Scholarly article on topic 'Cloud-based Integrated Shop-floor Planning and Control of Manufacturing Operations for Mass Customisation'

Cloud-based Integrated Shop-floor Planning and Control of Manufacturing Operations for Mass Customisation Academic research paper on "Agriculture, forestry, and fisheries"

Share paper
Academic journal
Procedia CIRP
OECD Field of science
{"Production Planning and Control" / "Mass Customisation" / Decision-making / "Software as a Service" / "Cloud Manufacturing"}

Abstract of research paper on Agriculture, forestry, and fisheries, author of scientific article — D. Mourtzis, M. Doukas, C. Lalas, N. Papakostas

Abstract The shift of traditional mass producing industries towards mass customisation practices is nowadays evident. However, if not implemented properly, mass customisation can lead to disturbances in material flow and severe reduction in productivity. This paper discusses the design and development of a Cloud-based production planning and control system for discrete manufacturing environments, referred to as i-MRP. The proposed approach takes into consideration capacity constraints, lot sizing and priority control in a ‘bucket-less’ manufacturing environment. The i-MRP system offers simultaneous shop scheduling and material planning, where material and capacity constraints are considered together in a continuous time environment. A number of feasible alternative shop schedules and material plan combinations are formed and are evaluated on the Cloud platform where the i-MRP engine is hosted. The Cloud platform enables mobility, since it is device and location independent, as well as it minimises the cost of IT infrastructure ownership, which is especially important for SMEs. The performance of the i-MRP system has been studied in an SME from the textile sector, using real production data. The system demonstrates high performance in cases of short production times, high value inventory and frequent, small deliveries by suppliers. The i-MRP can be easily integrated with legacy IT systems as an interfaced functional module under the Software as a Service (SaaS) architecture.

Academic research paper on topic "Cloud-based Integrated Shop-floor Planning and Control of Manufacturing Operations for Mass Customisation"

Available online at


Procedia CIRP 33 (2015) 9- 16

9th CIRP Conference on Intelligent Computation in Manufacturing Engineering - CIRP ICME '14

Cloud-based integrated shop-floor planning and control of manufacturing

operations for mass customisation

Mourtzis D.a*, Doukas M.a, Lalas C.a, Papakostas N.a

aLaboratory for Manufacturing Systems and Automation (LMS), Department of Mechanical Engineering and Aeronautics, University of Patras, Rio Patras,

Greece, 26500

* Corresponding author. Tel.: +30 2610 997262; fax: +30 2610 997744. E-mail address: Abstract

The shift of traditional mass producing industries towards mass customisation practices is nowadays evident. However, if not implemented properly, mass customisation can lead to disturbances in material flow and severe reduction in productivity. This paper discusses the design and development of a Cloud-based production planning and control system for discrete manufacturing environments, referred to as i-MRP. The proposed approach takes into consideration capacity constraints, lot sizing and priority control in a 'bucket-less' manufacturing environment. The i-MRP system offers simultaneous shop scheduling and material planning, where material and capacity constraints are considered together in a continuous time environment. A number of feasible alternative shop schedules and material plan combinations are formed and are evaluated on the Cloud platform where the i-MRP engine is hosted. The Cloud platform enables mobility, since it is device and location independent, as well as it minimises the cost of IT infrastructure ownership, which is especially important for SMEs. The performance of the i-MRP system has been studied in an SME from the textile sector, using real production data. The system demonstrates high performance in cases of short production times, high value inventory and frequent, small deliveries by suppliers. The i-MRP can be easily integrated with legacy IT systems as an interfaced functional module under the Software as a Service (SaaS) architecture.

© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (

Selection and peer-review under responsibility of the International Scientific Committee of "9th CIRP ICME Conference" Keywords: Production Planning and Control; Mass Customisation; Decision-making; Software as a Service, Cloud Manufacturing

1. Introduction

At the heart of currently used closed-loop Manufacturing Resource Planning (MRP-II) and Enterprise Resource Planning (ERP) systems in manufacturing enterprises, lies the fundamental Material Requirements Planning (MRP) logic [1][2]. Such IT systems entail major investments and involve extensive efforts and organisational changes in companies that decide to employ them. They integrate all business processes of the entire enterprise and tie the financial and marketing functions to the operations function, incorporating assets such as human resources, project management, product design, material and capacity planning [4].

Still, the classic time-phased material planning procedure is at the core of these systems as far as the production planning function is concerned [2] [2]. Despite the vast and increasing adoption of such commercial MRP-based systems [5], a growing number of authors criticize their poor performance in relation to implementation costs. Recent studies, such as Lapiedra et al., [6], showed that few manufacturers were able

to implement MRP-based systems successfully. In a survey conducted in [7], it was revealed that only 37% of the implementations achieved predicted budgets, and 66% of the companies realised less than half of the projected benefits. Moreover, while accurate percentages of unsuccessful implementations vary from study to study, nearly 20% of the times they are characterised as failures [7], with only a small number of companies achieving a Class A MRP operation [8]. The main reasons for that are commonly attributed to the fact that MRP-based systems do not produce detailed shopfloor schedules, since standard MRP method merely specifies the job release and completion dates in the context of time buckets [9]. Also, most of them assume infinite production capacity, thus using inflated, constant, and thus unrealistic lead times.

2. Literature Review

The integration of capacity limitations into the MRP planning process was one of the major areas of research in the past. Since the pure MRP logic is deployed in today's MRP-II,

2212-8271 © 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Selection and peer-review under responsibility of the International Scientific Committee of "9th CIRP ICME Conference" doi:10.1016/j.procir.2015.06.004

ERP and Supply Chain Management (SCM) systems, when they are used for production planning and control, their outputs suffer from the same false assumptions. The majority of the less recently published models and algorithms on capacity-sensitive production planning are complex and difficult to use in industrial practice. They employ either complex mathematical programming, as in Rahmnani et al., [10] and Wu and Shi [11], or heuristics in order to calculate workload-dependent planned lead times, as in Dobson and Karmarkar [12] and Aouam and Uzsoy [13]. Due to the fact that when the number of variables and constraints is raised the computational time increases rapidly, they operate under a lot of simplistic assumptions and restrict their use only to small problems. In addition, their performance under a dynamic production environment may be unreliable. As they are not easily understood by the planner, confidence in their results is limited [14]. Finally, the complications that product customisation introduces to the modern shop-floor, heavily its production planning and control functions, thus, the robustness of the deployed MRP-based solution is a catalytic factor for high performance.

These limitations have led to a lot of recently published research on the performance of different finite-capacitated production planning systems based on standard MRP. Pandey et al. [14] presented a capacitated material requirements planning algorithm that has been found to be superior to the existing standard MRP system in terms of mean job tardiness and inventory holding cost per part. However, the lot-sizing problem is not addressed and only a single resource for each part type is assumed to be available. Ho and Chang [15] proposed an integrated MRP and Just-In-Time (JIT) framework, modelled as an integer linear program in combination with forward and backward heuristics for finding detailed shop floor schedules with the objective of minimizing the total production cost without, however, providing any information related to actual implementations. Furthermore, they neither addressed the lot-sizing problem, nor did they deal with the refinement of the scheduling problem using multiple criteria. Koh et al. [16] presented the development and implementation of a generic model for simulating MRP-controlled finite-capacitated manufacturing environments in order to study the effects of uncertainty and production fluctuations on the performance of a company. The output of the infinite capacitated MRP planning acts as the main input to the simulation model. This link is problematic due to the fact that the simulation model does not recognize dependence relationships among the parts and hence it simply processes a part whenever the required resources are available. Iranpoor et al., [17] studied a general flexible flowshop scheduling problem minimising earliness and tardiness penalties deriving from the less or excess quantity produced. The model used deterministic processing times and fixed penalty costs for predefined production quantities. A capacitated master planning problem with inventory constraints over discrete multi-period horizons is presented in [18]. The paper focused solely on the determination of optimum pricing policies and schedules for a master facility planning level, thus the time components were aggregated, and possibly non-realistic. Another recent study, considered a drum-buffer-rope-based production planning

method as a control mechanism that exhibited high potential as a decision making tool especially for turbulent manufacturing environments [19]. However, a constraint of the method is the acquisition of real-time accurate monitoring data that require specialised high-investment equipment found only in state of the art shop-floors.

Literature on capacity-sensitive MRP fails to provide a comprehensive solution to all discrete manufacturing environments especially for finite short to medium-term horizon. Their application is usually limited to a number of constraints, such as the number of machines and processing stages, input from an interfaced MRP system, and bucketed planning horizon convention to name a few.

The proposed i-MRP production planning and control system incorporates a flexible workload and facility modelling, capable of representing the entirety of discrete manufacturing systems. In comparison to commercial solutions, it requires low implementation efforts, involves minimum organisational changes for its deployment, and is based on an easily maintainable. Moreover, the system considers finite production capacity in a bucket-less environment, thus it provides a detailed schedules without inflated constant time components. Finally, a Cloud infrastructure is designed for hosting and exposing the application as a service. This servitisation model offers benefits such as mobility and low investment costs, and allows the easy maintenance, synchronisation and version-control of material and production planning information.

3. Concept and design of the i-MRP system

This section presents the basic structure of the proposed i-MRP system. The i-MRP production planning and control tool is itself not MRP-based and supports the integration of shop scheduling and material planning under constraints imposed by the finite capacity of a manufacturing system. Items with relatively small cycle times and/or high inventory value, which is usually the case in textiles, can pass under its control.


Heuristic Dispatch Policy

Figure 1. i-MRP system main inputs and outputs

The i-MRP tool has been primarily developed for the textile industry, but, it can be also implemented in other multi-

stage and multi-product manufacturing environments with multiple parallel machines at each stage of production. Final or semi-final products can have linear or divergent structures with multi-level components. Alternative resources can be used to perform the same operation considering their specific quality, productivity, setup, and processing costs per time unit. Setup time in every resource is sequence dependent. Each final product must follow the precedence relationships in its routing. The major direct inputs / outputs of the i-MRP system are presented in Figure 1 and are described hereafter.

3.1. Inventory Record

The i-MRP method needs to be up-to-date regarding the exact inventory status of every item it controls before it plans order releases for each one of them. Information about projected on-hand quantities, scheduled order releases and receipts is stored in a file named Inventory Record (IR), as well as changes due to stock receipts, changed orders, stock withdrawals, scrap, corrections imposed by cycle counting and other similar events. Transient subassemblies or phantom items that are never stored but are directly consumed during the production of their parent items are not included. Transactions of such items are also not recorded in the IR module, thus reducing administrative efforts. Apart from the dynamic data that are stored and regularly updated, the IR module also contains static data that describe each item uniquely. These data are important in purchasing, cost accounting and other functions of a firm and include: part name, code/number, low-level code, unit of measure, raw materials' supply lead time, lot-sizing technique, safety stock, safety lead time, standard ordering cost, lot-sizing adjustment factors and linkage to the compact BOM module, which is the second basic input to i-MRP.

3.3. Finite Capacity Scheduling

Orders released to the shop floor are directed to a dynamic Finite Capacity Scheduling (FCS) module in order to allocate them to specific resources, thus producing the entire shop floor schedule. The FCS module creates a hierarchical model of both the production facility and the workload and operates under discrete event simulation. The production facility is divided into Job Shops that can produce a product family of similar semi-final and end products. Each Job Shop is further divided into Workcenters, which in turn consist of a number of Resources (Figure 2). The latter can be defined as individual production cells or parallel processors that can perform similar operations. Corresponding to the facility's hierarchy there is also the workload hierarchical breakdown. Orders are broken down into Jobs, which in turn consist of a number of Tasks. An order corresponds to the overall production facility and is divided into Jobs that based on their specifications, can be processed only by a suitable Job Shop. A Job consists of Tasks that can be released to one Work-centre only. Tasks can be dispatched to more than one of the Work-centre's Resources. Among the constraints taken into consideration in releasing and dispatching Jobs and Tasks are the facility's finite capacity and their precedence relationships.

I—Is divided intOyAre divided into-^


Are released to ^^ Facility

Are released to \ And dispatched to




I LJ(Machine Tools

.,.lTV- ■ Operators)

f I ' f I ^ f

Consists of-1 I-Consist of—1 1-Consist of-1

Figure 2. Hierarchical modelling of workload and facility

3.2. Compact Bill of Materials

In contrast to the simple BoM, the compact (c-BOM) combines both BOM and routing data in a single file, associating components and raw materials with the operation that requires them in the routing sequence. Thus, all parents of a material are operations and vice versa. The construction of c-BOM is appropriate for estimating Work-In-Process (WIP) inventory as well as for cost accounting purposes. The most important function of c-BOM is that it facilitates the integration of material and capacity planning. The function of the c-BOM module is critical for the integration of material planning and shop scheduling within the i-MRP algorithm. The c-BOM module is setup by combining single level compact BOMs in a matrix form. The latter is essentially a 'where-used' summarised list, where all items are listed along with the corresponding quantities or fractions with which they participate in their parent operations. Moreover, expensive tooling or chemicals consumable during material processing are included in a c-BOM. The i-MRP system plans replenishment orders for them as for any other item, based on the shop floor schedule as derived by the FCS module, which is the third basic input to i-MRP.

The operational policy behind the assignment of a task to a specific resource can be either a simple dispatching rule, or a multiple-criteria decision making technique [20][21][22]. The advantages of dispatching rules derive from their simplicity. Since they do not attempt to predict the future, they make decisions based on the present. Thus, these rules are very useful in factories that are extremely unpredictable, such as job shops. Also, dispatching rules are usually spatially local, requiring only the information available at the location where the decision will be implemented. When the multiple-criteria decision making technique is employed, several alternatives are formed and evaluated before assigning the available resources to pending production tasks. The choice of the best alternative is made by evaluating a set of criteria, such as cost, flowtime, earliness, reserve time, queue time, and tardiness, in a decision matrix. A utility function is applied to rank the alternatives and choose the best. Released orders are scheduled directly, without aggregation. Schedules are constructed on the basis of events occurring sequentially through time. The next scheduling decision is identified by moving along the time horizon until an event is scheduled to occur that will initiate a change in the status of the system. This would usually be the completion of a task on one of the resources or the arrival of a job to one of the work-centre

queues. All operations eligible for loading at the time a resource becomes available are considered. When there are multiple jobs competing for a resource, the selected operational policy is used to determine the highest priority operation. Hence, the schedule is constructed by simulating the detailed shop activity through real calendar time.

The FCS module has been specially adjusted in order to be able to schedule both forward and backward. The Forward Scheduling (FS) function will schedule all tasks of a job from its arrival date, starting with the first task. It aims at completing each job as early as possible. It can also be used to find out whether the earliest feasible completion time will meet customer's requirements. The Backward Scheduling (BS) function schedules all tasks of a job from its due date, starting with the last task. Its objective is to complete each job on or as close as possible to its due date, thus minimizing its slack time. Moreover, combinatory Forward/Backward Scheduling (F/BS) can be employed in order to schedule orders with different priorities either forward, or backward. Furthermore, the FCS module is capable of splitting an order to parallel resources, lap-phase orders by moving transfer batch sizes through successive operations, adding capacity through overtime and simulate deterministic / stochastic breakdowns.

3.4. The LCC lot-sizing technique

The lot-sizing technique built-in the i-MRP system, referred to as Least Cumulative Cost (LCC), computes cumulative requirements in order to determine an order's lot size. In order to support the continuous time operation of the i-MRP system it allows for both lot sizes and order time intervals to vary. The LCC technique is based on the premise that the sum of setup (for manufactured items) or ordering (for purchased items) cost and inventory carrying cost will be minimised when these are nearly equal. The exact lot size is computed within the following constraints: (i) portion of the order lot consumed in the day of order receipt incurs no carrying cost (the carrying cost of the rest is proportional to the time required to be consumed, according to the SFS), (ii) cumulative carrying cost must be within a predefined percentage range of the ordering and total costs should be below a given amount specified by the company's policy, (iii) specific minimum, maximum and multiple batch sizes per item imposed, based on vendor/process equipment considerations, and (iv), safety stock levels adjustments applied based on material considerations.

3.5. The i-MRP algorithm

The i-MRP algorithm, the engine of the system, coordinates the operation of the system's main input modules. First, an operational policy is selected from the ones available (heuristic rule or a multi-criteria configuration). The FCS module schedules orders on a level-by-level basis starting with the operations at the bottom or top level of their respected c-BOMs, depending on whether they are being scheduled forward (FS function) or backward (BS function) [26], respectively. As soon as the FCS forms the SFS, the i-MRP

system produces the material plan by relating every scheduled operation to the corresponding materials in the c-BOM module. Requirements are accumulated in continuous time from every scheduled parent operation in the shop schedule and netted against IR's data. The LCC lot-sizing technique is employed in order to group them and determine an order's lot size. The output data include a set of alternative shop schedule / material plan combinations, their evaluation reports, and the capacity load profiles per resource. The dispatch rules used in the experiments are include in below.

Table 1. Description of the dispatching rules utilised as operational policies


SPT Task with the shortest processing time is selected LPT Task with the longest processing time is selected EDD Task with the earliest due date is selected FIFO Task which first arrives at the factory is selected LIFO Task which last arrives at the factory is selected MOPNR Task which has the most operations remaining to

be performed is selected FOPNR Task which has the fewest operations remaining to

be performed is selected FASFS Task which arrives first in the job shop is selected LWRK Task which has the least work remaining to be

performed is selected MWRK Task which has the most work remaining to be _performed is selected_

4. Cloud-based deployment

A cloud infrastructure has been designed for hosting the i-MRP tool. Cloud-based implementations have been recently reported for hosting various manufacturing applications for machine availability monitoring [31], collaborative and adaptive process planning [31], and for online tool-path programming based on real-time machine monitoring [33].

The selection of a Cloud-based infrastructure has been made due to the following reasons. It offers increased mobility and information availability to a company, by being device and location independent. Moreover, the relatively low cost of purchasing the application as a service is advantageous for SMEs who cannot afford investing on high-performance computing installations [34]. The main challenge for the adoption of Cloud in manufacturing is the lack of awareness on security issues and standards. This major issue, can be addressed using security concepts and inherently safe architectures, such as private Clouds. The security concept will include availability of Information Technology (IT) systems, network security, software application security, data security and finally operational security [35].

The designed platform is service-centred; the three modules of the i-MRP system are decoupled from one another. A "User Cloud" is designed for exposing the systems functionalities to the planner, through Graphical User Interfaces (GUIs). The user requests are handled by the "Cloud Manager", which comprises of a broker agent that is hosted on a web-server and acts as an intermediary between the requested and the provided service [29][30].

The Apache HTTP server handles the http requests and operates as a connector between the i-MRP application and Internet. The communication between the applications and the database is achieved through the Model-View-Controller (MVC) architectural pattern. This is a robust architecture that allows the management of the database without any straight database-language programming and also allows custom made GUIs as mentioned above through the combination of HTML, CSS (Cascading Style Sheets), JavaScript and Java programming languages.

Client's Client's Client's

browser browser browser

Resolution of requests to a specific folder, containing the developed Web Application

Visualization for end-user (charts, data input-output) Necessary services to "connect" with database

Implementation of algorithms and rest functions

Spinning, and Weaving department, modelled as job-shops. The hierarchical facility model breakdown of the production line and the tasks associated with each work-centre, based on [27][28], are listed in Table 2 and are depicted in Figure 3.

Table 2. Hierarchical model of the blend carpets production facility

Job Shop Work-centre Tasks Resource No. of Resources

Dyeing-JS DYE-WC Dyeing DYE-R# 1

(20 types) PRESS-WC Hydroextraction PRESS-R# 1

DRY-WC Drying DRY-R# 1

Spinning-JS PREP-WC Preparatory Tasks PREP-R# 2

(60 types) BLEND-WC Blending BIN-R# 6

CARD-WC Carding CARD-R# 3

SPIN-WC Spinning SPIN-R# 7


WIND-WC Cleaning WIND-R# 3

Weaving-JS SPOOL-WC Spooling SPOOL-R# 4

(200 types) WARP-WC Warping WARP-R# 2

WEAV-WC Weaving WEAV-R# 27

3 Job-shops 12 Workcenters 59 Resources

Stored information, Workload/facility model/..

Figure 3. The Cloud Architecture

The backbone of the system consists of the storage cloud that manages a manufacturing data model. The data model stores and manages information regarding the workload and the facility model, as well as data necessary for the i-MRP modules, as described in section 3 above. The storage cloud is supported by a versioning agent. This agent periodically checks for changes in the Inventory Record (IR) and in the c-BOM structures, prevents and corrects discrepancies in the entered data. This environment allows the maintenance, synchronisation and version-control for (IR) and c-BOMs across multiple users and companies that potential share manufacturing resources [23]. Finally, the scheduling charts and the calculated performance indicators associated with specific facility conditions are stored.

5. Case study from the textile industry

The performance of the i-MRP production planning and control system has been studied through a set of simulation experiments in a vertically organised European textile industry. The under study company operates in the woollen textile system and its product range includes a wide variety of yarns for clothing, carpeting, knitting and wool/synthetic carpets. The proposed i-MRP system has been applied to the production line of blend carpets. The selected production line consists of three discrete departments, namely the Dyeing,

The first processing stage in the production line of blend carpets is the dyeing of the required quantities of raw materials (fibre mass) in the appropriate colours. This operation is performed in the Dyeing job shop. The second stage is the spinning of the required quantities of yarn types in the Spinning job shop. Finally, the third processing stage is the weaving of the carpets that is performed in the Weaving job shop. However, depending on the yarn type, the exact material flow may vary in the Spinning job shop.

Figure 3. The organisation structure of the facility of the textile industry

The workload model of the spinning and weaving job shops selected production area consists of more than 200 different job types for the weaving job shop and 60 job types for the spinning job shop. Each job type in the weaving job shop

corresponds to a single carpet type as defined by its colour set, quality (surface density), shape and dimensions. Each job type in the spinning job shop corresponds to a single yarn type as defined by its colour, quality (type of fibres selected) and title.

6. Results and Discussion

A set of simulation experiments have been conducted in order to validate the efficiency of i-MRP. Real data were collected from the sales department of the under study textile industry, covering a planning horizon of 60 days, plus 30 days' data for simulation initialisation purposes. In all different scenarios, 147 weaving job orders and 104 spinning job orders were scheduled following the BS procedure, resulting in more than 1,600 tasks, assigned to 54 resources in every simulation run. The number of 14 different assignment policies that was employed resulted into more than 22,400 task assignments. The production facility operates two shifts a day, six days a week. The mean capacity utilisation level was kept constant at 75% in all experimental simulations. The relative performance of 14 different assignment policies was evaluated, for the same workload, through a set of scheduling performance indices. The mean values of tardiness, queue time, flowtime, earliness, and reserve time, were used as performance indicators for evaluating the obtained detailed shop floor schedules. The formulas for the calculation of the mean values of these indices are included below:

ME(t,) = : 1


^ max (o,


Tardiness MT(1n) = —• £ max (o, t,comp - t"")

Reserve time

MR(t„) =

Ncolm> 1

•l(r -1," )

Flowtime MF(t ) = • £ (tr" " t," )

NcomP 1

Queue time MQ(t, )=N^ • z l(tr - j )

¡=1 j=1


comp dd

is the number of completed jobs up to time is the completion time of Job i is the due date of Job i is the arrival time of Job i is the start time of Job i is the start time of Task j that belongs to Job i is the arrival time of Task j that belongs to Job i is the time point when indices are calculated

The same operational policy was assigned to all work-centres in each simulation scenario. Four different adaptations of the multiple-criteria decision making technique (MULTI1-4), proposed in [20], were introduced. While the specific criteria in each of them were kept the same, their relative importance in the decision making process varied through the adjustment of their weight factors, wc, wf, and wt, respectively. The weight factor triples for the first configuration were MULTI1: (0.2, 0.4, 0.4), in the second MULTI2: (0.1, 0.8,

0.1), in the third MULTI3: (0.1, 0.1, 0.8), and in the fourth MULTI4: (0.8, 0.1, 0.1).

The setting of orders' due dates can derive form the delivery times promised to customers, MRP processing or managerial decisions based on various due date setting policies [24][25]. In this study, job due dates (DDj) were calculated using the number of operations rule (NOP), as follows: DDj=ADj+kNj, where ADj is the arrival date and time of job j, k is the allowance factor in days and Nj is the number of tasks of job j. Since time in queue is usually the largest component of a job's lead time, the number of tasks comprising it can be used as an indicator of the required flowtime. The value of k was set at 0.7. This configuration results in a set of relatively tight due dates. The selection of such a tight condition was based on the premise that the relative performance of different operational policies can be depicted more clearly in tight due date environments. Also, tight due dates can provide a competitive advantage by allowing the firm to offer an improved level of customer service, as well as achieve lower costs through reductions in WIP inventory.

The customer service level were used as indicator, calculated by the following formula:

( Ntardy ^

Service level SL(t„ ) = I 1 - — I • 100 % where:

N""al is the total number of scheduled job orders up to time Ntai-dy is the number of late job orders up to time tn is the time point when indices are calculated

Results in terms of mean tardiness, mean flowtime and mean queue time are presented in Figure 4 that follows, and results regarding mean earliness and mean reserve time are presented in Figure 5. The actual values from the experiments are included in Table 3.

Table 3. Values of performance indicators as derived from the experiments

Mean Tardiness

Mean Queue Time

Mean Flowtime

Mean Earliness

SPT 18:17:46 60:19:01 101:20:37 4:59:07

EDD 10:31:13 60:24:03 96:45:23 5:19:03

FIFO 9:40:57 59:09:03 94:07:21 5:26:42

LIFO 26:20:40 68:47:16 110:17:44 5:07:47

MOPNR 40:13:31 93:59:07 128:04:02 6:01:02

FASFS 10:04:58 60:28:08 95:30:54 5:17:11

LPT 13:03:49 68:13:17 99:14:46 6:06:51

FOPNR 10:52:34 59:15:53 94:59:12 5:14:55

LWRK 11:03:30 52:57:03 91:24:25 5:00:54

MWRK 27:24:10 80:58:15 115:03:59 6:03:19

MULTI1 10:14:01 57:58:58 92:59:28 4:59:24

MULTI2 14:29:24 55:02:36 96:32:43 4:44:30

MULTI3 10:19:26 61:01:17 94:56:20 5:18:28

MULTI4 15:11:52 64:29:40 100:25:34 5:27:42

The four variations of the multi-criteria decision making technique (MULTI1-4) and the LWRK and FOPNR dispatching rules produced the best results, in terms of mean tardiness, mean queue and mean flowtime, in general. The exceptional performances of MULTI1-4 can be attributed to

the fact that they directly address due dates and attempt to minimize lateness. As it was expected, the MOPNR, MWRK and LIFO rules performed poorly with respect to these measures, in order of deficiency. The reason is that these rules try to promote orders that are less likely to finish on time due to their number of operations, work remaining and late arrival, respectively. Moreover, time in queue accounted for a large part of an order's total flowtime in all experimental scenarios. The first three variations of the multi-criteria decision making technique MULTI1-3 and the LWRK rule restricted queue times in front of workcenters.

Reserve time is defined as the time difference between an order's arrival and its actual start time. It can be utilised as an indicator of a schedule's flexibility, or else its ability to reserve capacity in the near term in order to be able to respond more efficiently to new customer demands or rush orders. High mean reserve times also correspond to low WIP inventories. This is a basic advantage of the BS function due to the fact that it attempts to minimize jobs' slack times and thus it produces high reserve times. This effect would be more obvious in the case of a more relaxed due date setting (k>0.7).

144:00:00 120:00:00 96:00:00 72:00:00 48:00:00 24:00:00 0:00:00

■ Mean Tardiness Mean Queue Time HMean Flowtime

Figure 4. Mean tardiness, mean queue time and mean flow-time






n n ppppppppppppp

Q O O OS Q u- u- Z

en ^ ^ <h <vi m «t

Z OS Q£ i_ i_ i_ i_

a. § § □ □ □ □

O 3 i 3 3 3 3

5 5 5 5 5

finished very close to their due dates, which is consistent with the 'pull' production scheduling concept. The mean earliness indicator has a direct correlation with inventory performance, since low slack times usually correspond to just in time procurements, WIP and end products inventories reductions. Generally, those operational policies that produced a high mean reserve time, also produced a low mean earliness and vice versa.

The derived customer service levels in the tight due dates setting (k=0.7) can be characterised as unsatisfactory in all experiments, as shown in Table 4. However, the main objective was to reveal the relative performance of different assignment policies in strict conditions. The LIFO, MOPNR and MWRK heuristics produced the poorest performances, where service level fell below 75%. It should be noted that these are the same heuristics that also performed rather poorly in the case of the mean tardiness, mean queue and mean flowtime indicators. In the same table the derived mean tardiness to queue time ratio is presented for all experimental scenarios. This is another mean to evaluate different operational policies based on a comparison between the mean job tardiness they induce and the mean time lost in queue. An efficient policy would tend to minimize the time a job stays in queue in each work-centre in order to reduce its tardiness. The most effective policies in minimizing this ratio were MULTI3, EDD, FIFO, FASFS and MULTI1, in descending order.

Table 4. Customer service level and mean tardiness / queue time ratio

Policy Customer Service Level (%) Mean Tardiness / Queue time ratio

SPT 79.50 0.226

EDD 75.37 0.117

FIFO 80.36 0.118

LIFO 74.57 0.341

MOPNR 64.94 0.355

FASFS 79.52 0.130

LPT 77.54 0.146

FOPNR 78.66 0.161

LWRK 83.66 0.178

MWRK 66.59 0.318

MULTI1 80.73 0.140

MULTI2 80.74 0.199

MULTI3 78.92 0.070

MULTI4 76.22 0.164

■ Mean Earliness I Mean reserve time

Figure 5. Mean earliness and mean reserve time

Increased mean reserve times were achieved using the SPT, FIFO, MULTI1 and MULTI2 policies, while the MOPNR, MWRK, MULTI4 and LPT policies provided less flexible schedules. Mean earliness was kept relatively low in all experimental scenarios, owing to BS logic (Figure 5). Orders

7. Conclusions

Integrated material planning and shop scheduling solves material and capacity constraints together. The proposed i-MRP system is an integrated tool for simultaneous detailed scheduling and material planning. Important scheduling performance indices were implemented to assist the selection of the more efficient shop floor schedule and material plan alternative combination. The proposed system is best suited for production planning and control of discrete manufacturing environments, especially for those characterised by batch production processes with high product and volume variety,

where production times are relatively short. In such cases, the expected benefits include an efficient shop floor schedule, smoother flow of materials, and lower WIP inventory. To sum up the results, it has been found through a real-life case study that in order to simultaneously achieve an efficient shop schedule and an efficient material plan, the i-MRP system should be used together with a multiple-criteria decision making policy. Preinstalled MRP-based systems can still be used to effectively control items with relatively lower inventory value. Moreover, in environments where fixed lead times and infinite capacity assumptions are valid, as in the case of Distribution Requirements Planning (DRP) systems, the MRP scheduling logic may still be used with satisfactory results.

Directions for future work include the evaluation of dampening strategies to confront 'nervousness' caused by uncertainty in demand and supply or rescheduling of open orders. Furthermore, special adaptations of the LCC lot-sizing technique are needed in case of deteriorating inventory and where quantity discounts are available or transportation savings are realised when shipping full carload lots.


[1] Chryssolouris G., Manufacturing Systems: Theory and Practice, 2nd Ed.,

2006, Springer Verlag, New York.

[2] Olhager J., Evolution of operations planning and control: from

production to supply chains, Int J of Prod Res, 51(23-24):2013, pp. 6836-6843.

[3] Benton W.C., Shin H., Manufacturing planning and control: The

evolution of MRP and JIT integration. Eur J of Oper Res, 110:1998, pp. 411-440.

[4] Umble E.J., Haft R.R., Umble M.M., Enterprise resource planning:

Implementation procedures and critical success factors, Eur J of Oper Res, 146(2):2003, pp. 241-257.

[5] Damand D., Derrouiche R., Barth M., Parameterisation of the MRP

method: automatic identification and extraction of properties, Int J of Prod Res, 51(18):2013, pp. 5658-5669.

[6] Lapiedra R., Alegre J., Chiva R., The importance of management

innovation and consultant services on ERP implementation success, The Service Ind J, 31(12):2011, pp. 1907-1919.

[7] Panorama Consulting, 2014 ERP Report, A Panorama Consulting

Solutions Research Report, Denver, Colorado.

[8] Humphreys P., McCurry L., McAleer E., Achieving MRPII Class A

status in an SME. Benchmarking: An Int J, 8(1):2001, pp. 48-61.

[9] Plossl G.W., Orlicky's Material Requirements Planning, 3rd Ed., 2011,


[10] Rahmani D., Ramezanian R., Fattahi P., Heydari M., A robust

optimisation model for multi-product two-stage capacitated production planning under uncertainty, Appl Math Model, 37(20-21):2013, pp. 8957-8971.

[11] Wu T., Shi L., Mathematical models for capacitated multi-level

production planning problems with linked lot sizes, Int J of Prod Res, 49(20):2011, pp. 6227-6247.

[12] Dobson G., Karmarkar U.S., Planning Production and Inventories in the

Extended Enterprise, Int Ser in Oper Res & Mgmt Sci, 152:2011, pp. 1-14.

[13] Aouam T., Uzsoy R., Chance-Constraint-Based Heuristics for

Production Planning in the Face of Stochastic Demand and Workload-Dependent Lead Times, Dec Pol for Prod Net, 2012, pp 173-208

[14] Pandey P.C, Yenradee P., Archariyapruek S., A finite capacity material

requirements planning system, Prod Plan & Con, 11:2000, pp. 113121.

[15] Ho J.C., Chang Y.L., An integrated MRP and JIT framework. Com and

Ind Eng, 41(2):2001, pp. 173-185.

[16] Koh S.C.L., Saad S.M., Padmore J., Development and implementation of

a generic order release scheme for modelling MRP-controlled finite-capacitated manufacturing environments. Int J of Comp Int Mfg, 17(6):2004, pp. 561-576.

[17] Iranpoor M., Fatemi Ghomi S.M.T., Mohamadinia A., Earliness

tardiness production planning and scheduling in flexible flowshop systems under finite planning horizon, App Math & Comp, 184(2):2007, pp. 950-964.

[18] Smith N.R., Limon Robles J., Cârdenas-Barron L.E., Optimal Pricing

and Production Master Planning in a Multiperiod Horizon Considering Capacity and Inventory Constraints, Math Prob in Eng, 2009, Article ID 932676

[19] Georgiadis P., Politou A., Dynamic Drum-Buffer-Rope approach for

production planning and control in capacitated flow-shop manufacturing system, Comp & Ind Eng, 65(4):2013, pp. 689-703.

[20] Chryssolouris G., Lee M., An Approach to Real-Time Flexible

Scheduling, Int J of Flex Mfg Sys, 6(3):1994, pp. 235-253.

[21] Mourtzis D., Doukas M., Psarommatis F., A multi-criteria evaluation of

centralised and decentralised production networks in a highly customer-driven environment, CIRP Annals-Mfg Tech, 61(1):2012, pp. 427-430.

[22] Mourtzis D., Doukas M., Psarommatis F., Design and Operation of

Manufacturing Networks for Mass Customisation, CIRP Annals - Mfg Tech, 63(1):2013, pp. 467-470.

[23] Wang L., Machine availability monitoring and machining process

planning towards Cloud manufacturing, CIRP J of Mfg Sci & Tech, 6(4):2013, pp. 263-273.

[24] Kuroda M., Shin H., Zinnohara A., Robust scheduling in an advanced

planning and scheduling environment, Int J of Prod Res, 40(15):2002, pp. 3655-3668.

[25] Saad S.M., Picket N., Kittiaram K., An integrated model for order

release and due-date demand management, J. Mfg Tech, 15(1):2004, pp. 76-89.

[26] Lalas C., Mourtzis D., Papakostas N., Chryssolouris G., A Simulation-

Based Hybrid Backwards Scheduling Framework for Manufacturing Systems, Int J of Comp Int Manuf, 19(8):2006, pp. 762-774.

[27] Chryssolouris G., Papakostas N., Mourtzis D., A Decision Making

Approach for Nesting Scheduling: A Textile Case, Int J of Prod Res, 38(17):2000, pp. 4555-4564.

[28] Papakostas N., Mourtzis D., Makris S., Michalos G., G. Chryssolouris,

An agent-based methodology for manufacturing decision making: a textile case study, Int J of Comp Int Manuf, 25(6):2012, pp. 509-526

[29] Xu X., From cloud computing to cloud manufacturing, Robotics and

Computer-Integrated Manufacturing, 28(1):2012, pp. 75-86.

[30] Li B.-H., Zhang L., Wang S.-L., Tao F., Cao J.-W., Jiang X.-D., Song

X., Chai X.-D., Cloud manufacturing: A new service-oriented networked manufacturing model, Computer Integrated Manufacturing Systems, 16(1):2010, pp. 1-7+16.

[31] Mourtzis D., Doukas M., Vlachou K., Xanthopoulos N., Machine

availability monitoring for adaptive holistic scheduling: A conceptual framework for mass customisation, 8th International Conference on Digital Enterprise Technology - DET 2014, March 25 - 28, 2014, Stuttgart, Germany, ISBN: 9783839606971.

[32] CAPP-4-SMEs, Collaborative and Adaptive Process Planning for

Sustainable Manufacturing Environments - CAPP4SMEs, EC Funded Project, 7th Frammework Programme, Grant Agreement No.: 314024.

[33] Tapoglou N., Mehnen J., Doukas M., Mourtzis D., Optimal tool-path

programming based on real-time machine monitoring using IEC 61499 function blocks: A case study for face milling, 8th ASME 2014 International Manufacturing Science and Engineering Conference, June 9-13, 2014, Detroit, Michigan.

[34] Lewis G., Basics about Cloud Computing, Software Engineering

Institute, Carnegie Mellon University, September 2010, Fifth Avenue, Pittsburgh.

[35] Global Cloud Security Software Market, Technavio, Accessed online on

09-05-2014, URL: