Method for optimizing multi-target maintenance strategy of infrared moisture meter of tobacco primary processing line and storage medium

Abstract

This invention discloses a method and storage medium for optimizing multi-objective maintenance strategies of infrared moisture analyzers in silk-making lines. The method includes: step S1, acquiring the status x of the infrared moisture analyzer. i Step S2, based on state x i Determine priorities; based on the proportion of infrared moisture meters requiring maintenance to the total number of infrared moisture meters and the consequences of not maintaining them, determine the completion time and cost of maintenance tasks; Step S3, determine the comprehensive analysis index of the infrared moisture meters based on priorities, completion time, and cost; Step S4, determine the status x of the infrared moisture meters. i A Markov decision model is constructed using a comprehensive analysis index; step S5 involves generating the optimal maintenance strategy based on the Markov decision model. This method, tailored to the specific needs and characteristics of moisture meters, can improve equipment operating efficiency, reduce failure rate, and extend equipment lifespan; simultaneously, by incorporating maintenance time factors, the optimal maintenance time window is determined to minimize disruption to the production process.

Description

Technical Field

[0001] This invention relates to the field of cigarette manufacturing technology, specifically to a method for optimizing multi-objective maintenance strategies for infrared moisture analyzers used in cigarette manufacturing lines and a storage medium thereon. Background Technology

[0002] In cigarette manufacturing, an infrared moisture meter is a crucial piece of equipment used to measure the moisture content of tobacco leaves, ensuring the flavor and filling properties of the finished cigarette product. Therefore, its maintenance strategy must ensure the infrared moisture meter operates normally, including accuracy, reliability, and stability, to guarantee precise control of the tobacco moisture content during cigarette production.

[0003] Current methods primarily rely on statistical or empirical formulas to optimize maintenance strategies, lacking a comprehensive consideration of the reliability analysis of infrared moisture meters. Furthermore, optimization is based on equipment maintenance needs, with little consideration for integration with production planning. However, in cigarette manufacturing, infrared moisture meters are critical equipment, and their maintenance strategies need to be coordinated with production plans to ensure maintenance is performed within the optimal time window, minimizing disruption to the production process.

[0004] Therefore, a maintenance strategy is needed to address the specific needs and characteristics of infrared moisture meters. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a multi-objective maintenance strategy optimization method and storage medium for infrared moisture meters used in silk refining lines, specifically tailored to the needs and characteristics of infrared moisture meters. This method is more targeted, can improve equipment operating efficiency, reduce failure rate, and extend equipment lifespan.

[0006] According to a first aspect, the present invention provides a method for optimizing a multi-objective maintenance strategy for an infrared moisture analyzer used in silk production lines, comprising:

[0007] Step S1: Obtain the status x of the infrared moisture meter i Its status includes: design feature status, operating condition status, degradation feature status, and maintenance consequences status.

[0008] Step S2: Determine the priority based on the design feature state, operating condition state, or degradation feature state;

[0009] The maintenance task completion time and cost are determined based on the proportion of infrared moisture meters requiring maintenance to the total number of infrared moisture meters and the consequences of not maintaining them.

[0010] Step S3: Determine the comprehensive analysis index of the infrared moisture meter based on priority, maintenance task completion time, and maintenance task completion cost;

[0011] Step S4: Construct a Markov decision model based on the design characteristics, operating conditions, degradation characteristics, consequences of lack of maintenance, and comprehensive analysis index of the infrared moisture meter. The Markov decision model includes a state space, an action space, and a reward function.

[0012] Step S5: Generate the optimal maintenance strategy based on the Markov decision model.

[0013] Furthermore, in step S2, for the design characteristic state, operating condition state, and degenerate characteristic state, the priority P is output according to the following rules:

[0014] If x i_min If the value is ≤0.85, then output the first priority P1;

[0015] If 0.85 <x i_min If the value is ≤0.9, then output the second priority P2;

[0016] If 0.9 <x i_min If the value is ≤0.95, then output the third priority P3;

[0017] If 0.95 <x i_min If so, then output the fourth priority P4;

[0018] x i_min The status of the infrared moisture meter x i The smallest value in.

[0019] Furthermore, the design feature states include:

[0020] Moisture meter filter motor speed design features x1 and zero point offset x2;

[0021] Among them, x1 is a design feature used to describe the filter capability of the moisture meter. By checking whether the rotation speed reaches a uniform standard, the stability of the motor speed and the stability of the filter performance can be evaluated.

[0022]

[0023] Discrete(1000) is used to define the discrete space, which ranges from 7000 to 8000.

[0024] Zero offset x2 is a parameter used to describe the magnitude of the zero offset of a moisture meter. Zero offset is a form of damage caused by repeated loading, light source attenuation, aging of electronic components, random factors, etc. This parameter can be used to evaluate the stability of moisture meter measurements.

[0025]

[0026] x2 has a value range between 0 and 1 and is a floating-point number.

[0027] The operating condition states include:

[0028] This month's correction value offset rate for each grade x3 and the oven offset for the same month x4;

[0029] Among them, x3 is a parameter used to measure the variation of a moisture meter with different brands within a 1-month period.

[0030] It can be used to evaluate the repeatability and stability of moisture meter measurements.

[0031]

[0032] x3 has a value range between 0 and 1, and is a floating-point number.

[0033] x4 is a parameter used to describe the change of a certain grade in a certain month within a certain period. By comparing the ratio of the change of a certain grade in a certain month to the change in the current month within a certain period, the performance degradation of the moisture meter can be assessed.

[0034]

[0035] x4 has a value between 0 and 1 and is a floating-point number.

[0036] The degradation characteristic states include:

[0037] The deviation of the center value of the moisture meter motor speed from the standard value x5; the deviation of the center value of the moisture meter light source illuminance from the standard value x6;

[0038] x5 is an indicator used to measure the degradation of the motor speed of a moisture meter over a one-month period. It can be used to assess the degradation characteristics of the moisture meter motor performance.

[0039]

[0040] x5 has a value range between 0 and 1, and is a floating-point number.

[0041] x6 is an indicator used to measure the degradation of the light source of a moisture meter over a one-month period. It can be used to assess the degradation characteristics of the moisture meter's light source.

[0042]

[0043] x6 has a value range between 0 and 1, and is a floating-point number.

[0044] The consequences of not maintaining the equipment include:

[0045] The quality risks caused by failure to perform maintenance are x7, the economic risks caused by failure to perform maintenance are x8, and the human resources required for maintenance are x9.

[0046] The value range of x7 is between 1 and 3, and it is an integer. 3 represents high, 2 represents medium, and 1 represents low.

[0047] The value range of x8 is between 1 and 3, and it is an integer. 3 represents high, 2 represents medium, and 1 represents low.

[0048] The value range of x9 is between 1 and 3, and it is an integer. 3 represents senior staff, 2 represents intermediate staff, and 1 represents junior staff.

[0049] Furthermore, in the step S2, the completion time of the maintenance task is determined according to the following rules:

[0050] If R ≤ 0.2, then output T = 2 / x9;

[0051] If 0.2 < R ≤ 0.4, then output T = 3 / x9;

[0052] If 0.4 < R ≤ 0.6, then output T = 4 / x9;

[0053] If 0.6 < R, then output T = 5 / x9;

[0054] Where, R is the proportion of the infrared moisture meters to be maintained in the total infrared moisture meters, and T is the completion time of the maintenance task.

[0055] Furthermore, in the step S2, the maintenance task completion cost C = C T + C R ;

[0056] Where, C T is the time cost, and C R is the labor cost;

[0057] If R ≤ 0.2, then C T = 2 * A; C R = B1 * X9;

[0058] If 0.2 < R ≤ 0.4, then C T = 3 * A; C R = B2 * X9;

[0059] If 0.4 < R ≤ 0.6, then C T = 4 * A; C R = B3 * X9;

[0060] If 0.6 < R, then C T = 5 * A; C R = B4 * X9;

[0061] Where, A is the cost per time unit, and B1 to B4 are the costs of human resource units corresponding to different proportions R.

[0062] Furthermore, the comprehensive analytical index F of the infrared moisture meter is:

[0063]

[0064] Where P is the priority of the task, T is the maintenance task completion time, and C is the maintenance task completion cost. F takes an integer value between 0 and 1000.

[0065] Furthermore, the action space (actions) is:

[0066] actions=[action_1,action_2,action_3,action_4,action_5];

[0067] Among them, action_1: check the speed of the filter motor of the moisture meter, which may include operations such as cleaning, calibration or replacement of the motor;

[0068] action_2: Adjust the zero point offset of the moisture meter, which may involve resetting the zero point, repairing sensor offset, or performing calibration.

[0069] action_3: Correcting parameters for different grades may involve adjusting measurement parameters, updating calibration curves, or performing calibration based on grade changes;

[0070] action_4: Perform light source maintenance operations, which may include cleaning, servicing, replacing or adjusting the light source, etc.

[0071] action_5: Doing nothing may cause the moisture meter's performance to randomly decrease or have no effect whatsoever.

[0072] The design principles of the reward space include:

[0073] Performance stability: Based on the various state characteristics of the moisture meter (such as motor speed, zero-point offset, correction rate, etc.), positive rewards can be considered when these characteristics approach their target values ​​or are within acceptable ranges. For example, a decay function can be designed so that the smaller the difference between the state characteristics and the target value, the higher the reward.

[0074] Maintenance Costs: Considering economic risks and human resource requirements, negative rewards can be given based on the costs and resource consumption of maintenance. If maintenance costs are high, the agent can be penalized for choosing actions that require significant resources or incur high costs.

[0075] Risk assessment: Based on the assessment of quality and economic risks, corresponding rewards or penalties can be given. For example, if the quality risk is high, a negative reward can be given to encourage the agent to take action to reduce the risk.

[0076] Action selection: For action_5 (no maintenance), a positive reward can be given because this can significantly reduce costs. However, choosing various maintenance methods will increase costs, causing the agent to tend to choose maintenance.

[0077] Furthermore, the reward function R total for:

[0078] R total =R perfomance +R maintenance_cost+Rquality_risk +R action

[0079] Among them, R perfomance As a performance stability bonus; R maintenance_cost As a maintenance cost incentive; R quality_rike Rewards for risk assessment; R action Action selection reward.

[0080]

[0081] R maintenance_cost =weight(′maintenance_cost′)·action

[0082] R quality_risk =weight('quality'_risk)·s('quality'_risk)

[0083]

[0084] Where s is the state space; target_value(s) is the target value; weight(·) is the weight; action is the action space; decay(·) is the decay function; and s('quality'_risk) is the value of x7 of the quality risk caused by the lack of maintenance.

[0085] Furthermore, in step S5, the optimal maintenance strategy is generated by combining random search with a fixed strategy.

[0086] Among them, random search: at each time step, an action is randomly selected from the action space with equal probability;

[0087] Fixed strategy: When a specific condition is met, the corresponding action must be performed.

[0088] For example: when x1 satisfies the first priority P1, action_1 is executed;

[0089] If x2 satisfies the first priority P1, then action_2 is executed;

[0090] When x3 satisfies the first priority P1, action_3 is executed;

[0091] When x4 satisfies the first priority P1, action_4 is executed;

[0092] When x5 satisfies the first priority P1, action_5 is executed.

[0093] Alternatively, the optimal action can be obtained using a value iteration method. This involves first generating an action value table (number of states * number of actions), then finding the optimal value function, and finally deriving the optimal strategy (the one with the highest value in the action value table). The implementation is as follows:

[0094] Number of loop iterations:

[0095] Loop through each state into the state list:

[0096] Loop through each action in the action list:

[0097] Next state = Real environment(state, action) # Get the next state Immediate reward = Reward function(state, action) # Get the immediate reward

[0098] #Update action value function

[0099] If the state is not in the action value table:

[0100] Action Value Table [State] = {}

[0101] If the action is not in the action value table:

[0102] Action Value Table [State][Action] = 0

[0103] Action Value Table [State][Action] = Action Value Table [State][Action] + Learning Rate * (Instant Reward + Discount Factor * max(Action Value Table [Next State].values()) - Action Value Table [State][Action])

[0104] The number of iterations is usually chosen to be 10,000, and the actual environment is determined by the state of the moisture meter.

[0105] As the number of iterations increases, if other changes occur in the state space, this invention can automatically monitor the changes in the state space and selectively perform maintenance actions, thereby automatically forming a maintenance strategy for the entire lifecycle.

[0106] According to a second aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, the computer program being executable by a processor to implement the steps of the method described above.

[0107] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0108] (1) This invention provides a maintenance strategy specifically for infrared moisture meters, which is more targeted, can improve the operating efficiency of the equipment, reduce the failure rate, and increase the service life of the equipment.

[0109] (2) The present invention incorporates the maintenance time factor to determine the optimal maintenance time window in order to minimize the interference with the production process.

[0110] (3) Through in-depth fault diagnosis and theoretical analysis, equipment faults can be identified and resolved more accurately. This not only avoids the costs of repeated repairs and parts replacements, but also prevents production delays caused by equipment failures, thereby improving production efficiency. Attached Figure Description

[0111] Figure 1 This is a flowchart of the method of the present invention;

[0112] Figure 2 This is a block diagram of the multi-objective maintenance strategy optimization system for the infrared moisture analyzer in the silk-making line in Example 1;

[0113] Figure 3 This is a diagram showing the action value table in Example 2. Detailed Implementation

[0114] The present invention will be further described in detail below through specific embodiments.

[0115] Example 1

[0116] like Figure 1-2 As shown, this embodiment provides a multi-objective maintenance strategy optimization system for an infrared moisture meter in a silk refining line, which includes: a moisture meter status monitoring module 1, a moisture meter status analysis module 2, and a strategy generation and optimization module 3. The moisture meter status monitoring module 1 includes: a design feature status monitoring module 11, an operating condition status monitoring module 12, a degradation feature status monitoring module 13, and a non-maintenance consequences status monitoring module 14. The design feature status monitoring module 11 is used to acquire the design feature x1 of the moisture meter filter motor speed and the zero-point offset x2; the operating condition status monitoring module 12 is used to acquire the monthly grade correction value offset rate x3 and the same month oven offset x4; the degradation feature status monitoring module 13 is used to acquire the offset x5 of the moisture meter motor speed center value from the standard value and the offset x6 of the moisture meter light source illuminance center value from the standard value; the non-maintenance consequences status monitoring module 14 is used to acquire the quality risk x7 caused by non-maintenance, the economic risk x8 caused by non-maintenance, and the human resources required for maintenance x9.

[0117] The moisture analyzer status analysis module 2 includes: a priority analysis module 21, a completion time calculation module 22, a completion cost calculation module 23, and a comprehensive analysis module 24. The priority analysis module 21 outputs a priority P according to the following rules:

[0118] If x i_min ≤ 0.85, then output the first priority P1;

[0119] If 0.85 < x i_min ≤ 0.9, then output the second priority P2;

[0120] If 0.9 < x i_min ≤ 0.95, then output the third priority P3;

[0121] If 0.95 < x i_min , then output the fourth priority P4;

[0122] x i_min is the minimum value in the status x of the infrared moisture analyzer i among them.

[0123] The completion time calculation module 22 determines the maintenance task completion time T according to the following rules:

[0124] If R ≤ 0.2, then output T = 2 / x9;

[0125] If 0.2 < R ≤ 0.4, then output T = 3 / x9;

[0126] If 0.4 < R ≤ 0.6, then output T = 4 / x9;

[0127] If 0.6 < R, then output T = 5 / x9;

[0128] Among them, R is the proportion of the infrared moisture analyzers to be maintained in the total infrared moisture analyzers.

[0129] The completion cost calculation module 23 determines the maintenance task completion cost C in the following manner:

[0130] C = C T + C R ;

[0131] Among them, C T is the time cost, and C R is the labor cost;

[0132] If R ≤ 0.2, then C T = 2 * A; C R = B1 * X9;

[0133] If 0.2 < R ≤ 0.4, then C T = 3 * A; C R= B2 * X9;

[0134] If 0.4 < R ≤ 0.6, then C T = 4 * A; C R = B3 * X9;

[0135] If 0.6 < R, then C T = 5 * A; C R = B4 * X9;

[0136] Where, A is the cost per time unit, and B1 to B4 are the costs of the human resource units corresponding to different ratios R. Preferably, A = 10, B1 = 10, B2 = 15; B3 = 20; B4 = 25.

[0137] The comprehensive analysis module 24 calculates the comprehensive analysis index F in the following way:

[0138]

[0139] Where, P is the priority of the task, T is the maintenance task completion time, and C is the maintenance task completion cost. The value range of F is between 0 and 1000, and it is an integer.

[0140] The policy generation and optimization module 3 includes: a state space module 31, an action space module 32, and a reward space module 33.

[0141] The state space module 31 mainly uses the data (x1 - x9 and F) output by the moisture meter status monitoring module 1 as input. After appropriate processing, it is used as the discretized state space s, and the state space s consists of x1 - x9 and F.

[0142] The action space module 32 has an internal action list actions = [action_1, action_2, action_3, action_4, action_5], where each element represents an optional action.

[0143] action_1: Repair the rotation speed of the moisture meter filter motor, which may include operations such as cleaning, calibration, or replacement of the motor.

[0144] action_2: Adjust the zero offset of the moisture meter, which may involve operations such as resetting the zero point, repairing the sensor offset, or calibration.

[0145] action_3: Correct the parameters of different grades, which may involve operations such as adjusting the measurement parameters according to grade changes, updating the calibration curve, or calibration.

[0146] action_4: Perform light source maintenance operations, which may include operations such as cleaning, maintenance, replacement, or adjustment of the light source, etc.

[0147] action_5: Doing nothing may cause the moisture meter's performance to randomly decrease or have no effect whatsoever.

[0148] Reward space module 33 is used to calculate rewards, specifically:

[0149] R total =R perfomance +R maintenance_cost +R quality_risk +R action

[0150] Among them, R total R is the reward function; perfomance As a performance stability bonus; R maintenance_cost As a maintenance cost incentive; R quality_risk Rewards for risk assessment; R action Action selection reward.

[0151]

[0152] R maintenance_cost =weight(′maintenance_cost′)·action

[0153] R quality_risk =weight('quality'_risk)·s('quality'_risk)

[0154]

[0155] Where s is the state space; tar get_value(s) is the target value; weight(·) is the weight; action is the action space; and decay(·) is the decay function. s('quality'_risk) is the value of x7 of the quality risk caused by the lack of maintenance.

[0156] The system works as follows:

[0157] Step S1: Obtain the design feature x1 of the moisture meter filter motor speed and the zero-point offset x2 through the design feature status monitoring module 11; obtain the monthly correction value offset rate x3 and the oven offset x4 through the operation condition status monitoring module 12; obtain the offset x5 of the moisture meter motor speed center value from the standard value and the offset x6 of the moisture meter light source illuminance center value from the standard value through the degradation feature status monitoring module 13; obtain the quality risk x7 caused by not performing maintenance, the economic risk x8 caused by not performing maintenance, and the human resources required for maintenance x9 through the non-maintenance consequences status monitoring module 14.

[0158] Step S2: Priority analysis module 21 determines priority P based on design feature state, operating condition state, or degradation feature state;

[0159] The completion time calculation module 22 determines the maintenance task completion time T based on the proportion R of the infrared moisture meters that need maintenance to the total number of infrared moisture meters and the consequences of not maintaining them.

[0160] The cost calculation module 23 calculates the maintenance task completion cost C based on the proportion R of the infrared moisture meters that need maintenance to the total number of infrared moisture meters and human resources x9.

[0161] Step S3: The comprehensive analysis module 24 determines the comprehensive analysis index F of the infrared moisture meter based on the priority P, the maintenance task completion time T, and the maintenance task completion cost C;

[0162] Step S4: The moisture meter maintenance strategy generation module 3 constructs a Markov decision model based on the infrared moisture meter's design characteristic state, operating condition state, degradation characteristic state, consequences of non-maintenance state (x1-x9), and comprehensive analysis index F. The Markov decision model includes a state space s, an action space a, and a reward function R. total ;

[0163] Step S5: The moisture meter maintenance strategy generation module 3 generates the optimal maintenance strategy based on a Markov decision model. The moisture meter maintenance strategy generation module 3 generates the strategy using a value iteration method, that is: first, a value table of actions (number of state spaces * number of action spaces) is pre-generated; starting from a random value function, a new (improved) value function is found during the iteration process until the optimal value function is reached, thus deriving the optimal strategy.

[0164] Example 2

[0165] There are 23 infrared moisture analyzers on the production line of a cigarette factory's tobacco processing workshop. Let's take one of them as an example.

[0166] Its design features a rated speed of 7250 rpm, and at a certain moment its speed is 7240 rpm, then x1 = 0.998.

[0167] At a certain moment, upon inspection, its calibrated zero point is -0.5, and the current zero point is -0.46, then x2 = 0.92.

[0168] If the moisture content of a moisture meter varies by 0.3 for all brands in a certain month, and the moisture content of a certain brand is 0.2, then x3 = 0.667.

[0169] Over a 5-year period, the average monthly change for this brand is 0.28, and the change this month is 0.2, so x4 = 0.71.

[0170] The standard value for the motor speed of the moisture meter is 7250 rpm. The motor speed is 7241 rpm within a one-month period, so x5 = 0.998.

[0171] The illuminance of the light source is generally 300 lx, and the current measured illuminance is 270 lx, so x6 = 0.9.

[0172] The quality risk caused by not performing maintenance is x7 = 3; the economic risk caused by not performing maintenance is x8 = 3; currently, one intermediate-level personnel is needed for maintenance, so x9 = 2.

[0173] First priority P1 = 4; second priority P2 = 3; third priority P3 = 2; fourth priority P4 = 1. Because x i _ min =x3=0.667, therefore the first priority output is P1=4.

[0174] Currently, one moisture meter needs maintenance, accounting for 0.04% of the 23 meters in the entire silk production line, with an expected completion time of 1 minute.

[0175] In this embodiment, assuming that the cost A for each time unit is 10, and the cost B1 for the human resource unit corresponding to R = 0.04 is 10, then the completion cost C is the time cost 20 plus the human resource cost 20, C = 40.

[0176] The comprehensive analysis index F = 600 can be calculated.

[0177] Taking the current state of this embodiment as an example, if action_1 is selected, it will incur a motor replacement cost of 20 and a mid-level personnel cost of 20. After maintenance, its performance will decrease at a rate of 0.8% per year. For ease of explanation, the weight (weight of each indicator) is 1. Therefore:

[0178] R perfomance =100; R maintenance_cosr =-40; R quality_risk =10; R action =0

[0179] Total reward R total =70.

[0180] This embodiment uses an action value table for strategy generation and action selection. Therefore, an action value table (number of state spaces * number of action spaces) is pre-generated. In this invention, the number of state spaces is 9, but all are continuous variables, such as the design feature of the motor speed of the moisture meter filter (x1). Therefore, discretization is required, using a value of 1000 for each variable. This results in a three-dimensional action value table of (9 * 1000 * 5), with the form Value = V(D(s, 1000), a), where D represents uniform discretization. For example, the range of motor speed from 7000 to 8000 is uniformly discretized into (7000, 7001, 7002….8000). The form of the action value table is as follows: Figure 3 As shown.

[0181] Taking the current state of this embodiment as an example, after considering factors such as algorithm calculation cost, time, and personnel, it is decided to select action 1, that is, to replace the motor.

[0182] The above strategy was optimized and adjusted for each infrared moisture meter to form the optimal strategy for 23 infrared moisture meters. The optimal strategy formed by a series of optimized actions is [action_1, action_5, ..., action_2, action_5].

[0183] As time progresses, if other changes occur in the state space, this invention can automatically monitor these changes and selectively perform maintenance actions, thereby automatically forming a full lifecycle maintenance strategy. Compared to the invention's description, where the state space of the real environment is a computer simulation, in the actual embodiment, the real environment is the moisture meter itself. Therefore, the iteration in the invention's description is simply the flow of time in the embodiment.

[0184] Example 3

[0185] Table 1 shows the maintenance status of 22 moisture meters in the tobacco processing workshop of a cigarette factory during a certain period. As can be seen from Table 1, the method of the present invention optimized the maintenance strategy and executed it 1320 times (5 years * 12 months * 22 units). Among them, action_5 did not perform any operation 1312 times, action_4 replaced the light bulb 4 times, and action_1 replaced the motor 4 times. Among them, the two instances of action_4 replacing the light bulb were adjusted to be performed during the next month's maintenance under the strategy optimization.

[0186] The lifespan of each vulnerable part of a typical moisture meter is 50,000 hours. Due to some differences, the general strategy is to replace vulnerable parts every 5 years. Based on this calculation, the total cost of spare parts is 22 units * 15,000 = 330,000 yuan, and the labor time required is 22 units * (1 + 1.5) hours = 55 hours.

[0187] After adopting the method of this invention, the total cost was 56,000 yuan and the labor time was 10 hours. Under the premise of ensuring that production and process testing stability are not affected, the cost was reduced by 274,000 yuan and the labor time was reduced by 45 hours, realizing the state of maintenance on demand and planned maintenance.

[0188] Since the two light bulb replacement actions (action_4) have been adjusted to be performed during the next month's maintenance under the strategy optimization, the cost of a single maintenance has been reduced from 5,000 yuan to 3,000 yuan in terms of spare parts turnover and capital utilization.

[0189] Table 1

[0190]

[0191]

[0192]

[0193] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. A method for optimizing multi-objective maintenance strategies for infrared moisture analyzers used in silk refining lines, characterized in that, Including: Step S1: Obtain the status of the infrared moisture meter, and its status includes: design feature status, operating condition status, degradation feature status, and consequence status of non-maintenance. Step S2: Determine the priority according to the design feature status, operating condition status, and degradation feature status. Determine the completion time and cost of the maintenance task according to the proportion of infrared moisture meters that need to be maintained in the total infrared moisture meters and the consequence status of non-maintenance. Step S3: Determine the comprehensive analysis index of the infrared moisture meter according to the priority, completion time of the maintenance task, and cost of the maintenance task. Step S4: Construct a Markov decision model based on the design feature status, operating condition status, degradation feature status, consequence status of non-maintenance, and comprehensive analysis index of the infrared moisture meter. The Markov decision model includes a state space, an action space, and a reward function. Step S5: Generate an optimal maintenance strategy based on the Markov decision model. The status of the infrared moisture meter is indicated by x. i express; The design feature status includes: the design feature x1 of the filter light motor speed of the moisture meter and the zero-point offset x2. The operating condition status includes: the correction value offset rate x3 of each grade this month and the oven offset x4 in the same month. The degradation feature status includes: the offset amount x5 of the center value of the moisture meter motor speed from the standard value and the offset amount x6 of the center value of the light source illuminance of the moisture meter from the standard value. The consequence status of non-maintenance includes: the quality risk x7 caused by non-maintenance, the economic risk x8 caused by non-maintenance, and the human resources x9 required for maintenance.

2. The method according to claim 1, wherein ； ； ； ； ； 。 3. The method as described in claim 2, characterized in that, In step S2, for the design feature status, operating condition status, and degradation feature status, output the priority P according to the following rules: If x i _ min If the value is ≤0.85, then output the first priority P1; If 0.85 <x i _ min If the value is ≤0.9, then output the second priority P2; If 0.9 <x i _ min If the value is ≤0.95, then output the third priority P3; If 0.95 <x i _ min If so, then output the fourth priority P4; x i _ min The status of the infrared moisture meter x i The smallest value in.

4. The method as described in claim 2, characterized in that, In step S2, the completion time of the maintenance task is determined according to the following rules: If R ≤ 0.2, then output T = 2 / x9; If 0.2 < R ≤ 0.4, then output T = 3 / x9; If 0.4 < R ≤ 0.6, then output T = 4 / x9; If 0.6 < R, then output T = 5 / x9; Where R is the proportion of infrared moisture meters that need to be maintained in the total infrared moisture meters, and T is the completion time of the maintenance task.

5. The method as described in claim 4, characterized in that, In step S2, the maintenance task completion cost C = C T +C R ; Among them, C T For time cost, C R For labor costs; If R ≤ 0.2, then C T =2*A, C R =B1*X9; If 0.2 < R ≤ 0.4, then C T = 3 * A, C R = B2 * X9; If 0.4 < R ≤ 0.6, then C T = 4 * A, C R = B3 * X9; If 0.6 < R, then C T = 5 * A, C R = B4 * X9; Where A is the cost per time unit, and B1 to B4 are the costs per unit of human resources corresponding to different proportions of R.

6. The method as described in claim 1, characterized in that, In step S3, the comprehensive analysis index F of the infrared moisture meter is: ； Where P is the priority of this task, T is the completion time of the maintenance task, C is the cost of the maintenance task, and N is the total number of infrared moisture meters that need to be maintained.

7. The method as described in claim 1, characterized in that, The action space is represented as actions: actions = [action_1, action_2, action_3, action_4, action_5]; Where action_1 is to repair the filter light motor speed of the moisture meter; action_2 is to adjust the zero-point offset of the moisture meter; action_3 is to correct the parameters of different grades; action_4 is to perform the light source maintenance operation; action_5 is to perform no operation.

8. The method as described in claim 5, characterized in that, The reward function is represented as R. total : ； in, As a reward for performance stability; As a maintenance cost incentive; Rewards for risk assessment; Action selection reward.

9. The method according to claim 8, wherein ； ； ； ； in, For state space; The target value; As weight, For the action space, that is , It is a decay function; This represents the value of the quality risk caused by the lack of maintenance. This refers to maintenance costs, specifically the cost of completing the maintenance task, C. This refers to quality risk.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program can be executed by a processor to implement the steps of the method as described in any one of claims 1-9.

Images