Process flow and operations management

within food-processing factories (e-Track)

Muna Hamdi♣, Alireza Mousavi and Mansoor Sarhadi

School of Engineering and Design, Brunel University

Abstract
This paper describes the work currently undergoing as part of DEFRA♠sponsored feasibility study ‘Resource
Management and Product Traceability (e-Track)’. This project addresses the process flow and operations
management within the current food industry. It aims to improve the food production process by providing
techniques for real-time resource management, product quality assessment and traceability. This is to be
achieved by Firstly, to gain an understanding of the interdepartmental and organisational relationships and
interdependencies within the food processing industry. Secondly, to identify key factors, that affects the entire
production process, and the logical relationship between them. Thirdly, to investigate the feasibility to provide
suitable methods and propose solutions in the form of a generic suite of intelligent decision and
management support tool. The transformation must be gradual to avoid any prohibitive costs and
consequences. It should take into account the current technologies, moving gradually towards efficient,
automated, adaptable, cost effective and user friendly systems.

1 Introduction and literature survey

Meeting efficiency and productivity targets under low profit margins whilst conforming to ever-
increasing customer demands are the challenges facing today’s food manufacturers. The interrelated
factors of raw material, final product and process quality have been the focus of research and
commercialisation of many decision-making tools and techniques from shopfloor data acquisition to
enterprise resource planning systems. These tools have been developed to assist system managers
to deal with problems associated with such volatile and often conflicting needs. Although integrated
systems like SAP have attempted to act as a comprehensive decision-making and production
management toolkit, serious limitations continue to exist.
The key efficiency indicators that contribute to high labour and processing overheads can be
specified as: tedious manual quality checks, labour force absenteeism, product yield, stoppages due
to product or production anomalies or changeovers, and material handling systems. A preliminary
analysis conducted by one of the industrial partners shows that continuous improvements in labour
skills through training and plant efficiency will result in savings of up to £12m in five years. For
example, due to machine and manual quality check shortcomings in a typical sandwich line, on
average, 10% of products that contain labelling errors are rejected by the supermarkets. This
contributes to a costly wastage through product recalls and penalties imposed upon manufacturers.
Traceability is also a major concern for all supermarket suppliers as it provides the only serious
opportunity for due diligence defence.
Experience to date shows, that attempts to solve production efficiency issues using standard ERP and
shopfloor data acquisition technology have had little success, mainly due to its generality and
complexity [1]. However, ERP systems do provide powerful solutions for retrieval, dissemination, and
preliminary process of data.
It seems that a balance between appropriate technical innovation and advanced management
techniques needs to be struck. The development of an intelligent electronic system that links
operational experience with the interacting elements i.e. resources to identify and offer practical
solutions to deal with major anomalies and non-conformances is desirable.

♣ Email: [email protected]
♠ Department for Environment Food and Rural Affairs
Mermelstein[2000] gives an overview of the current trends in software for the food industry,
covering intelligent sensors, process modelling and simulation. The main focus here is on two main
issues. Firstly, to develop where possible, data acquisition and transfer techniques to reduce
dependency on paper work. Secondly, to enable manufacturers to use market forecasts and
consumer requirements for reduced wastage and resource utilisation. With resources we mean
operators, machines, material handling systems and logistics.
The decision-making process needs to incorporate a number of factors and their associated data,
where some of these factors may be conflicting. Hence, it is important to identify suitable algorithms
that can deal with large number of highly correlated variables as well as the ability to account for
deterministic and random behaviour of a volatile manufacturing environment. Cho et al.[2003]
explain that intelligent algorithms such as Bayesian and Neural Networks, are now preferred as a
data fusion technique, used to identify the underling relationships between common and unique
attributes. Their experiments showed that AI-based approaches generally outperform the correlation
approach, improve predictability, and that the relative performance of AI-based approaches tends to
improve as the amount of missing data increases.
As well as optimising and controlling production process, other factors such as quality control and
traceability [11] are now taking major priority within food industries. Yi et al.[2002] include quality
assurance in their proposed online-automatic shopfloor control system. The system is responsible for,
matching sensory and quality inspection data, Knowledge acquisitions and monitoring, process
diagnosis and optimisation for quality improvement, and scheduling to satisfy objectives such as
timely product delivery. The paper addresses large scale industries, where an efficient method for
optimal parameter adjustment is required for product quality, due to large number of controllable
and uncontrollable (such as ambient temperature) process parameters. To achieve process
optimisation, the system fixes the uncontrollable factors and then solves nonlinear programming to
minimise quality variations. The authors identify three main problems: First, the difficulty to reach
high quality of prediction using “short and fat” data sets, with large number of variables (columns)
but few observations (rows). Second, the difficulty for operators to keep changing these parameters,
therefore, the number of changing parameters must be minimised. Third, large-scale optimisation
can’t be adequately solved in real-time. It is unclear if the quality inspection process is done
continuously through the production process or only restricted to final product assessment.
Therefore, it is important to identify and prioritise key parameters in order to reduce the number of
changes needed to achieve the best in terms of process performance and product quality all through
the production process. Moreover, data update and preliminary processing should become as
automated as possible. Once key parameters are identified, process simulation can then be used to
predict future performance and evaluate alternative schedules.
In dynamic and volatile industries, fast scheduling and resource distribution systems are required to
deal in real-time with unexpected events, such as rush orders and machine malfunctions. Hence,
multiple rules scheduling/rescheduling is preferred to fixed scheduling as it become invalid over time.
Drake et al. [1995], describes a Multi-Pass Expert Control System, where the production cycle is
divided into several intervals separated by decision points. At each decision point, the current factory
status is evaluated and a series of deterministic simulations are run to determine which control rule
provides the best performance. The simulation is divided into process simulation modules, one for
real-time task generator, and the other for analysis or look-ahead mode. This allows simultaneous
task distributions and schedule comparison, to avoid any production delays during the analysis
process. To allow for ‘plug and play’ analysis, the simulation, planning, scheduling and execution
functions are usually separated within a flexible manufacturing systems (FMS). As a result, the
execution functions are dependent only on the physical configuration of the equipment and not on
the specific parts and production volumes currently required[6]. Simulation is particularly preferred
for its ability to capture complex interactions within a particular FMS, where analytical methods
fail[7].
Other management aspects include customer orders negotiations and supply chains. Choi et
al.[2003] propose a multi-agent based negotiation system to be part of modern and dynamic
electronic commerce. Mula et al.[2003 and 2004], takes the automobile industry with its complex
supply structure as an example to present a tool that can detect and control the demand variations
and its costs along the supply chain in order to decrease uncertainty (inventory excess or shortages,
urgent transport and overtime) as well as obsolete part costs.
In this paper, the proposed system should be capable of utilising the existing software and hardware
tools within the factory, and where required offer novel modelling and technological solutions to
provide production management with workable action plans for the rapidly changing production
conditions.
The following sections explain production problems within the food industry and identify the key
influential points that affect efficiency and quality of production, and suggest possible solutions.

2 Key influential factors within a food processing factory

In order to have intelligent, adaptable, user friendly and cost effective solutions for optimising the
existing food manufacturing systems, key elements contributing to high labour and processing
overheads within the food production industry needs to be identified, taking into account:

• the existing technology currently used in the management of a food factory;

• key performance indicators that affect the entire production process and the relationship
between them. For example, wastages and yield, resource availability, stoppages, quality,
traceability and data accuracy.
The focus is on observing key operations that affect the efficiency of the production process and the
quality of the end product. Several food processing factories were examined and the following
influential factors were identified:
1. The number of customers dealing with each factory. Having more than one customer
increases the complexity of the production process. Some customers insist upon using their own
approved suppliers and may request more stringent checks on product quality. Hence, the
prediction and scheduling process increases in complexity with the number of customers. It is
therefore important to have the in-built capability to assess product acceptability within the
communication channels between the consumer and producer [12].
2. The type of product, warm, spicy, cold and sandwiches. Production of warm meals peaks in
winter while chilled food production is mainly for the summer season sales.
3. Day of the week, month and season. More labour is required for peak production seasons.
Depending on the management policy and their flexibility the distribution of skilled staff within
nearby company sites is one way of keeping skilful staff working in the company.
4. Product promotions, sales and consequently production may also increase with the
introduction of promotions or new products.
5. Management policy and the level of interaction between different management hierarchies.
6. The variation in ingredients and raw material, for example, in size and texture due to the
season or their origins. This may affect a number of factors such as their weight, taste, oil
absorption when cooking or frying and cooking times. The change over between products varies
depending on product type and the variation of ingredients. For example, if both products use
the same type of ingredients the machines and trays may not need cleaning.
7. The number of staff available, particularly skilled staff.
8. Maintenance policy, machine and spare part stock management.
9. The type of communication link with a customer. For example, the forms in which orders
and complaints are received. Some customers use Internet based electronic link EDI, others still
depend upon Royal Mail. Compatibility issues must be considered when using on-line
applications.
10. The type of software packages used, ease of use and adaptability to changing requirement.
11. The degree of automation and modernisation, which includes.

• Data quality and management. The collection of data using paper forms affects their
accuracy upon which prediction and forecasting models are highly dependent. Prediction
quality is of high importance to avoid delays and wastages. Even when modern production
lines are available, the product records produced by those machines are not saved or
connected to a data network for real-time or later analysis. Reference figures such as
product process time are reviewed every three to six months depending on the current and
past production figures. Excel and Access are the two mainly used software packages;
information are usually shared by more than one department, at different levels of
management and by more than one site within a company. However, there is a need for
departments to use different levels of detail from; summarised to fully comprehensive
depending on their hierarchy and requirement.

• Machine age and type Each factory seems to have an accumulation of machines and
production lines of different age and type. This affects the number of staff working on a
line as less automated lines may require more labour. The type and age of the machines
dictate the type and amount of product that a line can produce.

• Resource management Real-time scheduling and resource management are done mainly
on the factory floor by production, shift and line managers. The operation is highly complex
and includes high degrees of predictability. It depends mainly on the manager’s policy,
experience and capability to solve problems. A number of factors influence this process
such as machine type, stocks availability, number of staff available, (as well as skilful staff)
and the type of product, as similar less messy products are preferred to start with and this
affects change over times.

There is lack of maintenance either due to continuous production; hence little time is left
for maintenance, or due to shortages in skilled engineers. As a consequence, machine
malfunction or stoppages causes wastages in material and time. The decision as to when a
machine should be maintained or when spare parts get ordered seems totally dependent on
human observation skills.

Production managers are challenged by the difficult task of incorporating all related information in a
decision-making process while managing resources and solving various problems that involve
several departments and different issues.
The proposed e-track system will be capable of:
• Automatically validating and verifying data presented to higher layer of information system.
• Providing high quality prediction to help assessing if a particular decision is the best choice,
considering all related circumstances.
The type, availability, complexity, and the extent of information vary from one management level to
another. Having data available on-line and in real-time helps reporting down times and monitoring
wastages. It allows visibility, faster and more efficient accounting and purchasing process, as well as
stock and store management.
Most factories are now moving towards using SAP, which has to some extent improved shop floor
management. The ERP system has also improved inter-organisational interactions. These include
accounting and purchasing, where SAP acts as a control centre between different departments.
However, SAP is only as good as the data supplied to it, hence, collection of accurate, validated and
properly formatted data is crucial to making timely decisions.
Modern ERP systems are a first major step towards modernisation, but in order such systems to
properly work data accuracy and complexity issues need to be addressed.
3 Proposed decision Support and management system
This project aims to create a technology that utilises the combination of field data acquisition and
information modelling for a predictive shopfloor monitoring and control system.
Once a full appreciation of shopfloor activities is obtained, it will enable us to tackle the following
issues:
Identification of key elements for efficiency improvement. Once the key elements are
recognised, a cause-effect modelling blueprint will be established for transferring meaningful
data from shopfloor to higher levels of information system.
Development of an online systems analysis tool to assess the current shopfloor condition and
offer an ahead of time picture of the system for improved scheduling and resource
assignment.
Measures will be taken to evaluate hardware and software technology applications e.g.
Object-Oriented programming philosophy to modularise and automate software elements and
its interface with data acquisition equipment.
Shopfloor product tracking and traceability is now becoming an increasingly important factor
for the players in the entire value chain. Rejected products resulting in wastage are due to
labelling and traceability issues. e-Track will study the feasibility of introducing new
practicable and affordable methods for challenges facing product traceability in the shopfloor.
Finally, this feasibility analysis will assess where applicable the use of wireless data acquisition
and monitoring tools and two-way communication devices for real-time shopfloor control and
scheduling applications.
Where uncertainty issues apply, the feasibility of using advanced predictive techniques such as belief
networks, simulation, and/or neural fuzzy analysis will be studied to make the data analysis and
prediction mode as close to reality as possible. The models will be validated and verified using a
combination of historical and real-time data capture.
The proposed system should be able to selectively and intelligently process critical data from the
shopfloor and other sources of data and information such as customer demands and weather
forecasts. The data are then refined and used to update the simulation and scheduling processes, to
accurately predict near-term performance such as production changes, breakdowns, and resource
status. As well as production floor management the Decision Support System extends to other highly
influential operation managements such as inventory control, human resources and machine
maintenance. Figure1, gives an illustration of the Decision Support and Management System.

4 Conclusion
The main objective of the proposed project is to provide a robust decision maker for food
manufacturing which will turn myriad of interlinked information into action plans leading to an
improved trading result whilst production is still underway. System integration requirements covering
traceability technology and management tools will be identified to enhance monitoring, decision-
making and production quality.
Although the work takes into account the full picture of the future food industry, the transformation
must be gradual to avoid any prohibitive costs and consequences. It should take into account the
current technologies, moving gradually towards efficient, automated, adaptable, cost effective and
user friendly systems.

5 Acknowledgement
This project is funded by DEFRA and supported by our partners, Northern Foods and Harford Control.
 Decision Support and Resource Management

Prediction
Scheduling Multi-pass Simulation

Information Constructor

Data Acquisition

Shopfloor Data Sources

e.g. Customers, organisation and
Environment

Figure 1 Decision support and management system

6 References
1 Vogt C.,”Interactable ERP, a comprehensive analysis of failed enterprise-recourse-planning projects”,
Software Engineering Notes ,vol. 27, no. 2, pp. 62-68, March 2002.
2 Mermelstein N. H., “Software for food processing”, FOODTECHNOLOGY, vol. 54, no. 1, pp. 56-60, January
2000.
3 Cho S., Baek S. and Kim J.S., “Exploring artificial intelligence-based data fusion for conjoint analysis”,
Expert Systems with Applications, Elsevier, 24, pp. 287-294, 2003.
4 Yi G., Shin J., Cho H. and Kim K., “Quality-oriented shop floor control system for large-scale manufacturing
processes: functional framework and experimental results“, Journal of Manufacturing systems, vol. 21, no.
3, 2002.
5 Drake G.R., Smith J.S., and Peters B.A., “Simulation as a Planning and Scheduling Tool for Flexible
Manufacturing Systems,” Proceedings of the 1995 Winter Simulation Conference, Washington DC, pp. 805-
812, 1995.
6 Smith, J. S.,"Design and Implementation of FMS Control Software," Proceedings of the 1996 FAIM
Conference, Atlanta, GA, pp. 859-867, 1996.
7 Peters B. A. and Smith J. S.,” Real-time, simulation-based shop floor control” SAT, simulation and
Automation Technologie GmbH, Gesellschaft für IT-Consulting und System-Integration.
8 Choi H. R., Park J. P., Kim H. S., Park B. J., and Park Y. S., “Multi-Agent based negotiation support systems
for order based manufacturers”, Proceedings of the 5th international conference on Electronic commerce,
Pittsburgh, Pennsylvania, pp. 479-487, 2003, ISBN:1-58113-788-5.
9 Mula, J., Poler, R., García, J.P., and Ortiz, A., 2003a, Supply planning and demand behaviour in an
automobile industry supply chain. International Conference on Industrial Engineering and Production
Management (IEPM’03), Porto (Portugal), Mayo, 26-28.
10 Mula, J., Poler, R., and García, J.P., “Supply Chain Production Planning in a Mass Customization
Environment”, Second World Conference on POM and 15th Annual POM, Cancun, Mexico, April 30 – May 3,
2004.
11 A. Mousavi, M. Sarhadi, A. Lenk and S. Fawcett (2002), “Tracking and Traceability in meat processing
industry: a solution”, British Food Journal, Vol. 104, No. 1 pp. 7-19.
12 Mousavi, A., Adl, P., Rakowski, R.T., and Gunasekaran, A. (2001), “Customer Optimisation Route and
Evaluation (CORE)”, International Journal of Computer Integrated Manufacturing, Volume 14, Number 2,
pp. 236 – 243.