections and numbers represent the workstation numbers within the work sections;

[0045] B2.2: Assign a unique identifier to each workstation according to the established identifier rules, and establish a workstation information table; the workstation information table includes the workstation's name, function, and identifier information;

[0046] B2.3: Perform conceptual mapping, treating each workstation as a queuing system, work-in-process as customers, and the processing equipment and operators at the workstation as service counters. At the same time, define the queuing rule for work-in-process in front of the workstation as a first-come, first-served rule.

[0047] B2.4: Collect historical data on the arrival of work-in-process at each workstation during the pipe fitting production process, record the timestamp of the arrival of work-in-process, and calculate the time interval between the arrival of two adjacent work-in-process at each workstation based on the timestamp. ;

[0048] B2.5: Calculation time interval average and the average value Find the reciprocal to obtain the arrival rate of each queuing system. ;

[0049] B2.6: Analyze the results of waste motion identification, assess the impact of wasteful motions on the arrival rate of work-in-process, and correct the calculated arrival rate according to the degree of impact;

[0050] B2.7: Determine the standard processing time for each station based on the process parameters of the pipe forming process. The analysis of wasted motion identification results assesses the impact of wasted motions on the efficiency of work-in-process handling at workstations, and adjusts the standard processing time based on the degree of impact to obtain the actual processing time. ;

[0051] B2.8: Yes Find the reciprocal to obtain the service rate of each queuing system. .

[0052] Specifically, the steps of B3 include:

[0053] B3.1: Determine the decision to be made in each action state, load the pre-trained M / M / 1 model, and simultaneously obtain the arrival rate. Service rate As parameters of the M / M / 1 model;

[0054] B3.2: Set the minimization of the production cycle as the objective function, and derive the state transition equation based on the M / M / 1 model and decision rules;

[0055] B3.3: Based on the objective function and the state transition equation, establish the Bellman equation and solve the Bellman equation using the value iteration method;

[0056] B3.4: By solving the Bellman equations, the optimal decision, i.e., the optimal policy, is obtained for each action state. A dynamic bottleneck workstation identification model was obtained.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] 1. This invention proposes a real-time monitoring and optimization system for intelligent factory production processes, and optimizes and improves its architecture, operation steps and processes. The system has the advantages of simple process, low investment and operating costs, and low production operating costs.

[0059] 2. This invention proposes a real-time monitoring and optimization system for intelligent factory production processes. By collecting process parameters and action sequence data of pipe forming process through the Industrial Internet of Things, and using Markov chain theory to construct a motion waste analysis model, it can accurately identify motion waste. This process achieves in-depth insight into the details of the production process, quantifies the waste situation, and helps to optimize the production process from the source of operation, reduce unnecessary motion consumption, and improve production efficiency.

[0060] 3. This invention proposes a real-time monitoring and optimization system for intelligent factory production processes. In terms of bottleneck station identification and production rhythm adjustment, the production management system combines the process parameters of the pipe forming process with the results of motion waste identification. It uses queuing theory and operations research to construct a dynamic bottleneck station identification model, which can identify bottleneck stations in real time and adjust the production rhythm accordingly. This effectively avoids the accumulation of work-in-process, coordinates and matches the production rhythm of each station, ensures a smooth and efficient production process, and thus improves overall production quality, reduces production costs, and enhances the lean production level of the intelligent factory. Attached Figure Description

[0061] Figure 1 This is a system architecture diagram of a smart factory production process real-time monitoring and optimization system according to the present invention;

[0062] Figure 2 This is a flowchart illustrating the principle of a real-time monitoring and optimization system for intelligent factory production processes according to the present invention.

[0063] Figure 3 This is a flowchart illustrating the action waste identification process of a real-time monitoring and optimization system for intelligent factory production processes according to the present invention.

[0064] Figure 4This invention provides a flowchart for determining dynamic bottleneck workstations in a real-time monitoring and optimization system for intelligent factory production processes. Detailed Implementation

[0065] Example 1:

[0066] Please see Figure 1 and Figure 2 One embodiment of the present invention provides a real-time monitoring and optimization system for intelligent factory production processes, comprising:

[0067] Data acquisition module, motion waste analysis module, bottleneck station identification module, production cycle adjustment module;

[0068] The data acquisition module is used to collect process parameters of the pipe fitting forming process and action sequence data during the pipe fitting production process, and transmit the data to the data processing center; the process parameters of the pipe fitting forming process include temperature, pressure, and speed; the action sequence data includes the sequence of actions of workers operating the equipment and the operating data of the equipment itself;

[0069] The motion waste analysis module is used to analyze motion sequence data in the production process based on Markov chain theory, identify motion waste, and feed the results back to the production management system.

[0070] The bottleneck station identification module is used to combine the motion waste identification results with the process parameters of the pipe forming process to build a dynamic bottleneck station identification model, monitor the status of each station in real time, and determine the current dynamic bottleneck station.

[0071] The production cycle adjustment module is used to adjust the production cycle based on the identification results of dynamic bottleneck workstations, so as to optimize the production process and improve production efficiency.

[0072] The motion waste analysis module includes: a model building unit, a model calculation unit, and a motion waste identification unit;

[0073] The model building unit is used to build a motion waste analysis model based on the collected motion sequence data according to Markov chain theory.

[0074] The model computation unit is used to process and compute action sequence data, determine the state transition probability from one action state to another, and compute the steady-state distribution of the Markov chain.

[0075] The motion waste identification unit is used to analyze the frequency of occurrence of each motion state in the actual production process based on the steady-state distribution and state transition probability of the Markov chain, compare it with the standard motion flow, and identify motion waste.

[0076] The bottleneck workstation identification module includes: a probability calculation unit, a status monitoring unit, and a bottleneck judgment unit;

[0077] The probability calculation unit is used to calculate in real time the probability that each workstation will become a bottleneck workstation at different times using the constructed dynamic bottleneck workstation identification model; the dynamic bottleneck workstation identification model is constructed based on the motion waste identification results and the process parameters of the pipe forming process.

[0078] The status monitoring unit is used to monitor the amount of work-in-process inventory and production efficiency at each workstation in real time, and to obtain the actual operating status of the workstation.

[0079] The bottleneck identification unit is used to determine the current dynamic bottleneck station when the number of work-in-process items at any station exceeds the set threshold for the number of work-in-process items and its production efficiency is lower than the demand efficiency of the downstream station, based on the calculation results of the model.

[0080] The production cycle adjustment module includes: an adjustment decision unit and an adjustment execution unit;

[0081] The adjustment decision unit is used to formulate strategies and plans for adjusting production cycle time based on the identification results of dynamic bottleneck workstations, and to determine the magnitude and direction of the adjustment;

[0082] The adjustment execution unit is used to perform production rhythm adjustment operations according to the plan formulated by the adjustment decision unit. For example, it adjusts the operating speed and rhythm of equipment to make the production process smoother and reduce the impact of bottleneck stations on production.

[0083] Specifically, the overall process of a smart factory production process real-time monitoring and optimization system includes:

[0084] S1: Acquire process parameters for pipe fitting forming process and action sequence data during pipe fitting production process, and transmit them to the data processing center via the Industrial Internet of Things;

[0085] S2: Based on the action sequence data in the pipe fitting production process, a motion waste analysis model is constructed based on Markov chain theory. By processing and calculating the action sequence data, the state transition probability from one action state to another is determined, and the steady-state distribution of the Markov chain is calculated. Based on the steady-state distribution and state transition probability of the Markov chain, the frequency of occurrence of each action state in the actual production process is analyzed. By comparing with the standard action flow, motion waste is identified, and the motion waste identification results are fed back to the production management system.

[0086] S3: After receiving the motion waste identification results, the production management system combines the process parameters of the pipe forming process with the motion waste identification results to build a dynamic bottleneck station identification model. It calculates in real time the probability of each station becoming a bottleneck station at different times. At the same time, it monitors the work-in-process inventory and production efficiency of each station in real time. When the work-in-process inventory of any station exceeds the set work-in-process inventory threshold and its production efficiency is lower than the demand efficiency of the downstream station, it determines that the station is the current dynamic bottleneck station based on the model calculation results.

[0087] S4: Adjust the production cycle based on the identification results of dynamic bottleneck workstations.

[0088] Example 2:

[0089] Please see Figure 3 In this embodiment, the motion waste analysis module constructs a motion waste analysis model based on Markov chain theory according to the motion sequence data, and identifies motion waste situations, including:

[0090] A1: Obtain the action sequence data during the pipe fitting production process and perform preprocessing;

[0091] A2: Based on the pipe fitting production process, a unique discrete state code is defined for each action, and a mapping rule from action to discrete state is formulated. The mapping is performed according to the name of the action. By traversing the preprocessed action sequence data, each action is converted into the corresponding discrete state code according to the mapping rule.

[0092] For example, if the action name contains the word "screening", it is mapped to the status code corresponding to "raw material classification and screening".

[0093] A3: Analyze the discrete action sequence data, classify the actions into different categories according to their type, and combine K actions into an action state based on the execution order of the actions; the action state includes raw material classification and screening, processing technology, multi-dimensional quality inspection, and finished product classification and handling;

[0094] Each action state is formed by a combination of consecutive actions with the same function.

[0095] For example, all series of raw material screening actions can be merged into the "raw material classification screening" state.

[0096] A4: After removing duplicate action states from all the action states obtained in A3, summarize them and form the state space of the Markov chain. ,in, This represents the nth action state, where n represents the number of action states.

[0097] A5: Starting from the first action state in the preprocessed action sequence data, check adjacent action states sequentially. and ;

[0098] If two adjacent action states are different, then a state transition has occurred, that is, from one action state... Transition to another action state ;

[0099] A6: The transition frequency is obtained by counting the number of state transitions in the preprocessed action sequence data using an initialized counter. ;

[0100] A7: By calculating the transfer frequency With Transfer to all non The ratio of the total number of times the action state is used to obtain the state transition probability. ,in, Indicates from Transfer to non The frequency of transitions in the action's state;

[0101] A8: Create The matrix, with the state transition probabilities calculated in A7. Assigning values ​​to the matrix yields the action transition matrix. ;

[0102] A9: Based on the properties of steady-state distribution, establish the Markov chain equilibrium equations, and solve the Markov chain equilibrium equations using a system of linear equations to obtain the steady-state distribution vector. ;

[0103] The Markov chain equilibrium equations in A9 simultaneously satisfy:

[0104] The system transitions from all possible previous action states to the action state. The sum of the weighted state transition probabilities equals the system's state of action. The steady-state probability;

[0105] The sum of the steady-state probabilities of the system in all action states is 1.

[0106] The steady-state distribution is a stable state that a Markov chain reaches after running for a long time, where the probability of each state no longer changes with time. The steady-state distribution vector is usually a one-dimensional array with the same number of elements as the number of action states.

[0107] For example, suppose we have a Markov chain with a state space of... The state transition probability matrix is ;

[0108] Let the steady-state distribution vector be... According to the specific formula of the Markov chain equilibrium equation We can obtain the following system of equations: By solving this system of equations, the steady-state distribution vector can be obtained. This means that after a long period of operation, the system is in an active state. The steady-state probability is In action state The steady-state probability is .

[0109] A10: Based on the steady-state distribution vector Determine the frequency of occurrence of each action state in the actual production process. The frequency of occurrence of each action state is the steady-state probability of each action state. ,and ;

[0110] A11: Based on the process requirements for pipe fitting production, a standard frequency range is set for each operating state. At the same time, a preset standard threshold is set for each action state. ,in, and Let represent the minimum and maximum values ​​of the standard frequency of the i-th action state, respectively;

[0111] A12: For each action state , action state Frequency of occurrence With standard frequency range Compare;

[0112] like ,and Then determine the action state. There is wasted action;

[0113] A13: Feed back the results of motion waste identification to the production management system through the data interface.

[0114] The specific steps of A10 include:

[0115] A10.1: Obtaining the steady-state distribution vector The state space of Markov chains ;

[0116] A10.2: Traversal The index of each action state, from the steady-state distribution vector. Extract the element value at the corresponding position, and this element value is the steady-state probability of the action state.

[0117] Example 3:

[0118] Please see Figure 4 In this embodiment, the bottleneck station identification module, based on the received action waste identification results and combined with the pipe forming process parameters collected by the data acquisition module, constructs a dynamic bottleneck station identification model to quantitatively evaluate the probability of each station evolving into a bottleneck station at different time points, including:

[0119] B1: Acquire the results of motion waste identification and the process parameters of the pipe forming process. At the same time, monitor the amount of work-in-process and production efficiency at each station in real time. The results of motion waste identification include wasted motions and the degree of waste.

[0120] B2: Assign a unique identifier to each workstation, treat each workstation as a queuing system, and treat work-in-process as a customer. Based on the process parameters of the pipe forming process and the results of motion waste identification, determine the arrival rate and service rate of each queuing system; the arrival rate is the rate at which work-in-process arrives at the workstation; the service rate is the rate at which the workstation processes work-in-process.

[0121] B3: Combining the M / M / 1 model, a dynamic bottleneck workstation identification model is established using dynamic programming methods;

[0122] B4: Integrate the acquired motion waste identification results, pipe forming process parameters, and real-time monitoring of the work-in-process accumulation quantity and production efficiency of each station to form pipe data. Input the pipe data into the dynamic bottleneck station identification model and use the dynamic bottleneck station identification model to calculate the probability value of each station evolving into a bottleneck station at different time nodes in real time.

[0123] Furthermore, the specific steps of B4 include:

[0124] B4.1: Integrate the results of motion waste identification, process parameters of pipe forming process, work-in-process accumulation quantity and production efficiency to form pipe data containing multiple features. These data can be organized into a data table, where each row represents the data of a workstation at a certain point in time, and each column corresponds to different features, such as workstation number, motion waste index, process parameter value, work-in-process accumulation quantity, and production efficiency.

[0125] B4.2: Input the integrated pipe fitting data into the dynamic bottleneck station identification model. Based on the input data, combined with its own algorithm and parameters, calculate the probability value of each station evolving into a bottleneck station at different time points, specifically including:

[0126] (1) Check the integrated pipe fitting data, remove missing values ​​and outliers, and standardize them. For example, if the number of work-in-process inventory is negative, or the production efficiency is an unreasonable maximum value, it needs to be corrected or removed.

[0127] (2) Load the pre-trained dynamic bottleneck station identification model and set the parameters required for model operation, such as discount factor and time step;

[0128] (3) Determine the initial state of the model based on the input pipe fitting data. The initial state usually includes information such as the quantity of work-in-process at each station and the current time.

[0129] (4) Based on the input pipe fitting data and the state transition equation of the dynamic bottleneck station identification model, calculate the probability and result of the system transitioning from the current state to the next state at each time node;

[0130] (5) Based on the state transition results, calculate the objective function value for each state. The objective function is usually to minimize the production cycle or maximize the production efficiency.

[0131] (6) By solving the Bellman equation, the probability value of each work station evolving into a bottleneck work station at different time points is obtained.

[0132] B5: Real-time monitoring of the work-in-process inventory and production efficiency at each workstation, and setting a threshold for the work-in-process inventory.

[0133] If the work-in-process inventory at any workstation exceeds the work-in-process inventory threshold, and its production efficiency is lower than the demand efficiency of downstream workstations, the workstation is identified as the current dynamic bottleneck workstation based on the probability value calculated by the dynamic bottleneck workstation identification model. Specifically, this includes:

[0134] (1) The number of work-in-process items at each workstation is collected in real time by sensors installed at the workstation;

[0135] (2) Monitor the production efficiency of each workstation, such as recording the number of products completed or the amount of work processed per unit time. At the same time, determine the demand efficiency of downstream workstations based on their production capacity, production cycle time and order demand.

[0136] (3) Compare the number of work-in-process inventory at each workstation with the pre-set threshold for the number of work-in-process inventory, compare the production efficiency of the workstation with the demand efficiency of the downstream workstation, and determine whether the condition of production efficiency being lower than the demand efficiency of the downstream workstation is met.

[0137] (4) When any workstation simultaneously meets the two conditions of the number of work-in-process exceeding the threshold of the number of work-in-process and the production efficiency being lower than the demand efficiency of the downstream workstation, check the probability value of the workstation evolving into a bottleneck workstation at the current time node calculated by the dynamic bottleneck workstation identification model. In this invention, the probability value of the bottleneck workstation is obtained by processing and calculating the process parameters of the pipe forming process and the results of motion waste identification by the dynamic bottleneck workstation identification model.

[0138] (5) Make a comprehensive judgment based on the set rules and probability values. For example, if the probability value exceeds the set threshold, the workstation is determined to be the current dynamic bottleneck workstation; if the probability value does not reach the threshold, continue to observe and monitor, and do not determine it as a bottleneck workstation.

[0139] B6: Output the identification results of dynamic bottleneck workstations in chart form and feed them back to the production management system.

[0140] The specific steps of B2 include:

[0141] B2.1: Develop a set of identifier rules based on the characteristics of the workstations; the identifier rules adopt a combination of letters and numbers, where letters represent work sections and numbers represent the workstation numbers within those work sections;

[0142] B2.2: Assign a unique identifier to each workstation according to the established identifier rules, and establish a workstation information table; the workstation information table includes the workstation's name, function, and identifier information;

[0143] B2.3: Perform conceptual mapping, treating each workstation as a queuing system, work-in-process as customers, and the processing equipment and operators at the workstation as service counters. At the same time, define the queuing rule for work-in-process in front of the workstation as a first-come, first-served rule.

[0144] B2.4: Collect historical data on the arrival of work-in-process at each workstation during the pipe fitting production process, record the timestamp of the arrival of work-in-process, and calculate the time interval between the arrival of two adjacent work-in-process at each workstation based on the timestamp. ;

[0145] B2.5: Calculation time interval Average value and the average value Find the reciprocal to obtain the arrival rate of each queuing system. ;

[0146] B2.6: Analyze the results of motion waste identification, assess the impact of wasteful motions on the arrival rate of work-in-process, and correct the calculated arrival rate according to the degree of impact. For example, if any motion waste causes work-in-process to stay in the production process for too long, it will affect the rate at which it arrives at the subsequent workstation; repeated handling may cause delays or damage to work-in-process during handling, affecting its timely arrival at the next workstation.

[0147] Furthermore, by analyzing historical data, we can determine how much the rate at which work-in-process reaches the next workstation decreases for every minute the waiting time increases, thus quantifying the impact of wasteful actions on the arrival rate. Based on the quantified impact, we can correct the originally calculated work-in-process arrival rate by taking into account the effects of different wasteful actions, and obtain the corrected arrival rate.

[0148] B2.7: Determine the standard processing time for each station based on the process parameters of the pipe forming process. The analysis of wasted motion identification results assesses the impact of wasted motions on the efficiency of work-in-process handling at workstations, and adjusts the standard processing time based on the degree of impact to obtain the actual processing time. For example, wasted actions may lead to equipment failure, operator error, etc., thereby reducing the processing efficiency of the workstation;

[0149] Furthermore, the specific steps in B2.7 include:

[0150] B2.71: Obtain the process parameters for pipe forming, and calculate the standard processing time based on these parameters. The calculation of the standard processing time varies depending on the specific process of each workstation. For example, for a simple machining workstation, if the machine processes the pipe at a fixed speed v, the processing length is X, and each clamping and unclamping of the pipe requires a fixed amount of time... Standard processing time ;

[0151] B2.72: Analyze the operational actions of operators at workstations to identify which actions are wasteful, such as unnecessary walking, unnecessary waiting, and repetitive operations, and record the frequency and duration of each wasteful action. Among them, unnecessary walking refers to walking other than workstation switching, and unnecessary waiting refers to non-work waiting.

[0152] B2.73: Based on the characteristics and frequency of wasteful actions, analyze their impact on the efficiency of work-in-process handling at the workstation. For example, frequent walking may lead to worker fatigue, thereby reducing work speed; unnecessary waiting will cause equipment to be idle and prolong the overall processing time. Expert evaluation methods can be used to determine the degree of impact of each wasteful action on the efficiency of work-in-process handling at the workstation.

[0153] B2.74: Adjust the standard processing time based on the impact of wasteful actions on the efficiency of work-in-process handling obtained from the assessment;

[0154] If a wasteful action leads to reduced efficiency The actual processing time .

[0155] B2.8: Yes Calculate the mean and then the reciprocal to obtain the service rate of each queuing system. .

[0156] The specific steps of B3 include:

[0157] B3.1: Determine the possible decisions for each action state and load the pre-trained M / M / 1 model. At the same time, the arrival rate will be obtained. Service rate As parameters of the M / M / 1 model, The M / M / 1 model represents the average number of work-in-process items in the system. This model is existing technology in the field and is not an inventive solution of this application. Therefore, it will not be described in detail here.

[0158] For example, whether to increase the number of operators at a certain workstation or adjust the processing speed of the equipment, the decision space includes all possible decision options.

[0159] It should be noted that the M / M / 1 model is a classic and fundamental queuing model in queuing theory. It belongs to the single-server queuing model. In this model, M represents a Markov process, that is, the arrival process of customers follows a Poisson distribution and the service time follows an exponential distribution. 1 indicates that there is only one server in the system.

[0160] B3.2: Set the minimization of the production cycle as the objective function, and derive the state transition equation based on the M / M / 1 model and decision rules;

[0161] Furthermore, the derivation of the state transition equation includes: Let... The current action state, For the current decision, If the next action state is given, then the state transition probability is... Describes the action state Make a decision Then transition to the action state The probability of.

[0162] B3.3: Based on the objective function and the state transition equation, establish the Bellman equation and solve it using the value iteration method. The specific form of the Bellman equation is as follows: , Indicates the state of action Make a decision The immediate cost, min represents the minimum function, and D represents the decision space. and These represent the action states respectively. and action state The optimal value function under the action state, i.e. or action state Initially, identify the optimal long-term cumulative reward or cost that can be obtained by adopting the optimal strategy;

[0163] B3.4: By solving the Bellman equations, the optimal decision, i.e., the optimal policy, is obtained for each action state. This leads to a dynamic bottleneck workstation identification model. The Bellman equation is solved using the value iteration method, which is existing technology in this field and not an inventive solution of this application. Therefore, it will not be elaborated here.

[0164] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of the present invention without departing from the spirit and scope of the present invention. All of these variations are within the protection scope of the present invention.

Claims

1. A real-time monitoring and optimization system for intelligent factory production processes, characterized in that, include: Data acquisition module, motion waste analysis module, bottleneck station identification module, production cycle adjustment module; The motion waste analysis module constructs a motion waste analysis model based on Markov chain theory according to the motion sequence data collected by the data acquisition module. By calculating the state transition probability and the steady-state distribution of the Markov chain, it identifies motion waste situations. The bottleneck station identification module, based on the received action waste identification results and combined with the pipe forming process parameters collected by the data acquisition module, constructs a dynamic bottleneck station identification model to quantitatively evaluate the probability value of each station evolving into a bottleneck station at different time points. The production cycle adjustment module adjusts the production cycle based on the identification results of the dynamic bottleneck station; The motion waste analysis module constructs a motion waste analysis model based on Markov chain theory according to the motion sequence data, and identifies motion waste situations, including: A1: Obtain the action sequence data during the pipe fitting production process and perform preprocessing; A2: Based on the pipe fitting production process, a unique discrete state code is defined for each action, and a mapping rule from action to discrete state is formulated. The mapping is performed according to the name of the action. By traversing the preprocessed action sequence data, each action is converted into the corresponding discrete state code according to the mapping rule. A3: Analyze the discrete action sequence data, classify the actions into different categories according to their type, and combine K actions into an action state based on the execution order of the actions; the action state includes raw material classification and screening, processing technology, quality inspection, and finished product classification and handling; A4: After removing duplicate action states from all the action states obtained in A3, summarize them and form the state space of the Markov chain. ,in, This represents the nth action state, where n represents the number of action states. A5: Starting from the first action state in the preprocessed action sequence data, check adjacent action states sequentially. and ; If two adjacent action states are different, then a state transition has occurred, that is, from one action state... Transition to another action state ; A6: The transition frequency is obtained by counting the number of state transitions in the preprocessed action sequence data using an initialized counter. ; A7: By calculating the transfer frequency With From Transfer to all non The ratio of the total number of times the action state is used to obtain the state transition probability. ; A8: Create The matrix, with the state transition probabilities calculated in A7. Assigning values ​​to the matrix yields the action transition matrix. ; The bottleneck station identification module, based on the received action waste identification results and combined with the pipe forming process parameters collected by the data acquisition module, constructs a dynamic bottleneck station identification model to quantitatively evaluate the probability of each station evolving into a bottleneck station at different time points, including: B1: Acquire the results of motion waste identification and the process parameters of the pipe forming process. At the same time, monitor the amount of work-in-process and production efficiency at each station in real time. The results of motion waste identification include wasted motions and the degree of waste. B2: Assign a unique identifier to each workstation, treat each workstation as a queuing system, and treat work-in-process as a customer. Based on the process parameters of the pipe forming process and the results of motion waste identification, determine the arrival rate and service rate of each queuing system; the arrival rate is the rate at which work-in-process arrives at the workstation; the service rate is the rate at which the workstation processes work-in-process. B3: Combining the M / M / 1 model, a dynamic bottleneck workstation identification model is established using dynamic programming methods; B4: Integrate the acquired motion waste identification results, pipe forming process parameters, and real-time monitoring of the work-in-process accumulation quantity and production efficiency of each station to form pipe data. Input the pipe data into the dynamic bottleneck station identification model and use the dynamic bottleneck station identification model to calculate the probability value of each station evolving into a bottleneck station at different time points in real time.

2. The intelligent factory production process real-time monitoring and optimization system as described in claim 1, characterized in that, The motion waste analysis module, based on the motion sequence data and Markov chain theory, constructs a motion waste analysis model to identify motion waste situations, and further includes: A9: Based on the properties of steady-state distribution, establish the Markov chain equilibrium equations, and solve the Markov chain equilibrium equations using a system of linear equations to obtain the steady-state distribution vector. ; A10: Based on the steady-state distribution vector Determine the frequency of occurrence of each action state in the actual production process. The frequency of occurrence of each action state is the steady-state probability of each action state. ,and ; A11: Based on the process requirements for pipe fitting production, a standard frequency range is set for each operating state. At the same time, a preset standard threshold is set for each action state. ,in, and Let represent the minimum and maximum values ​​of the standard frequency of the i-th action state, respectively; A12: For each action state , action state Frequency of occurrence With standard frequency range Compare; like ,and Then determine the action state. There is wasted action; A13: Feed back the results of motion waste identification to the production management system through the data interface.

3. The intelligent factory production process real-time monitoring and optimization system as described in claim 2, characterized in that, The Markov chain equilibrium equations simultaneously satisfy: The system transitions from all previous action states to the action state. The sum of the weighted state transition probabilities equals the system's state of action. The steady-state probability; The sum of the steady-state probabilities of the system in all action states is 1.

4. The intelligent factory production process real-time monitoring and optimization system as described in claim 3, characterized in that, The specific steps of A10 include: A10.1: Obtaining the steady-state distribution vector The state space of Markov chains ; A10.2: Traversal The index of each action state, from the steady-state distribution vector. Extract the element value at the corresponding position, and the element value is the steady-state probability of the action state.

5. The intelligent factory production process real-time monitoring and optimization system as described in claim 4, characterized in that, The bottleneck station identification module, based on the received action waste identification results and combined with the pipe forming process parameters collected by the data acquisition module, constructs a dynamic bottleneck station identification model to quantitatively evaluate the probability of each station evolving into a bottleneck station at different time points. It also includes: B5: Real-time monitoring of the work-in-process inventory and production efficiency at each workstation, and setting a threshold for the work-in-process inventory. If the amount of work-in-process at any workstation exceeds the work-in-process quantity threshold and its production efficiency is lower than the demand efficiency of the downstream workstation, the workstation is determined to be the current dynamic bottleneck workstation by combining the probability value calculated by the dynamic bottleneck workstation identification model. B6: Output the identification results of dynamic bottleneck workstations in chart form and feed them back to the production management system.

6. The intelligent factory production process real-time monitoring and optimization system as described in claim 5, characterized in that, The specific steps of B2 include: B2.1: Identifier rules shall be formulated according to the characteristics of the workstations; the identifier rules shall adopt a combination of letters and numbers, where letters represent work sections and numbers represent the workstation numbers within the work sections; B2.2: Assign a unique identifier to each workstation according to the established identifier rules, and establish a workstation information table; the workstation information table includes the workstation's name, function, and identifier information; B2.3: Perform conceptual mapping, treating each workstation as a queuing system, work-in-process as customers, and the processing equipment and operators at the workstation as service counters. At the same time, define the queuing rule for work-in-process in front of the workstation as a first-come, first-served rule. B2.4: Collect historical data on the arrival of work-in-process at each workstation during the pipe fitting production process, record the timestamp of the arrival of work-in-process, and calculate the time interval between the arrival of two adjacent work-in-process at each workstation based on the timestamp. ; B2.5: Calculation time interval Average value and the average value Find the reciprocal to obtain the arrival rate of each queuing system. ; B2.6: Analyze the results of waste motion identification, assess the impact of wasteful motions on the arrival rate of work-in-process, and correct the calculated arrival rate according to the degree of impact; B2.7: Determine the standard processing time for each station based on the process parameters of the pipe forming process. The analysis of wasted motion identification results assesses the impact of wasted motions on the efficiency of work-in-process handling at workstations, and adjusts the standard processing time based on the degree of impact to obtain the actual processing time. ; B2.8: Yes Find the reciprocal to obtain the service rate of each queuing system. .

7. The intelligent factory production process real-time monitoring and optimization system as described in claim 6, characterized in that, The specific steps of B3 include: B3.1: Determine the decision to be made in each action state, load the pre-trained M / M / 1 model, and simultaneously obtain the arrival rate. Service rate As parameters for the M / M / 1 model; B3.2: Set the minimization of the production cycle as the objective function, and derive the state transition equation based on the M / M / 1 model and decision rules; B3.3: Based on the objective function and the state transition equation, establish the Bellman equation and solve the Bellman equation using the value iteration method; B3.4: By solving the Bellman equations, the optimal decision, i.e., the optimal policy, is obtained for each action state. A dynamic bottleneck workstation identification model was obtained.

Images