Optimization method for selective maintenance decision of manufacturing system considering operational robustness

Abstract

The application provides a manufacturing system selective maintenance decision optimization method considering operation robustness, and the steps are as follows: 1, collecting and analyzing the data of the manufacturing system after a task cycle of the manufacturing system ends; 2, determining the performance state of the manufacturing system, the deviation of the product quality characteristics and the product quality; 3, determining the current robustness level of the manufacturing system; 4, collecting the data about the machine maintenance, the corresponding relationship of the maintenance cost, the maintenance time and the maintenance quality of each machine; 5, establishing a selective maintenance decision model with the maximum robustness of the next task stage as the target and the maintenance time and the cost as the constraint conditions; 6, establishing an ASA-PSO algorithm model to obtain the best maintenance strategy. The application considers the relationship between the system robustness and the machine performance decline, the task delay and the product quality non-compliance risk, analyzes the function risk and the mechanism of the manufacturing system, optimizes the use of the maintenance resources and realizes a better maintenance plan.

Description

Technical Field

[0001] This invention provides a selective maintenance decision optimization method for manufacturing systems that takes into account operational robustness, belonging to the field of system state assessment and maintenance decision-making. Background Technology

[0002] In the rapidly evolving manufacturing industry, ensuring the reliability of multi-stage manufacturing systems is crucial for producing consistently high-quality products that meet requirements. These complex systems typically consist of multiple workstations configured in series and parallel. In such systems, work-in-process quality is influenced by various factors and is directly related to the performance status of the processing equipment. The direct interaction between processing equipment and work-in-process means that any degradation in equipment performance directly impacts workpiece quality. Workpiece quality during production can also lead to performance degradation in downstream equipment components. The generation, accumulation, and propagation of variations at these stages create a coupling effect that can disrupt product quality stability. Therefore, studying the robustness of multi-stage manufacturing systems is essential for achieving consistent product quality, and researching maintenance decision optimization for equipment in multi-stage manufacturing systems is crucial for ensuring the robustness of the manufacturing system's operation.

[0003] Previous studies on the functional risks of multi-stage manufacturing systems have primarily focused on product quality and the reliability of processing equipment, using these as key indicators of functional risk. However, these studies have overlooked the importance of robustness as a crucial indicator for multi-stage manufacturing systems. The operational robustness of a manufacturing system typically reflects the stability of product quality and the performance status of processing equipment. When modeling the functional risks of multi-stage manufacturing systems, the relationship between robustness and machine performance degradation, product quality, and task delay risks must be considered to comprehensively assess the operational status of the manufacturing system. Furthermore, multi-stage manufacturing systems consist of various types of processing machines, each with different degradation patterns and states. Therefore, appropriate maintenance activities must be developed for different types of processing machines based on their degradation patterns. Selective maintenance, when limited maintenance resources are available, can rationally allocate limited maintenance costs and time to each processing machine to ensure the operational robustness and product quality of the next stage of the manufacturing system. However, most existing maintenance methods ignore the operating mechanism of multi-stage manufacturing systems and the relationship between task requirements and outputs. Robustness indicators are important indicators for measuring the production capacity of multi-stage manufacturing systems. They have a significant impact on production scheduling, product qualification rate and production costs. They directly reflect the production status of the manufacturing system and help guide the selective maintenance of the system.

[0004] To address this problem, this invention develops a selective maintenance decision optimization method for manufacturing systems that considers operational robustness. It aims to elucidate the functional risk connotation of manufacturing systems from the perspective of operational robustness, and establish a functional risk model that considers operational robustness. This model takes into account the impact of process parameter deviations, machine performance degradation, work-in-process and final product quality, and task delay risks on system robustness. Furthermore, the selective maintenance method for manufacturing systems considering operational robustness analyzes the impact of parameter changes caused by machine performance degradation risks on the robustness of the manufacturing system, and further studies the impact on work-in-process quality. This invention is applicable to maintenance decisions in manufacturing systems. Through quantitative assessment of robustness and under the constraint of limited resources, it proposes a selective maintenance optimization method, which has guiding significance for manufacturing system operation and maintenance decisions. Summary of the Invention

[0005] The purpose of this invention is to provide a selective maintenance decision optimization method for manufacturing systems that considers operational robustness. The robustness of multi-stage manufacturing systems is fundamental to ensuring task reliability and stable output of qualified final products. Product quality is a direct reflection of task reliability and functional risk. Furthermore, the operational robustness of a manufacturing system reflects the performance of the processing machines and the conditions of work-in-process (WIP). However, current maintenance strategies often neglect the functional risks and operational robustness of the manufacturing system. Therefore, this invention proposes a selective maintenance decision optimization method for manufacturing systems that considers operational robustness. First, based on the definition of operational robustness of multi-stage manufacturing systems, the functional risk formation mechanism is defined, and the principle of selective maintenance is proposed. Second, the synergistic effect between functional risk and operational robustness is explained, and a comprehensive evaluation model of functional risk considering operational robustness is established. Third, a selective maintenance method for multi-stage manufacturing systems that considers robust parameters is proposed, and the optimal maintenance decision is obtained using the Adaptive Simulated Annealing Particle Swarm Optimization (ASA-PSO) algorithm, maximizing the operational robustness of the multi-stage manufacturing system. Finally, the effectiveness of the proposed method is verified using a ferrite phase-shifting unit manufacturing system as an example.

[0006] (2) Technical solution:

[0007] This invention proposes a selective maintenance decision optimization method for manufacturing systems that considers operational robustness, based on the following assumptions:

[0008] Assumption 1: The degradation of all processing machines in this manufacturing system follows a semi-Markov process;

[0009] Assumption 2: The performance status of each processing machine refers to the amount of work that the processing machine can process under normal conditions;

[0010] Assumption 3: The corresponding key quality characteristics processed by each piece of machinery are considered to be independent of each other;

[0011] Assumption 4: There is an absolutely reliable checkpoint after each processing step to inspect the quality status of the output product;

[0012] Assumption 5: The processing capacity of the machine has a finite number of discrete states, and the corresponding probabilities are proportional;

[0013] Assumption 6: The robustness of a manufacturing system affects the probability of a processing machine operating in various performance states;

[0014] Assumption 7: The robustness of the manufacturing system is not affected by human or environmental factors;

[0015] Assumption 8: The cost and time of manufacturing system maintenance are related to the state before and after maintenance, and there is a one-to-one correspondence.

[0016] Based on the above assumptions, this invention proposes a selective maintenance decision optimization method for manufacturing systems that considers operational robustness, the steps of which are as follows:

[0017] Step 1: After a task cycle of the manufacturing system is completed, collect and analyze the data of the manufacturing system after the completion of this task;

[0018] Step 2: Based on Step 1, determine the performance status of the manufacturing system equipment, the deviation of product quality characteristics, and product quality;

[0019] Step 3: Determine the current robustness level of the manufacturing system based on Step 2;

[0020] Step 4: Collect data on machine maintenance, including the relationship between maintenance cost, maintenance time, and maintenance quality for each machine;

[0021] Step 5: Based on Step 2 and Step 4, establish a selective maintenance decision model with the objective of maximizing robustness in the next task phase and maintenance time and cost as constraints.

[0022] Step 6: Establish the ASA-PSO algorithm model to obtain the optimal maintenance strategy;

[0023] In step 1, “collecting and analyzing data from the manufacturing system after a task cycle is completed” means that after a task cycle is completed, the equipment in the manufacturing system may degrade to varying degrees, and it is necessary to collect data such as machine performance status and work-in-process quality deviations to facilitate subsequent assessment of the robustness of the manufacturing system.

[0024] Specifically, in step 2, "determining the performance status of the manufacturing system equipment, the deviation of product quality characteristics, and product quality based on step 1" involves determining the work-in-process quality characteristic deviation and product quality during the previous task's execution, and statistically analyzing the performance status of each processing device at the start and end of the previous task. This invention assumes that the normal operating time of the processing equipment follows a Weibull distribution. Where β is the shape parameter, η is the scale parameter, and β ≥ 0; furthermore, the failure probability function of the Weibull distribution is... The cumulative failure probability of the Weibull distribution is Generally, the larger the task load, the smaller the scale parameter. The more robust the process, the larger the scale parameter, and the closer it is to 1. Since the dwell time of the machining machine in any state follows a Weibull distribution, the probability of the machining machine being in each state at any time t is expressed as:

[0025] p(t)={p1(t),p2(t),p3(t),…,p w (t)} and

[0026] The performance of the machine tool continues to remain at s within the time interval Δt. x The probability of the state is:

[0027]

[0028] Here, we assume that the processing machine has w performance states, i.e., {s1, s2, ..., s}. w Given that at any time t, the performance state of the machining machine must be one of these w performance states, x∈{1,2,…,w}, therefore, at any time t, assume the performance state of the machining machine is s. x F x (t) represents the state of the processing machine at time t. x The probability of a state, F x (t+Δt) represents the state of the machining machine at time t+Δt. x The probability of [the outcome]. Furthermore, this invention suggests that the critical quality characteristic deviation of a manufactured workpiece can be expressed as [the following]. Where ε i,a It is a constant. and These are the sets of influence vectors of machine degradation factors and noise factors on the key quality characteristics of work-in-process; χ i (t) and ψ i (t) represents the vector sets of the controllable factor and the noise factor, respectively; Θ i,a This refers to the mutual influence between the two aforementioned effects; assuming there is l at workstation i. a Key quality characteristics, when χ iWhen (t) = 0, the machine performance is at its optimal state. Therefore, the deviation of the key quality characteristics under ideal conditions is: Therefore, the deviation of the key quality characteristics is:

[0029]

[0030] During the operation of a manufacturing system, due to the existence of uncertainties, χ i (t) follows a Weibull distribution, i.e., χ i (t)~f(t;β,η), therefore

[0031]

[0032] in

[0033]

[0034] Λ=Α i,a T cov(χ i (t))Α i,a The expression for the work-in-process output pass rate at station i is:

[0035]

[0036] Among them, USL a and LSL a These are the upper and lower limits in the design of key quality characteristics of the workpiece, respectively. ζ (0 < ζ ≤ 1) is an adjustment parameter based on the robustness of the manufacturing system to ensure that the pass rate meets the requirements.

[0037] In step 3, "determining the current robustness level of the manufacturing system based on step 2" refers to assessing the current robustness of the manufacturing system. The robustness of the processing equipment at time t... Therefore, the robustness of the entire manufacturing system can be derived. Here, the random variable J represents the length of time from the start of machine operation to failure. Robustness can be defined as the machine's operating time t being less than the random variable. Therefore, the probability that the failure time is greater than time t is defined as the robustness of the manufacturing system at that moment, i.e., P(J>t).

[0038] Specifically, step 4, "collecting data on machine maintenance, including the correspondence between maintenance costs, maintenance time, and maintenance quality for each machine," involves collecting maintenance data for each processing device in the manufacturing system, and establishing the correspondence between maintenance costs, maintenance time, and machine performance before and after maintenance. By performing maintenance activities within each task interval, faulty or aging machines in the manufacturing system are restored, ensuring the robustness of the manufacturing system in the next task. Each device has M... c+1 optional maintenance activity. During the m-th maintenance period, machine i can select the maintenance activity g. i,m And g i,m ∈{0,1,…,M c}, g i,m The larger the value, the better the maintenance effect. When g... i,m When g = 0, the machine is not repaired; when g = 0, the machine is repaired. i,m =M c At that time, the machine is perfectly repaired. Therefore, the maintenance cost of performing maintenance activities on machine i is: in, For the fixed maintenance cost of machine i, c i (g i,m ,s i,m ) represents the variable maintenance cost of maintenance activities, assuming that the processing machine i is in state s at the end of the m-th task. i,m This cost depends on the level of maintenance and the condition of the machine. Therefore, the total maintenance cost of the manufacturing system in the m-th task cycle is... The maintenance time for machine i to perform maintenance activities during the m-th task interval is in, It is the fixed maintenance time of machine i, t i (g i,m ,s i,m Let f(T1,m,T2,m,…,Tp,m) be the variable maintenance time for machine i, which depends on the maintenance level and machine condition. Therefore, assuming that maintenance of all equipment in the manufacturing system is parallel within any task interval, the total maintenance time of the manufacturing system is defined as Tm = f(T1,m,T2,m,…,Tp,m), where Tp,m represents the time required to maintain machine p in the m-th maintenance period. If the maintenance activity for each machine is performed once within the task interval, the maintenance time of the manufacturing system is the sum of the individual maintenance times. If the maintenance activities for each machine are independent and parallel, then the maintenance time of the manufacturing system is equal to the longest maintenance time among all machines.

[0039] Specifically, in step 5, "based on steps 2 and 4, establishing a selective maintenance decision model with the objective of maximizing robustness in the next task phase and maintenance time and cost as constraints" refers to establishing a selective maintenance decision model based on steps 2 and 4, with the objective of maximizing robustness in the next task phase and maintenance time and cost as constraints. The Kijima II model is used, and the maintenance effectiveness U of maintenance behavior is characterized by service life regression. l,m+1 =b l,m ·V l,m , among which, U l,m+1 V represents the effective service life of machine l after maintenance is performed during the m-th task interval. l,mLet b be the duration of the m-th task. i,m ≥1 is the service life regression factor, which is affected by maintenance resource input, b i,m The smaller the value, the better the repair effect. Therefore, the conditional survival probability of machine l after repair is:

[0040]

[0041] Among them, T m η represents the duration of task m. lx Let β be the scaling parameter of the Weibull distribution of machine l in state x. lx Let be the shape parameter of the Weibull distribution of machine l in state x. The robustness of the machine in the m-th task depends on the conditional survival probability r(t) and the state s of the machine at the end of the m-th task. break,m The robustness expression for machine l is R. l (m)=r l (m)s break,m ,in, Therefore, the robustness of the entire manufacturing system is... Assume the pre-allocated maintenance resources are: maintenance cost C0, and maintenance time T0. Therefore, the maintenance decision model is:

[0042] Objective function: Constraints:

[0043]

[0044]

[0045] Equations (2) and (3) show that the overall maintenance cost and maintenance time of the manufacturing system cannot exceed the resource constraints, while equations (4) and (5) show the variable constraints in the selective maintenance decision problem.

[0046] In step 6, "establishing the ASA-PSO algorithm model to obtain the optimal maintenance strategy" refers to the following table, which shows the pseudocode for selective maintenance decisions based on ASA-PSO:

[0047]

[0048]

[0049] Through the above steps, a selective maintenance decision optimization method for manufacturing systems considering operational robustness is proposed. This invention overcomes the limitations of traditional manufacturing system health status prediction. By analyzing the relationship between machine performance degradation and robustness, the robustness of the manufacturing system is maximized, thereby guiding selective maintenance decisions. The relationship between product qualification rate and machine performance status is examined, the functional risks and mechanisms of the manufacturing system are analyzed, and the use of maintenance resources is optimized to achieve better maintenance plans. The relationship between system robustness and the risks of machine performance degradation, task delays, and product quality non-conformity is considered. This method can be applied to various complex systems, and the ASA-PSO optimization algorithm effectively solves multi-constraint problems.

[0050] (3) Advantages and benefits:

[0051] This invention is an optimization method for selective maintenance decisions in manufacturing systems that considers operational robustness. Its advantages are:

[0052] i. This invention implements the concept of systems engineering, defines the connotation of functional risk, and explains the factors affecting the operational risk of manufacturing systems from the perspective of operational robustness.

[0053] ii. Based on machine status data, work-in-process quality data, task execution status data, and operational robustness data, this invention proposes a functional risk model.

[0054] iii. This invention considers the impact of operational robustness on the product quality, machine performance status, and task execution status of a manufacturing system, and proposes a robust selective maintenance decision model for manufacturing systems.

[0055] iv. The modeling method described is scientific and reasonable, has good manufacturability, and can be widely promoted in the industry. Attached Figure Description

[0056] Figure 1 This is a flowchart of the selective maintenance decision optimization method for manufacturing systems that considers operational robustness, as described in this invention.

[0057] Figure 2 This is a schematic diagram of the selective maintenance principle proposed in this invention, which considers robustness.

[0058] Figure 3 This is the selective maintenance decision-making framework diagram proposed in this invention.

[0059] Figure 4 This is a flowchart of the selective maintenance decision-making process using the ASA-PSO algorithm in this invention. Detailed Implementation

[0060] The present invention will now be described in further detail with reference to the accompanying drawings and examples.

[0061] This invention is an optimization method for selective maintenance decisions in manufacturing systems that consider operational robustness, see [link to relevant documentation]. Figure 1 As shown, the steps are as follows:

[0062] Step 1: After a task cycle of the manufacturing system is completed, collect and analyze the data from the manufacturing system at the end of this task. See [link to relevant documentation]. Figure 2 As shown, selective maintenance decisions are made based on the collected data. Table 1 shows the key processes and key workstations for the products produced by this manufacturing system, and Table 2 shows the relevant parameters of the machine.

[0063] Table 1. Key Quality Characteristic Parameters and Related Information

[0064]

[0065] Table 2. Relevant data of the processing machine

[0066]

[0067] Based on the collected key quality characteristic data, the production pass rate of the four machine tools was determined. Furthermore, the machine tools were divided into several discrete states, from optimal state to complete failure. The performance state set and production pass probability are shown in Table 3.

[0068] Step 2: Determine the work-in-process quality characteristic deviation and product quality during the previous task execution, and statistically analyze the performance status of each processing equipment at the start and end of the previous task. Substitute the data from Table 2 into...

[0069]

[0070] The quantitative indicators of the key quality characteristics of the workpiece in production from the previous task are as follows:

[0071] ν1(t)=7.6571×10 -8 t 2 +1.504×10 -5 t

[0072] ν2(t)=9.433×10 -8 t 2 +1.939×10 -5 t

[0073] ν3(t)=4.761×10 -8 t 2 +1.202×10 -5 t

[0074] ν4(t)=5.647×10 -8 t 2 +1.092×10 -5 t

[0075] Based on the collected key quality characteristic data and formulas:

[0076]

[0077] The production pass rate of the machine tool was obtained. Furthermore, the machine tool was divided into several discrete states, from optimal state to complete failure. The performance state set and production pass probability are shown in Table 3.

[0078] Table 3. Machine Performance Status Data

[0079]

[0080] Based on the above data, the probabilities of each state of the machining machine M1 at the end of the previous task execution phase are shown in Table 4.

[0081] Table 4. State distribution and probability of processing machines M1, M2, M3, and M4.

[0082]

[0083] After one production cycle, the production qualification rates of each processing machine are q1 = 0.979, q2 = 0.975, q3 = 0.972, and q4 = 0.957, respectively. The minimum input flow rate of each processing machine in one cycle is f1. I =225.07,

[0084] Step 3: Assuming a task cycle of 40 days, based on the operational data, the states of the four processing machines and their corresponding Weibull distribution parameters can be obtained.

[0085]

[0086] Therefore, the robustness of the manufacturing system at this point is R(t) = r1(t)·r2(t)·r3(t)·r4(t) = 0.8193 (all processing machines are operating normally, i.e., the robustness of the four processing machines is s). break,m =1)

[0087] Step 4: Collect data on machine maintenance, including the relationship between maintenance cost, maintenance time, and maintenance quality for each machine. Also, determine the acceptable state machine and its corresponding values ​​for each processing machine. 1,j ≥240,s 2,j ≥260,s 3,j ≥280,s 4,j ≥210, total maintenance cost is C0=8000, total maintenance time is T0=4. The fixed resources and maximum resources consumed for the maintenance of each processing machine are shown in the table below. Figure 3The selective maintenance framework diagram determines maintenance decisions. The next task's runtime is t. m =50 days. Clearly, the total maintenance resources are insufficient to restore all machines to optimal condition.

[0088] Step 5: Based on Steps 2 and 4, establish a selective maintenance decision model, rationally allocate maintenance resources, and maximize robustness in the next stage.

[0089] Table 5. Fixed and Maximum Resources Consumed for Machine Maintenance

[0090]

[0091] Step 6: After determining all parameters of the selected maintenance optimization model, the optimal selected maintenance strategy is determined using the ASA-PSO algorithm. In addition to the ASA-PSO parameter settings mentioned above, the population size (50) and the number of iterations (1000) are also set. Table 6 shows the final maintenance strategy of the ferrite phase-shifting unit manufacturing system and the corresponding robustness of the manufacturing system.

[0092] Table 6. Optimal Maintenance Strategy and Results

[0093]

[0094] Traditional selective maintenance methods typically rely on worker experience, historical data, and regular inspections. Workers use intuition to determine the condition of the equipment and the tasks requiring maintenance. Based on opinions collected from experts and production line workers in this study, the acceptable condition for each machine is as follows: 1,j ≥180,s 2,j ≥195,s 3,j ≥210,s 4,j ≥140. Traditional selective maintenance strategies are shown in Table 7.

[0095] Table 7. Results of Traditional Selective Maintenance

[0096]

[0097] In the absence of information about the intrinsic relationship between manufacturing system robustness and machine and product quality, traditional selective maintenance methods cannot guarantee the robustness of the manufacturing system in the next stage. Furthermore, overestimation of the acceptable processing condition of the machining system leads to inefficient utilization of maintenance resources. These discrepancies result in unplanned downtime of the manufacturing system during task execution. Therefore, the proposed method fully considers the interaction between manufacturing system robustness, machine performance, and product quality, reasonably and effectively assesses the robustness of task completion in the next stage, improves the utilization rate of maintenance resources, and accurately guides maintenance activities.

[0098] Based on this, the selective maintenance decision optimization for a certain manufacturing system has been completed. This modeling method can be used for subsequent production planning and task scheduling.

Claims

1. A method for optimizing selective maintenance decision of a manufacturing system considering operational robustness, comprising the following steps: Setting 1: The degradation of all processing machines in the manufacturing system follows a semi-Markov process; Setting 2: The performance state of each processing machine refers to the workload that the processing machine can process under normal circumstances; Setting 3: The corresponding key quality characteristics processed by each machine are considered to be independent of each other; Setting 4: There is an absolutely reliable inspection point after each processing to check the quality state of the output product; Setting 5: The processing capacity of the machine has a limited number of discrete states, and the corresponding probabilities are proportional; Setting 6: The robustness of the manufacturing system affects the probabilities of the processing machines in each performance state; Setting 7: The robustness of the manufacturing system is not affected by human factors and environmental factors; Setting 8: The cost and time of maintenance of the manufacturing system are related to the state before and after maintenance, and are one-to-one corresponding; Based on the above settings, the method comprises the following steps: Step 1: Collect and analyze the data of the manufacturing system after a task cycle of the manufacturing system ends; Step 2: Determine the performance state of the manufacturing system, the deviation of the product quality characteristics, and the product quality according to Step 1; Step 3: Determine the current robustness level of the manufacturing system according to Step 2; Step 4: Collect data about machine maintenance, including the corresponding relationship between the maintenance cost, maintenance time, and maintenance quality of each machine; Step 5: Based on Step 2 and Step 4, establish a selective maintenance decision model with the goal of maximizing the robustness of the next task stage and the constraints of maintenance time and cost; Step 6: Establish an ASA-PSO algorithm model to obtain the best maintenance strategy; In Step 4, collect maintenance data of each processing equipment in the manufacturing system, including the corresponding relationship between the maintenance cost, maintenance time, and performance state of each machine before and after maintenance; restore the malfunctioning or aging machines in the manufacturing system by performing maintenance activities within each task interval to ensure the robustness of the manufacturing system in the next task; wherein, in step 2, the normal working time of the processing equipment is subject to a Weibull distribution, wherein β is a shape parameter, η is a scale parameter, and β≥0; in addition, the failure probability function of the Weibull distribution is The cumulative failure probability of the Weibull distribution is: The greater the task load, the smaller the scale parameter; the more robust the process, the greater the scale parameter, and the closer to 1; since the stay of the processing machine in any state is subject to a Weibull distribution, the probability of the processing machine being in each state at any time t is represented as: p(t) = {p1(t), p2(t), p3(t),..., pn(t)} and w (t)} and The performance of the machine tool continues to be maintained at s x The probability of the state is: Suppose the machining machine has w performance states, i.e., {s1, s2, ..., sn}. w Given that at any time t, the performance state of the machining machine must be one of these w performance states, x∈{1,2,…,w}, therefore, at any time t, assume the performance state of the machining machine is s. x ;F x (t) represents the state of the processing machine at time t, s. x The probability of a state, F x (t+Δt) represents the state of the machining machine at time t+Δt. x The probability of; in addition, the key quality characteristic deviation of the manufactured workpiece is expressed as: Where ε i,a It is a constant. and These are the sets of influence vectors of machine degradation factors and noise factors on the key quality characteristics of work-in-process; χ i (t) and ψ i (t) represents the vector sets of the controllable factor and the noise factor, respectively; Θ i,a This refers to the mutual influence effect between the two aforementioned influences; suppose there is a l at workstation i. a Key quality characteristics, when χ i When (t) = 0, the machine performance is at its optimal state. Therefore, the deviation of the key quality characteristics under ideal conditions is: Therefore, the deviation of the key quality characteristics is: wherein in step 3, the robustness of the processing equipment at time t is: The random variable J represents the length of time the processing machine runs from the start to the failure, and the robustness is defined as the time t that the machine runs is less than the random variable J, so the probability that the length of time to failure is greater than the time t is defined as the robustness of the manufacturing system at that time, i.e. P(J > t); The maintenance time of machine i performing maintenance activities in the mth task interval is: Each device has M c +1 optional maintenance activity. During the m-th maintenance period, machine i can select the maintenance activity g. i,m And g i,m ∈{0,1,…,M c }, g i,m The larger the value, the better the maintenance effect; when g i,m When g = 0, the machine is not repaired; when g = 0, the machine is repaired. i,m =M c At that time, the machine is perfectly repaired. Therefore, the maintenance cost of performing maintenance activities on machine i is: in, For the fixed maintenance cost of machine i, c i (g i,m ,s i,m ) is the variable maintenance cost of maintenance activities; When the machine i is in state s at the end of the mth task i,m The cost depends on the degree of maintenance and the state of the machine; thus, the total maintenance cost of the manufacturing system over the mth task cycle is: The maintenance decision model is: wherein, is the fixed maintenance time of machine i, t i (g i,m , s i,m is the variable maintenance time of the maintenance activity of machine i, whose value depends on the maintenance level and the machine status; Thus, the maintenance of each equipment of the manufacturing system is parallel in any task interval, and the total maintenance time of the manufacturing system is defined as T m = f (T 1,m , T 2,m ,..., T p,m ), where T p,m represents the time required to repair machine p during the mth repair period; if the repair activities of each machine are performed once in a task interval, the repair time of the manufacturing system is the sum of the individual repair times, and if the repair activities of each machine are independent and parallel, the repair time of the manufacturing system is equal to the longest repair time among all machines; In step 5, the selective maintenance decision model is established with the objective of maximizing the robustness of the next task phase and maintenance time and cost as constraints; the maintenance effectiveness U of the maintenance behavior is characterized by the service life regression through the Kijima II model l,m+1 = b l,m · V l,m , wherein U l,m+1 is the effective service life of machine l after the maintenance operation in the mth task interval, V l,m is the duration of the mth task, b i,m ≥ 1 is a service life regression factor affected by the maintenance resource input, and the smaller b i,m is, the better the effect of the maintenance is; therefore, the conditional survival probability of machine l after the maintenance is: where T m represents the duration of task m, η lx is the scale parameter of the Weibull distribution of machine l in state x, β lx is the shape parameter of the Weibull distribution of machine l in state x; the robustness of machine in the mth task depends on the conditional survival probability r(t) and the state s of the machine at the end of the mth task break,m ; the robustness of machine l is expressed as R l (m) = r l (m) s break,m , where, Therefore, the robustness of the entire manufacturing system is Let the pre-allocated maintenance resources be: maintenance cost Co, maintenance time To; Equations (2) and (3) indicate that the total maintenance cost and maintenance time of the manufacturing system cannot exceed the resource constraints, and equations (4) and (5) indicate the variable constraints in the selective maintenance decision problem. Objective function: Constraints: Step 1 specifically refers to the end of a task cycle, where the equipment in the manufacturing system may have different degrees of degradation, and the machine performance state data and in-process quality deviation data need to be collected to facilitate subsequent evaluation of the robustness of the manufacturing system.

2. The method of claim 1, wherein: In Step 6, the pseudo code for selective maintenance decision based on ASA-PSO is as follows:

3. The method of claim 1, wherein: Input: initial particle population pop, maximum number of iterations Max_iter, lower limit of optimal solution lb, upper limit of optimal solution ub, dimension dim, objective function fobj, maximum value of velocity vector Vmax, minimum value of velocity vector Vmin; Output: position of optimal solution, fitness of optimal solution; Initialize parameters, initialize individual extreme value and global extreme value of particles; ​ If the upper bound of the optimal solution is a scalar, convert the upper bound of the optimal solution value to a matrix of 1 row and "dimension" columns, where each element is the upper bound of the optimal solution value; convert the lower bound of the optimal solution value to a matrix of 1 row and "dimension" columns, where each element is the lower bound of the optimal solution value; Loop from 1 until maximum number of iterations is reached; update inertia weight and acceleration constant according to the following equations V max = l x X max , l ∈ [0.1, 1], where k is the iteration number, V id is the velocity of particle i in dimension d, X id is the position of the i-th particle in dimension d, ω is the inertia control coefficient, r1 and r2 are random numbers in the interval [0, 1], c1 is the self-learning factor, c2 is the social influence factor, X max is the maximum bound of the solution; Update the velocity and position of each particle according to the following formula: where e is the natural logarithm, E(i) and E(j) represent the internal energy of the object in state i and state j, respectively, and K is the Boltzmann constant; Loop from 1 to the number of particles to perform the following calculation: If the fitness of the jth particle is less than the current record of the jth dimension of the global optimal fitness; update the position of the jth particle to the global optimal position of the current dimension; update the global optimal fitness of the jth dimension; If the fitness of the jth particle is less than the current record of the global optimal fitness; update the global optimal position; update the global optimal fitness; Record the optimal fitness of the current iteration to the curve array, output the position of the optimal solution and the fitness of the optimal solution.

Images