Method of performing maintenance activities on a plurality of interconnected production line machines, and the system thereof

Abstract

A computer implemented method of performing maintenance activities on a plurality of interconnected production line machines and a system for performing maintenance activities on said plurality of interconnected production line machines.

Description

Title

[0001] . A method of performing maintenance activities on a plurality of interconnected production line machines, and the system thereof 5 Technical Field

[0002] . The present invention generally relates to the field of real-time maintenance on multiple production line devices. Technical Background 10

[0003] . In the majority of integrated production and maintenance planning and scheduling literature, it is assumed that resources to execute maintenance are in abundance and available at any time. Unfortunately, this is not in line with the situation typically observed in practice. As it is important to study problems that have impact on real-life production environments, the limited availability of resources should not be 15 neglected. Reviews on the topic of maintenance scheduling and planning in combination with production are provided by Budai et al. (2008) “Maintenance and production: A review of planning models" (Complex System Maintenance Handbook. Springer series in Reliability engineering) and Geurtsen et al. (2022) “Production, Maintenance and resource scheduling: A review” (European Journal of operational research, volume 305, 20 Issue 2, 29 March 2022, pages 501-529). In Geurtsen et al. (2022), it was indeed noted that literature on integrated scheduling of maintenance, production and resources is scarce. Two streams of research can be identified in the field of combining maintenance, production and resources. These streams are separated by their problem type, where one stream considers scheduling and the other addresses planning or dispatching. The 25 difference is observed in two ways: (1) the way production is modeled and (2) the randomness involved in the modelling of the problem. In scheduling, production is explicitly considered through jobs that have to be assigned to machines, whereas planning considers production in-explicitly, i.e. it’s not a decision variable to be dealtwith. Furthermore, in scheduling, problems are almost always deterministic, while planning deals more with stochastic problems. Maintenance scheduling 5

[0004] . In the first stream of research, which deals with the integrated production, maintenance and resource scheduling problem, variations to this problem are presented by Rebai et al. (2013) “Scheduling jobs and maintenance activities on parallel machines” (Journal of Operational Research, 18 December 2012, volume 13 pages 363-383), Tavana et al. (2015) “An integrated three-stage maintenance scheduling model for 10 unrelated parallel machines with aging effect and multi-maintenance activities” (Journal of Computers & industrial engineering, volume 83, 1 May 2015, pages 226-236), Aramon Bajestani and Beck (2015) “A two-stage coupled algorithm for an integrated maintenance planning and flowshop scheduling problem with deteriorating machines” (Journal of Scheduling, Volume 18, page 471-486, 20 January 2015), Wang and Liu 15 (2015) “Multi-objective optimization of parallel machine scheduling integrated with multi-resources preventive maintenance planning” (Journal of Manufacturing systems, Volume 37, part 1, 1 October 2015)and Geurtsen et al. (2023b) “Integrated maintenance and production scheduling for unrelated parallel machines with setup times” (Flexible Services and Manufacturing Journal, 18 October 2023). In Rebai et al. (2013), jobs are 20 scheduled on parallel machines and multiple maintenance activities are considered. At most one maintenance activity can take place at one time and it can be scheduled anywhere along the scheduling horizon. Every machine is maintained once. An MILP (mixed integer linear program), a heuristic solution and GA (genetic algorithm) are presented to solve the aforementioned problem. The study by Tavana et al. (2015) 25 considers a rate-modifying PM (preventive maintenance) activity and resources for the unrelated a parallel machine environment with machines having varying performance (commonly known as “unrelated machines”). Processing times on machines deteriorate over time and the rate modifying PM activity restores the processing time back to normal. The resource problem is described as a repairmen selection problem, i.e., whichrepairmen is going to maintain which machine at what time. Each machine is maintained by only one repairman whereas each repairman can take care of more than one machine. Furthermore, Tavana et al. (2015) presents a multi objective ILP (integer linear program). Deteriorating processing times are also considered by Aramon Bajestani and Beck 5 (2015). Maintenance must be scheduled on multiple machines and restore the processing times back to normal. Resources are modelled as highly skilled maintenance technicians required to perform a PM activity, of which there is a constant limited availability. The difference with the work of Tavana et al. (2015) is the way in which processing times are modeled. An MILP is developed for the problem and they show that the integrated 10 approach of scheduling maintenance and production simultaneously yields better results than the non-integrated approach. A different study altogether is presented by Wang and Liu (2015), in which PM activities must be scheduled for both the machines as well as the resources required for said machines. In the work of Geurtsen et al. (2023b), a novel maintenance scheduling problem is introduced, where maintenance on unrelated parallel 15 machines must be scheduled in a set of flexible windows. Multiple maintenance activities per machine are considered and the maximum number of maintenance activities that can be scheduled simultaneously is limited to the maximum number of available resources at that time. An MILP and Hybrid Genetic Algorithm are presented for addressing the problem. A case-study on real-world production data shows the 20 efficiency of scheduling maintenance flexibly, compared to fixed maintenance. The recent study by Mota et al. (2023) “Joint optimization of production and maintenance for a serial-parallel hybrid two-stage production system” (Journal of reliability engineering & system safety, Volume 226, October 2022) also develop a genetic algorithm to schedule production, maintenance, and resources in a flexible job shop 25 setting, aiming to cut energy and maintenance costs. Their approach treats maintenance flexibly, allowing scheduling outside designated windows. Using real production data, they demonstrate that their algorithm efficiently schedules maintenance by rearranging tasks to accommodate multiple activities and timing them at high production energy prices. Unfortunately, their focus is solely on reducing energy consumption and neglects 30 improvements in production efficiency.Maintenance planning

[0005] . In the second stream of research, we first look at the maintenance planning problem on a single production line without resources, before examining research with 5 multiple resources. Planning without resources

[0006] . Studies that consider a single production line with multiple machines and intermediate buffers in series are considered in Ni et al. (2015) “Preventive maintenance 10 opportunities for large production systems” (CIRP Annals, Volume 64, Issue 1, pages 447-450, 3 June 2015), Huang et al. (2020) “Deep reinforcement learning based preventive maintenance policy for serial production lines” (Journal of Expert systems with applications, Volume 160, 1 December 2020), Valet et al. (2022) “Opportunistic maintenance scheduling with deep reinforcement learning” (Journal of manufacturing 15 systems, Volume 64, 1 July 2022, pages 518-534), Geurtsen et al. (2023a) “Deep reinforcement learning for optimal planning of assembly line maintenance” (Journal of manufacturing systems, Volume 69, 1 August 2023, pages 170-188)). The study by Ni et al. (2015) examines a large complex production line of both series and parallel machines. They examine PM opportunity windows by utilizing starved and blocked 20 states of machines and predict opportunity windows in real time in response to downtime events. Their goal is to optimize the throughput of the production line. In Huang et al. (2020), a deep Reinforcement Learning approach is presented. Their state space includes machine ages, buffer levels and remaining maintenance durations of the ongoing maintenance activities. The policy obtained by the agent is compared to a run-to-failure 25 policy, an age-based replacement policy and an opportunistic policy. The agent learns a policy that combines aspects of both opportunistic maintenance and group maintenance, which outperforms the other policies in terms of maintenance cost. Maintenance opportunities induced by approaching condition-based breakdown, as well as opportunities triggered by external factors such as empty / full buffers are examined inValet et al. (2022). The decision to perform PM is based on a time-to-failure distribution for each machine and the content of the buffer prior to the machine. Decisions based on machine production states and the use of corrective maintenance activities as opportunities are ignored. In the recent study by Geurtsen et al. (2023a), a single 5 production line is considered where maintenance must be planned based on machine production states, buffer levels and a flexibility window prior to reaching the end of the PM due time. A novel deep reinforcement learning approach is presented and they show that by considering these production line elements, higher throughput improvement can be achieved compared to a fixed product based dispatcher. However, all these studies do 10 not consider the possible unavailability of resources. Planning with resources

[0007] . A sub-stream where production and maintenance planning are combined with a form of resource constraints is the literature on capacitated production lines. One of the 15 first to consider multiple production lines is the study by Aghezzaf and Najid (2008) “Integrated production planning and preventive maintenance in deteriorating production systems” (Journal of information sciences, volume 178, issue 17, 1 September 2008, pages 3382-3392). In their study, preventive maintenance is performed periodically and corrective maintenance activities are considered as well. Both maintenance tasks, as well 20 as production, require a certain capacity of resources. The goal is to minimize total costs and a Lagrangian heuristic is developed for this purpose. They are the first to model both maintenance and production planning simultaneously. Their study is extended by Yalaoui et al. (2014) “Integrated production planning and preventive maintenance in deteriorating production systems” (Journal of information sciences, volume 278, 10 25 September 2014, pages 841-861) by considering a new approach which deal with a broader range of problems. The study of Aghezzaf and Najid (2008) is also extended by Ettaye et al. (2017) “Integrating production and maintenance for a multi-lines system” (International journal of Performability Engineering, volume 13, number 1 January 2017, pages 29-44) by adding a set of additional constraints. In both extensions, corrective andpreventive maintenance activities that require a certain amount of resources are considered. However, preventive maintenance is not planned, but instead assumed to be performed periodically at fixed time intervals. Both production activities and maintenance activities require resources in terms of capacity. Unlike Aghezzaf and Najid 5 (2008), the study by Yalaoui et al. (2014) proposes a new formulation based on mixed numbers and Ettaye et al. (2017) develop an MIP (mixed integer program) to solve the problem. Unfortunately, studies that examine capacitated production lines don’t optimize the policy of when to execute maintenance, but instead take it as given. Another approach is to treat the problem of maintenance planning on production lines as a 10 workload balancing problem, as in Truong et al. (2023) “Modelling and application of joint maintenance grouping and workload smoothing for an automotive plant” (International journal of production research 62(8), 2832-2852, 16 July 2023). In their study, multiple production lines are considered, both correct maintenance (CM) and PM activities are considered, but only PM requires a resource to perform. The aim in Truong 15 et al. (2023) is to plan maintenance for a group of production lines and to assign resources on these production lines to minimize total maintenance costs. A case-study on an automotive plant is performed and it shows that both the maintenance cost and workload variance can be reduced. The study by Gao et al. (2021) “Joint optimization on maintenance policy and resources for multi-unit parallel production system, Journal 20 of computers & industrial engineering, volume 159, September 2021) on the other hand considers PM activities on production lines where both repairmen and spare-parts need to be assigned to perform a PM activity. Their goal is to define a PM plan that minimizes total maintenance cost and adopt a GA to solve the problem. Corrective maintenance or other stochastic down times are not considered in Gao et al (2021). A series-parallel 25 production line with maintenance and replacement activities is considered by Sheikhalishahi et al. (2016) “Maintenance Scheduling Optimization in a Multiple production line considering human error” (Human Factors and Ergonomics in Manufacturing & Service Industries, Volume 26, issue 6, August 2016). They also include human error, where each activity or task has a specific probability of human 30 error. Machines in the production lines have a failure rate and the goal is to create amaintenance plan to optimize cost, human error and reliability. In all 3 aforementioned studies, the main focus is on maintenance costs and the effect of maintenance on production is completely neglected. Another perspective on maintenance planning involves examining limited capacity for maintenance in multi-machine systems, rather 5 than focusing solely on production lines. Optimizing production and maintenance for deteriorating multi machine systems is studied by Koopmans and de Jonge (2023) “condition-based maintenance and production speed optimization under limited maintenance capacity” (Journal of computers & industrial engineering, volume 179, may 2023). They introduce condition-based production and demonstrate, through a Markov 10 decision process (MDP), that adjusts production speeds to reduce simultaneous maintenance needs and increase profit. Their findings highlight the importance of synchronized production and maintenance decisions in maximizing system availability and profit. Hu et al. (2023) “Knowledge-enhanced reinforcement learning for multi- machine integrated production and maintenance scheduling” (computers & industrial 15 engineering, volume 185, 1 November 2023) also study a multi-machine system where each machine operates independently and deteriorates with production. They conduct PM activities every period based on machine condition. With one scarce resource, their aim is to maximize production through optimal scheduling of production and maintenance operations. They propose a novel reinforcement learning technique, which 20 surpasses baseline methods by improving failure avoidance. Unfortunately, the scheduling is performed within deterministic scenario’s and real-time planning is not considered.

[0008] . In view of the challenges mentioned above in this field, there is a growing demand for innovative solutions that address the limitations of conventional 25 maintenance techniques and overcome the hurdles associated with these techniques.Summary

[0009] . It would be advantageous to achieve a leadless package that overcomes the difficulties as addressed above. It would further be advantageous to achieve a corresponding method of manufacturing such a leadless package. 5

[0010] . In a first aspect of the present disclosure, there is provided a computer implemented method of performing maintenance activities on a plurality of interconnected production line machines, the method comprising obtaining, by a server, at least one resource configured to perform maintenance, obtaining, by the server, at least one maintenance value from each of the plurality of machines, obtaining, by the server, 10 a list of maintenance activities for performing maintenance on the plurality of interconnected production line machines, determining, by the server, whether to perform a maintenance activity on the list of maintenance activities on at least one of the plurality of machines based on the at least one resource and the at least one maintenance value, wherein the determining comprises, for each maintenance activity on the list of 15 maintenance activities: calculating a threshold value, based on the at least one maintenance value, indicating a remaining interval for each of the plurality of machines to complete the maintenance activity, sorting the threshold value of the plurality of machines in ascending order, wherein when the at least one resource is available to perform maintenance, allocating a maintenance activity on said one resource on a first 20 machine of the plurality of machines currently not undergoing maintenance, wherein the first machine has a shortest expected remaining end time for the maintenance activity determined for said first machine of the plurality of machines, performing said maintenance activity, wherein the first machine has a smallest threshold value of the plurality of machines, and when the at least one resource is not available to perform a 25 maintenance activity, identifying a second machine of the plurality of machines currently not undergoing maintenance activity with the smallest threshold value of the plurality of machines, wherein when said smallest threshold value is greater than an expected time interval for the at least one resource to complete a maintenance activity for said machine, identifying said second machine as the next machine for performing maintenance and 30 allocating a next available resource for performing said maintenance activity, andrepeating this step for all machines in ascending order of expected remaining end time for the maintenance activity determined for each machine; when said smallest threshold value is smaller than the expected time interval for the at least one resource to complete a maintenance activity, adjusting, for all threshold values, the threshold value of said 5 machine when the maintenance activity has been performed, and adjusting for all maintenance activities currently in use by the resource, the expected time interval for the at least one resource to complete the maintenance activity for the plurality of machines by subtracting the smallest threshold values from both the threshold value and the time interval, and deallocating said one resource after completion of maintenance activity. 10

[0011] . The advantage of such a process is that it is able to simulate (re-enact) the entire system as a whole in with 100% resource utilization to determine the optimum maintenance activity sequence for the system of production line machines. Another benefit is that such a system is also scalable.

[0012] . In a further embodiment of the present disclosure, the at least one resource is a 15 list of available resources for performing maintenance. This allows for the specialization of resources for performing maintenance activities for a particular machine.

[0013] . In a further embodiment of the present disclosure, the at least one maintenance value is at least one of an expected time of a time interval for performing maintenance, a list of mean maintenance performance duration, list of maintenance operations 20 currently being performed, list of elapsed times of resource currently in use, and a stochastic factor.

[0014] . In a further embodiment of the present disclosure, wherein the threshold value is determined to be a sum of the list of mean maintenance duration for maintenance operation, for each machine and for each resource, with a stochastic factor, wherein the 25 stochastic factor delays or advances performing the maintenance activity with respect to the expected duration of the maintenance activity.

[0015] . In a second aspect of the present disclosure, there is provided a system (100) for performing maintenance activities on a plurality of production line machines, the system comprising, a server (101) configured to obtain at least one resource configured to 30 perform maintenance, obtain at least one maintenance value from each of the pluralityof production line machines (102a – 102n), obtain a list of maintenance activities for performing maintenance on the plurality of interconnected production line machines (102a – 102n), determine whether to perform a maintenance activity on the list of maintenance activities on at least one of the plurality of machines (102a – 102n) based 5 on the at least one resource and the at least one maintenance value, wherein the determining comprises, for each maintenance activity on the list of maintenance activities: calculate a threshold value, based on the at least one maintenance value, indicating a remaining interval for each of the plurality of machines to complete the maintenance activity, sort the threshold value of the plurality of machines (102a – 102n) 10 in ascending order, wherein when the at least one resource is available to perform maintenance, allocate a maintenance activity on said one resource on a first machine of the plurality of machines (102a – 102n) currently not undergoing maintenance, wherein the first machine has a shortest expected remaining end time for the maintenance activity determined for said first machine of the plurality of machines, perform said maintenance 15 activity, wherein the first machine has a smallest threshold value of the plurality of machines (102a – 102n), when the at least one resource is not available to perform maintenance activity, identify a second machine of the plurality of machines (102a – 102n) currently not undergoing maintenance activity with the smallest threshold value of the plurality of machines, when said smallest threshold value is greater than an 20 expected time interval for the at least one resource to complete a maintenance activity for said machine, identifying said second machine as the next machine for performing maintenance and allocating a next available resource for performing said maintenance activity, and repeating this step for all machines in ascending order of expected remaining end time for the maintenance activity determined for each of the plurality of 25 machines (102a – 102n), when said smallest threshold value is smaller than the expected time interval for the at least one resource to complete a maintenance activity, adjusting, for all threshold values, the threshold value of said machine when the maintenance activity has been performed, and adjusting for all maintenance activities currently in use by the resource, the expected time interval for the at least one resource to complete the 30 maintenance activity for the plurality of machines by subtracting the smallest thresholdvalues from both the threshold value and the time interval, and deallocating said one resource after completion of maintenance activity, and a plurality of production line machines (102a – 102n) configured to transmit, to the server (101), at least one maintenance value, receiving an allocation of a resource for performing maintenance 5 and transmitting a deallocation of the resource after maintenance.

[0016] . In a further embodiment of the present disclosure, the at least one resource is a list of available resources for performing maintenance.

[0017] . In a further embodiment of the present disclosure, the at least one maintenance value is at least one of an expected time of a time interval for performing maintenance, 10 a list of mean maintenance performance duration, list of maintenance operations currently being performed, list of elapsed times of resource currently in use, and a stochastic factor, wherein the stochastic factor delays or advances performing the maintenance activity with respect to the expected duration of the maintenance activity.

[0018] . In a further embodiment of the present disclosure, the threshold value is 15 determined to be a sum of the list of mean maintenance duration for maintenance operation, for each machine and for each resource, with a stochastic factor, wherein the stochastic factor delays or advances performing the maintenance activity with respect to the expected duration of the maintenance activity.

[0019] . In the figures, similar components and / or features may have the same reference 20 label. Further, various components of the same type may be distinguished by following the reference label by a dash and a second label that distinguishes among the similar components. If only the first reference label is used in the specification, the description is applicable to any one of the similar components having the same first reference label irrespective of the second reference label. 25

[0020] . The above and other aspects of the disclosure will be apparent from and elucidated with reference to the examples described hereinafter.Brief description of the Figures

[0021] . Figure 1 illustrates an embodiment performing preventive maintenance on a plurality of production line machines according to the present invention.

[0022] . Figure 2 illustrates the difference between a first come first serve (FCFS) 5 decision process and the heuristic process to determine when maintenance should be performed for a 2-production line machine system according to an embodiment of the present invention.

[0023] . Figure 3 illustrates the heuristic n-line approach for a one resource and three production line machines system according to an embodiment of the present invention. 10

[0024] . Figure 4 illustrates an embodiment according to the present invention where there are three production lines with different PM durations and two resources according to an embodiment of the present invention.

[0025] . Figure 5 illustrates a comparison of the effectiveness of different decision- making processes used to determine whether to perform maintenance on the line 15 according to an embodiment of the present invention.

[0026] . Figure 6 illustrates a comparison of the PM decision process for a heuristic, three line, two resource system according to an embodiment of the present invention.

[0027] . Figure 7 illustrates a system configured to implement the algorithm according to an embodiment of the present invention. 20 Detailed Description

[0028] . It is noted that in the description of the figures, same reference numerals refer to the same of similar components performing a same of essentially similar function.

[0029] . A more detailed description is made with reference to particular examples, some 25 of which are illustrated in the appended drawings, such that the features of the present disclosure may be understood in more detail. It is noted that the drawings only illustrate typical examples and are therefore not to be considered to limit the scope of the subject matter of the claims. The drawings are incorporated for facilitating an understanding of the disclosure and are thus not necessarily drawn to scale. Advantages of the subjectmatter as claimed will become apparent to those skilled in the art upon reading the description in conjunction with the accompanying drawings.

[0030] . The ensuing description above provides preferred exemplary embodiment(s) only, and is not intended to limit the scope, applicability, or configuration of the disclosure. 5 Rather, the ensuing description of the preferred exemplary embodiment(s) will provide those skilled in the art with an enabling description for implementing a preferred exemplary embodiment of the disclosure, it being understood that various changes may be made in the function and arrangement of elements, including combinations of features from different embodiments, without departing from the scope of the disclosure. 10

[0031] . Unless the context clearly requires otherwise, throughout the description and the claims, the words "comprise," "comprising," and the like are to be construed in an inclusive sense, as opposed to an exclusive or exhaustive sense; that is to say, in the sense of "including, but not limited to." As used herein, the terms "connected," "coupled," or any variant thereof means any connection or coupling, either direct or 15 indirect, between two or more elements; the coupling or connection between the elements can be physical, logical, electromagnetic, or a combination thereof. Additionally, the words "herein," "above," "below," and words of similar import, when used in this application, refer to this application as a whole and not to any particular portions of this application. Where the context permits, words in the Detailed 20 Description using the singular or plural number may also include the plural or singular number respectively. The word "or" in reference to a list of two or more items, covers all the following interpretations of the word: any of the items in the list, all of the items in the list, and any combination of the items in the list.

[0032] . These and other changes can be made to the technology considering the following 25 detailed description. While the description describes certain examples of the technology, and describes the best mode contemplated, no matter how detailed the description appears, the technology can be practiced in many ways. Details of the system may vary considerably in its specific implementation, while still being encompassed by the technology disclosed herein.

[0033] . As noted above, particular terminology used when describing certain features or aspects of the technology should not be taken to imply that the terminology is being redefined herein to be restricted to any specific characteristics, features, or aspects of the technology with which that terminology is associated. In general, the terms used in the 5 following claims should not be construed to limit the technology to the specific examples disclosed in the specification, unless the Detailed Description section explicitly defines such terms. Accordingly, the actual scope of the technology encompasses not only the disclosed examples, but also all equivalent ways of practicing or implementing the technology under the claims. 10

[0034] . A well-known method of planning maintenance is to schedule a specific time at which maintenance is performed on a machine for a specific duration of time. In the case of one machine and one maintenance mechanic, the ideal situation is that the time interval between the scheduled maintenance times is far greater than the time it takes to complete the maintenance. In the case of multiple machines but one mechanic, it is then 15 necessary to have the scheduled maintenance time interval to be sufficiently great such that the mechanic can perform maintenance on each machine to prevent unwanted shutdown.

[0035] . Another aspect to consider is to minimize the amount of maintenance which needs to be performed on each machine. The more the machine is out of service due to 20 maintenance, the less the machine outputs. Therefore, the specific times at which maintenance is performed should maximized and be as late as possible.

[0036] . The issue with a time-based approach for maintenance on such production lines, especially high-speed production lines configured to manufacture products is that the machine performance degrades over time. In such a fast-paced environment, a minor 25 degradation in the machine performance has serious implications in the machines production output. As the rate of degradation is different for each machine, this results in each production line producing a different number of products over the same time period.

[0037] . Therefore, according to an embodiment of the present invention, the 30 determination on whether to perform maintenance is based on the number of productsthe machine has (already) produced, and not based on a time interval. This can increase throughput of equipment as less maintenance is required (you are trying to prevent doing maintenance too early). An example of this is then shown in Figure 1.

[0038] . Figure 1 illustrates an embodiment according to the present invention. Figure 1 5 discloses a plurality of production lines (indicated as “Line 1” to “Line N”). Each production line has a timeline which, according to the invention, illustrates the number of products manufactured by the production line. Each production line has a maximum product limit to perform maintenance, indicated as Preventive Maintenance (PM) limits in the Figure 1. As the performance of each production line is different due to 10 deterioration and other factors, the calculation of the PM limits varies for each line. Here, the window W for PM may be preferably fixed for each production line (or would be a universally set variable), with W being number of products produced by the line. The production line starts counting W from the point at which PM has been successfully completed (PM execution, marked as crosses on the timelines), and determines what the 15 maximum PM limit for the subsequent maintenance. In case we would have the same number of resources / technicians as number of production lines, we would be able to let maintenance run until the end of the window. However, this is typically not the case and resources are scarce. Therefore, as displayed in Figure 1, sometimes you would like to initiate maintenance a bit earlier to prevent the multiple lines have to perform 20 maintenance at the same time, resulting in one or more lines having to wait until a resource becomes available. As such, it is desirable that the PM execution is performed before the PM limit.

[0039] . One of the advantages of maintaining a product timeline in this manner is that the PM execution is recorded as a point on the timeline, as during maintenance the output 25 of the production line is zero.

[0040] . In order to perform the maintenance procedure as described above, the most common practice would be to simply do a First Come First Served (FCFS) approach, which is what was described previously, where maintenance is performed at the latest possible time. Here, the latest possible time is a predefined (or predetermined limit 30 whereby, if said limit is reached without maintenance, the machine cannot performfurther production and must either shutdown or wait for maintenance. Figure 2 discloses some drawbacks to this approach for the case of 2 production lines, with 1 resource (maintenance engineer). Figure 2 illustrates the timeline in the form of elapsed time (different from Figure 1 where the timeline is based on the number of products produced 5 by the production line machines). As shown in Figure 2 for the FCFS approach, line 2 first undergoes maintenance, as it is closer to the PM limit (or end of PM interval) than line 1. The maintenance engineer would then expend their resource on maintaining line 2. During this time, line 1 is also approaching its PM limit. In this case, as the resource is already being used by line 2, line 1 must shut down in order to perform maintenance 10 at a later time. Once the maintenance in line 2 is completed (and production has resumed on this line) then line 1 undergoes maintenance.

[0041] . Instead of the FCFS approach above, the present invention adopts a Heuristic approach, also shown in Figure 2, which not only uses the expected PM limit to determine when to perform maintenance, but also takes into account how long 15 (timewise) such PM is expected to take.. In the embodiment described in Figure 2, both lines 1 and 2 have an expected PM time of 1 hour. In an ideal situation, a PM limit of 1 hour is required for PM for both lines with one resource without impeding the production lines. However, in reality a PM limit of slightly larger time period is required to account for external factors, such as the mechanic needing time to move from one line to another,20and so on. In such case, a threshold value, Trequired, may be determined to take this extra time period into account in addition to the PM limits of each line (“End PM interval”). This threshold value Trequired may also be (or additionally be) determined by a predicted latest possible time for PM without shutting any of the machine down, in order to not perform PM too early, based on historical data or other simulations. For the embodiment 25 described in Figure 2, the Trequiredis set to 1 hour. Once the PM timeline for one of the lines (in this case, line 2) approaches this threshold value, the system proceeds to inform the resource (maintenance engineer) to perform maintenance on line 2 (illustrated as a check mark) while line 1 continues to operate normally. At this point in time, line 1 is of 1.2 hours away from its PM limit, and can thus maintain production until PM on line 30 2 is completed. After maintenance on line 2 has been performed, the system informs theresource to perform maintenance on line 1 (illustrated as a check mark) before the PM limit of line 1 is reached. In this manner, the maintenance on both lines may be performed before the PM limits, and thus preventing any machines on either production line to shut down. 5

[0042] . An example pseudo-code to illustrate this Heuristic approach is shown in Table 1 below.Table 1 10

[0043] . In Table 1 it is assumed that both production lines require the same threshold value Trequired as input. As shown in Table 1, if there is an available resource (line 2), the heuristic approach maintains a list of remaining time intervals for a 2-line production line (lines 4-6) and sorts the list in ascending order (lines 7, 8). In line 9, the available 15 time, Tavailable,is determined to be the longest remaining time period of the two lines fromthe PM limit (i.e., the line that requires maintenance later). If the available time Tavailable is less than or equal to the threshold value Trequired , maintenance is performed on the line closest to the PM limit (lines 10 – 14).In line 16, the loop continues until a resource is no longer available.

[0044] . As an illustration, for the heuristic approach shown in Figure 2, let us assume that the remaining up time step for line 1 is 1.2 hrs, and 0.5 hrs for line 2, and Trequired for both lines to being 1 hr. The values for wn in line 1 of Table 1 would then be w1 = 1.2 hrs, w2 = 0.5 hrs. The T-list would then have two elements in ascending order (lines 7,8) where wn = [0.51.2]. According to line 9, Tavailable is 1.2 hrs. As the condition is not met in line 10, no maintenance is performed on line 2. Only when Tavailable reaches 1 hr in line 1 does maintenance begin in line 2. The conditional statement in line 10 maximizes the PM interval by restricting maintenance to as late as possible (and as close to the PM limit as possible).

[0045] . The heuristic approach shown in Figure 2 shows the situation when two lines with very similar PM limit times (Figure 2, left), one of the lines undergoes maintenance at the point where the other line reaches the threshold value (Figure 2, second from left), which gives ample time for said other line to maintain production (Figure 2, second from right) while the one line completes maintenance for the other line to undergo maintenance (Figure 2, right)

[0046] . For more than two production lines and more than one available resource, the above algorithm may lead to sub-optimal maintenance scheduling, as illustrated in Figure 3.

[0047] . Figure 3 discloses an embodiment where there are three production lines and one resource using the heuristic approach above. Here, when maintenance is performed on line 2, the threshold value Trequiredis initially taken to be 2 hours due to the remaining lines also nearing their PM limits with PM durations of 1 hr each (Figure 3, left). As the resource will take 1 hr for each line, the threshold value must be dynamically calculated to PM each line without shutting any of the lines down. When maintenance is completed on line 2, all three lines continue production (Figure 3, second form left), until line 3 reaches a threshold value of 1 hr (Figure 3, second from right) at which point line 1undergoes maintenance. Once line 1 completes maintenance all three lines resume production until line 3 reaches its PM limit. As the remaining PM time intervals of the other two lines are sufficiently large, line 3 can simply perform maintenance normally (Figure 3, right). 5

[0048] . The pseudo-code of Table 1 thus needs to be adapted to take the dynamic characteristics of the threshold value into account. An example pseudo-code to illustrate this heuristic approach is provided in Table 2 below.10 Table 2

[0049] . As shown in Table 2, two new variables have been introduced in the code to determine the threshold value.

[0050] . The first variable is factor L., which is a random factor to account for stochastics (this value is 0 for the deterministic / ideal case there the resource can perform maintenance seamlessly from one line to another). This is a value (a constant) set by a user. This factor, which is a constant variable to be determined by a user using this 5 algorithm, gives a user the possibility to delay or advance performing the maintenance activity with respect to the expected duration of maintenance activity. As the expected duration of the maintenance activity may change over the lifetime of the machine, this introduces an inherent randomness / stochastics to the algorithm, which can be compensated by for by introduction of this factor L. When L is a positive value, a time 10 is added to the expected duration of maintenance, which makes the 'expectation of the duration of the maintenance activity' longer (i.e. an indication is presented that the maintenance activity may take longer than the average, and thus maintenance activity may need to be performed sooner), ultimately leading to an earlier initiation of maintenance activity for other activities that are approaching the end of their interval. 15 Similarly, when we make L negative, we adjust the expected duration towards a value which is smaller, leading to later initiation of other maintenance activities.

[0051] . The second variable is Dn, which is the average PM duration. These two variables are added to the threshold value, which allows the threshold value to be adaptable, based on the number of production lines. In the case shown in Figure 3 (left),20as the initial Trequired and Dn are both 1 hour, the updated Trequired value is 2 hrs (when L=0). For Figure 3 (second from right), the threshold value is 1 hr, similar to the two line configuration shown in Figure 2 and Table 1.

[0052] . Figure 4 further illustrates an embodiment according to the present invention where there are three production lines with different PM durations and two resources. 25 The PM duration of lines 1 and 3 is one hour, whereas for line 2 the PM duration is four hours. Figure 4 (left) illustrates the situation line 3 is 5 hrs away from the PM limit while the other lines are undergoing maintenance (due to multiple resources). In this case the production (or activity) of line 3 is continued without interruption, even after the maintenance of line 1 (which takes one hour) has been completed (Figure 4, right). Withmultiple resources, it is essential to only perform maintenance when it is absolutely necessary to do so.

[0053] . Figure 5 illustrates how the decision-making process to determine whether to perform maintenance on the line may lead to sub-optimal results. For the ‘Bad decision’ 5 case, the system performs maintenance on two lines simultaneously because the system deemed both to lines 1 and 2 to have passed the threshold value. However, this is not the case for line 1, and the maintenance of line 1 could be postponed to a time period much closer to the PM limit as shown in the ‘Good decision’ case.

[0054] . According to an embodiment of the present invention, a good algorithm / heuristic 10 approach overcome the above problems by finding solutions to the question (PM decision): “Would any of the resources be available before reaching the end of the PM interval, if maintenance is continuously initiated in order of expected PM interval end time, referring to a scenario where all resources would be fully occupied?”.

[0055] . Maintenance should be initiated now on the machine that reaches the end of its 15 interval first if the answer to this question is no. To obtain the answer, this scenario should be played out, meaning that re-enactment must be applied to simulate the execution of maintenance.Table 3

[0056] . Table 3 illustrates the pseudo code implementing this PM decision.

[0057] . In contrast to focusing on one activity per production line, it is designed to handle any number of maintenance activities, defined by the set S.

[0058] . Prior to initiating the algorithm, data regarding the current state of the production line environment is collected or obtained (lines 1-5 and input). This encompasses gathering information on the anticipated time until an activity (S) reaches its maintenance interval end (ws), a list of available resources (Ra), a list of the elapsed maintenance times for resources currently engaged (Rt), the activities currently undergoing maintenance (SPM), and the average maintenance duration derived from the distribution associated with maintenance activity S (Ds).

[0059] . Herein, the term activity (S) refers to a list of maintenance activities. A singular form for activity S may be denoted with small s. As each machine is different, different maintenance activities must be performed on each machine. However, if the machines are similar initially, the general maintenance activity may be the same for all machines, with minor technical differences distinguishing each activity (such as the amount of lubrication necessary for maintaining the drive belt, and so on). The maintenance activity (S) can thus be the same, similar, or different, depending on the machine.

[0060] . The algorithm is executed only when resources are available (line 6). This execution results in the determination, by the algorithm in a server, whether to perform a maintenance activity (from the list of possible maintenance activities) on at least one of the machines (to which the server is communicatively coupled to) based on the current state of the production line environment.

[0061] . Similar to Algorithm 3 in Table 2, the algorithm begins by storing the anticipated time until reaching the end of the PM interval for each activity (s) not currently in maintenance (lines 8-11). Here, the anticipated times wsof the activities are stored in the list T, while the corresponding activities S are stored in the list TA. Following this, both lists T and TA are sorted in ascending order based on the values in list T (lines 12-13).

[0062] . Subsequently, for each activity currently in maintenance, the expected time until completing the PM activity is determined (lines 15-18). This value is then added to a threshold value Rr, which is a list that contains, for each activity in maintenance, the anticipated (or expected) remaining time until maintenance is finished (or completed). Rrmay be calculated from Ds, TA and a factor L. Factor L is the same as that described in Table 2. Rr is continuously sorted in ascending order of duration.

[0063] . The re-enactment procedure is initiated (lines 19-37) to ascertain (or determine) whether maintenance must be initiated immediately. This re-enactment procedure simulates a scenario where all resources are used, mimicking 100% resource utilization. Initially, if any resources are available, a maintenance activity is allocated to the available resource to initiate PM on the maintenance activity closest to reaching the end of their maintenance window, which are stored in the list of ordered activities TA (lines 20-23). The maintenance activity closest to reaching the end of the maintenance window refers to a machine, out of the plurality of machines, that has the smallest value T out of the machines, meaning that said machine has the shortest expected time until interval end for activity s. After the (available) resource is allocated to the machine, the maintenance activity is performed on said machine. Throughout the algorithm, a resource is deallocated from the machine once the maintenance is completed (i.e. the maintenance activity has been performed).

[0064] . If no resources are available, a check is performed to determine whether the remaining duration of the first maintenance activity in Rr (for a machine) exceeds the anticipated end time of the next maintenance activity in the ordered list of maintenance activities that require PM, stored in T. Said list is ordered in ascending order of Rr. If this condition is met, maintenance is initiated (lines 24-26) on said machine at the next resource availability, as the expectation is that the specified maintenance activity will reach the end of the PM interval before a resource becomes available, resulting in unnecessary wait times. If this condition is not met, the loop continues, and the duration of the maintenance activities in Rr and the end times of the remaining maintenance activities in T are adjusted by subtracting the duration of the first maintenance activityto reach the end of maintenance in Rr (lines 27-32). This subtraction simulates the (temporary) situation where the first maintenance activity has been completed.

[0065] . Subsequently, this first maintenance activity is removed from Rr, leading to a new available resource, which is then used to initiate PM on the next maintenance 5 activity in TA, which is added to Rr(lines 33-34).

[0066] . This entire process continues until all maintenance activities are processed. If the condition to initiate maintenance is triggered, maintenance is executed for the maintenance activities closest to reaching the end of their PM interval, stored in the ordered list TA, with the number of maintenance activities receiving maintenance being 10 equal to the available number of resources (lines 38-45).

[0067] . With this heuristic, there is no scenario in which only one of the multiple available resources should be used.

[0068] . The re-enactment procedure assumes that maintenance should only be initiated if, in a scenario of 100% resource utilization, there is a possibility that an activity might 15 need to wait for a resource. If this condition is not met, it is more effective to let the activities run closer to the end of their PM interval, improving long-term throughput.

[0069] . An example operation of the pseudo code in Table 3 is illustrated in Figure 6. The configuration described here has two PM events for line 1, each lasting one hour, one PM event for line 2 with a four-hour duration, and line 3 has one PM event of one 20 hour. There are two available resources.

[0070] . For the case where no PM decision is made, in Figure 6 (left), there are two available resources and two lines (lines 1 and 2) where the PM event is nearing the PM limit. The PM event on line 3 is 3 hours away, which is ample time as the two PM events on line 1 take 1 hour each. The two resources then perform maintenance on lines 1 and 25 2 (Figure 6, second from left). After the first PM event in line 1 is completed, maintenance is performed on the second PM event on line 1, as the PM event in line 3 is further away (Figure 6, second from right). This allows for both resources to be utilized. However, as shown in Figure 6 (right), the completion of the PM events without a PM decision has resulted in only one of the resources being utilized while the PM event 30 on line 3 is approaching its PM limit, which is not efficient management of the resources.

[0071] . For the case where PM decision is made it is shown that, even though the PM event on line 3 is closer to the PM limit than the second PM limit of line 1 (Figure 6, second from left), the PM is first performed on line 1, as the PM decision has identified that the PM event is sufficiently far from the PM limit to already start performing 5 maintenance (Figure 6, second from right), and can therefore proceed with PM after the second PM event of line 1 is complete (Figure 6, right).

[0072] . Figure 7 shows the system (100) according to an embodiment of the present invention. The production line machines 1 – N (102a – 102n) are configured to communicate with the server (101). The server (101) is configured to receive at least 10 one maintenance value from each production line machine (102a – 102n) constantly, during both the PM event and also outside said PM events. For each machine (102a – 102n), the server may obtain the following maintenance values: - When under maintenance, the expected (time) interval until the end of the maintenance period. 15 - When under maintenance, the elapsed time of resource currently in use - When under maintenance, the list of activities currently in maintenance - When not under maintenance, the expected time interval until maintenance is required. - When not under maintenance, what type of maintenance (activity) needs to be 20 performed on the machine.

[0073] . The server is configured to dynamically calculate the threshold value and the required value based on the above variables.

[0074] . During the lifetime of the system (100), the server (101) obtains a wealth of data to determine the rate of degradation (or stochastics) for each machine. As the rate of 25 degradation is not uniform and cannot be controlled, each machine will have its own stochastic factor L. This factor is always positive, as a degradation slows the system down.

[0075] . The server (101) is further configured to communicate with a maintenance / resource planner. Such planner may already be included in the server but 30 could also be an external device communicatively coupled to the server (101). Themaintenance planner, alongside the measurement variables, determines the type of maintenance to perform on each machine, and allocates resources for such function.

[0076] . Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, 5 the disclosure, and the appended claims. The provided figures and descriptions of the embodiments of the invention are illustrative and explanatory to the heart of the invention and should not be seen as limiting the invention thereto. In the claims, the word “comprising” does not exclude other elements or steps, and the indefinite article “a” or “an” does not exclude a plurality. The mere fact that certain measures are recited 10 in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope thereof.

Claims

Claims 1: A computer implemented method of performing maintenance activities on a plurality of interconnected production line machines, the method comprising - obtaining, by a server, at least one resource configured to perform maintenance, - obtaining, by the server, at least one maintenance value from each of the plurality of machines, - obtaining, by the server, a list of maintenance activities for performing maintenance on the plurality of interconnected production line machines, - determining, by the server, whether to perform a maintenance activity on the list of maintenance activities on at least one of the plurality of machines based on the at least one resource and the at least one maintenance value, wherein the determining comprises, for each maintenance activity on the list of maintenance activities: o calculating a threshold value, based on the at least one maintenance value, indicating a remaining interval for each of the plurality of machines to complete the maintenance activity, o sorting the threshold value of the plurality of machines in ascending order, wherein o when the at least one resource is available to perform maintenance, ^ allocating a maintenance activity on said one resource on a first machine of the plurality of machines currently not undergoing maintenance, wherein the first machine has a shortest expected remaining end time for the maintenance activity determined for said first machine of the plurality of machines, ^ performing said maintenance activity, wherein the first machine has a smallest threshold value of the plurality of machines, and o when the at least one resource is not available to perform a maintenance activity, ^ identifying a second machine of the plurality of machines currently not undergoing maintenance activity with the smallest threshold value of the plurality of machines, wherein^ when said smallest threshold value is greater than an expected time interval for the at least one resource to complete a maintenance activity for said machine, identifying said second machine as the next machine for performing maintenance and allocating a next 5 available resource for performing said maintenance activity, and repeating this step for all machines in ascending order of expected remaining end time for the maintenance activity determined for each machine; ^ when said smallest threshold value is smaller than the expected time interval for the at least one resource to complete a maintenance activity, adjusting, for all threshold values, the threshold value of said machine when the maintenance activity has been performed, and adjusting for all maintenance activities currently in use by the resource, the expected time interval for the at least one resource to complete the maintenance activity for the plurality of machines by subtracting the smallest threshold values from both the threshold value and the time interval, and ^ deallocating said one resource after completion of maintenance activity. 2: The computer implemented method according to claim 1, wherein the at least one resource is a list of available resources for performing maintenance. 3: The computer implemented method according to any one of the above claims, wherein the at least one maintenance value is at least one of an expected time of a time interval for performing maintenance, a list of mean maintenance performance duration, list of maintenance operations currently being performed, list of elapsed times of resource currently in use, and a stochastic factor, wherein the stochastic factor delays or advances performing the maintenance activity with respect to the expected duration of the maintenance activity. 4: The computer implemented method according to any of the above claims, wherein the threshold value is determined to be a sum of the list of mean maintenance duration for maintenance operation, for each machine and for each resource, with a stochastic factor, wherein the stochasticfactor delays or advances performing the maintenance activity with respect to the expected duration of the maintenance activity. 5: A system (100) for performing maintenance activities on a plurality of production line machines, the system comprising, - a server (101) configured to obtain at least one resource configured to perform maintenance, obtain at least one maintenance value from each of the plurality of production line machines (102a – 102n), obtain a list of maintenance activities for performing maintenance on the plurality of interconnected production line machines (102a – 102n), determine whether to perform a maintenance activity on the list of maintenance activities on at least one of the plurality of machines (102a – 102n) based on the at least one resource and the at least one maintenance value, wherein the determining comprises, for each maintenance activity on the list of maintenance activities: calculate a threshold value, based on the at least one maintenance value, indicating a remaining interval for each of the plurality of machines to complete the maintenance activity, sort the threshold value of the plurality of machines (102a – 102n) in ascending order, wherein when the at least one resource is available to perform maintenance, allocate a maintenance activity on said one resource on a first machine of the plurality of machines (102a – 102n) currently not undergoing maintenance, wherein the first machine has a shortest expected remaining end time for the maintenance activity determined for said first machine of the plurality of machines, perform said maintenance activity, wherein the first machine has a smallest threshold value of the plurality of machines (102a – 102n), when the at least one resource is not available to perform maintenance activity, identify a second machine of the plurality of machines (102a – 102n) currently not undergoing maintenance activity with the smallest threshold value of the plurality of machines, when said smallest threshold value is greater than an expected time interval for the at least one resource to complete a maintenance activity for said machine, identifying said second machine as the next machine for performing maintenance and allocating a next available resource for performing said maintenance activity, and repeating this step for all machines in ascending order of expected remaining end time for the maintenance activity determined for each of the plurality of machines (102a – 102n), when said smallest threshold value is smaller than the expected time intervalfor the at least one resource to complete a maintenance activity, adjusting, for all threshold values, the threshold value of said machine when the maintenance activity has been performed, and adjusting for all maintenance activities currently in use by the resource, the expected time interval for the at least one resource to complete the maintenance activity 5 for the plurality of machines by subtracting the smallest threshold values from both the threshold value and the time interval, and deallocating said one resource after completion of maintenance activity, and - a plurality of production line machines (102a – 102n) configured to transmit, to the server (101), at least one maintenance value, receiving an allocation of a resource for performing maintenance and transmitting a deallocation of the resource after maintenance. 6: The system of claim 5, wherein the at least one resource is a list of available resources for performing maintenance. 7: The system according to any one of claims 5 and 6, wherein the at least one maintenance value is at least one of an expected time of a time interval for performing maintenance, a list of mean maintenance performance duration, list of maintenance operations currently being performed, list of elapsed times of resource currently in use, and a stochastic factor, wherein the stochastic factor delays or advances performing the maintenance activity with respect to the expected duration of the maintenance activity. 8: The system according to any one of claims 5 to 7, wherein the threshold value is determined to be a sum of the list of mean maintenance duration for maintenance operation, for each machine and for each resource, with a stochastic factor, wherein the stochastic factor delays or advances performing the maintenance activity with respect to the expected duration of the maintenance activity

Images