Professional Documents
Culture Documents
Expert Systems With Applications: S.I. Lao, K.L. Choy, G.T.S. Ho, Y.C. Tsim, T.C. Poon, C.K. Cheng
Expert Systems With Applications: S.I. Lao, K.L. Choy, G.T.S. Ho, Y.C. Tsim, T.C. Poon, C.K. Cheng
a r t i c l e i n f o a b s t r a c t
Keywords: Food safety plan is being promoted in the food industry by the Hong Kong Government as a preliminary
Food safety quality control tool. However, it appears to be a challenging task for Distribution Centers (DC) that han-
Receiving operations dles food inventory since most of them are lack of knowledge and know how technology to manage infor-
Operating procedures mation in a real time base. This paper proposes a Radio Frequency Identification based Food Operations
RFID
Assignment System (RFID-FOAS) to help DC facilitates the food safety control activities in receiving areas
Case-Based Reasoning
by generating a proper safety plan. The system has adopted the Radio Frequency Identification (RFID)
technology and the Case-Based Reasoning (CBR) technique to facilitate the inventory data-capturing pro-
cess and assist in formulating decisions, respectively. The developed system aims to help reduce the dif-
ficulties in safety plan development using a knowledge-based expert system. The significance and
contribution of RFID-FOAS in the context of managing the inventory quality in DC for safety plan devel-
opment is demonstrated through the adoption of the system in a Hong Kong-based logistics company.
The generated results show that the decision-making process of the safety plan development is facili-
tated. Moreover, the real-time data capturing nature of RFID technology has further improved the effi-
ciency and timeframe requested for the actions. With the support of RFID-FOAS, the data capture
system and the decision-making time is minimized. As a result, inventory quality and customer satisfac-
tion level are significantly improved.
Ó 2011 Elsevier Ltd. All rights reserved.
0957-4174/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2011.08.105
S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548 2533
Fig. 1. Operation flow of raw food supply chain and receiving operation in a food distribution center.
medium-scale companies have difficulty in applying the opera- need to distinguish the critical control points among the control
tions and duties required to implement HACCP. Jobs required in- points. Finally, specified monitoring procedures and corrective ac-
clude hazard analysis, identification of critical control points, tions are required to be identified.
controlling and monitoring identification procedures, handling With the existing approach of a safety monitoring decision-
and keeping track of records, launching of workforce training pro- making process, DC managers face problems along the food safety
grams, and defining maintenance and waste disposal procedures plan implementation. In general, two major problems were found
(ISO/TS 22004, 2005). To monitor food quality, the government to exist in the process of formulating a food safety plan in the food
has emphasized the necessity of developing a food safety plan DC.
based on the principle of HACCP system for internal assessment
and control. Based on the food safety plan, companies can address
food-related problems, as well as aid the implementation of HACCP 1.1. Existence of too many parameters that should be considered in
in the future. formulating decisions
In Fig. 2, a food safety plan consists of potential hazards, control
points, critical control points, control limits, monitoring proce- Currently, different SKUs have different potential hazards; in
dures, and corrective action (HKSAR, 2002). In formulating these addition, based on the potential hazards, different control points
elements, a range of actions are required, such as analyzing the are identified. However, numerous parameters and rules are
potential hazard of food product, identifying all check points of required when analyzing elements in food safety plans. These
operations, identifying critical food safety check points, etc. Tradi- parameters include operator skills, operation procedures, company
tionally, a human decision maker is involved in the process of for- rules, food regulations, operation manual, etc. Scattered informa-
mulating a decision on the food safety plan for upcoming tion increases the difficulty of managing the process. Hence, during
inventories. When products are inbound to DC, different Standard the decision-making process, numerous SKU characteristics, such
Keeping Units (SKU) are present in the product lot. Decision makers as color, texture, storage period, storage condition, water level,
are required to decide on the present potential hazards. For each food category, etc. need to be considered. This further increases
potential hazard, different control points must be identified. Then, the complexity of the process. Hence, with numerous parameters
specific control limits are decided. Furthermore, decision makers to be considered, a long timeframe in decision making is needed.
Fig. 2. Existing safety monitoring, decision-making framework based on HACCP (HKSAR, 2002).
2534 S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548
1.2. Human errors occur while analyzing parameters in the more critical parameters, such as quality and safety controls. Foods
formulation of food safety plans easily deteriorate and have a short life span; hence, additional cau-
tion should be placed in maintaining quality and minimizing im-
While formulating the food safety plan, the decision maker is pacts on health (Erkoc et al., 2005; Iijima et al., 1996; Trienekens
involved in identifying the potential hazard, the critical control & Zuurbier, 2007; van Donk, 2001). In recent years, food safety
points among the control points, and specifying monitoring proce- has been managed and judged by society and the media (Henson,
dures, as well in corresponding corrective actions. As quality is Holt, & Northen, 1999; Houghton et al., 2008; Kuo & Chen, 2010).
intrinsically difficult to be defined, bias and errors may be involved Moreover, the increase in the recorded rate of food-borne patho-
in the decision-making process. On the other hand, if different gens and poisoning (Henson et al., 1999; Unnevehr & Jensen,
decision makers are involved in case of labor turnover or absence, 1999) changed the demand and characteristic of the food supply
the decision outcome may vary. As these decisions are interrelated, chain. This pushes DC operators to further consider the advantages
errors made at the beginning would affect decisions executed at of launching food safety management to boost competitiveness
latter stages, which would in turn amplify the negative impacts (Ziggers & Trienekens, 1999) which led to its adoption. In addition,
on the final outcome. The high occurrence of errors would increase with the growth of dyadic quality performance, companies aid the
the number of substandard goods and customer complaints. development of mutually agreed operations and quality require-
In relation to the two previously mentioned problems, a proper ments. Based on the request of the customers, companies imple-
and comprehensive system for food safety management is needed. ment new techniques, all of which are specially-tailored to meet
For this reason, we focus on facilitating food safety controls in DC specific requirements (Yang, Wong, Lai, & Ntoko, 2009). Under this
by proposing a Radio Frequency Identification (RFID) technology- situation, a number of DCs have included production elements
based Food Operations Assignment System (RFID-FOAS). The pro- supported by technologies (Garcia, Chang, & Valverde, 2006; Ross
posed system has adopted: (i) the RFID technology to visualize & Droge, 2002). Therefore, the operations performed in DC are
and gather real-time inventory information for inventory arrival changing to embrace the nature of manufacturing, which highly in-
notification and (ii) Case-Based Reasoning engine (CBR) technology creased its importance in quality control (Rong, Akkerman, &
to arrange inventory characteristic-based safety plan. With the Grunow, in press).
help of RFID-FOAS, a shortlist of critical control points, measure-
ment, and appropriate actions in food handling is generated. In 2.2. Implementation of safety plans based on HACCP and its difficulties
summary, the diversified safety control can be monitored, to facil-
itate the food safety plan implementation and increase the level of Quality is defined as a key element adhered to by consumers to
customer satisfaction. make choices from a wide array of products and features that pro-
The remainder of the paper is organized as follows. In Section 2, vide a closer match to individual desires (Ho, Lau, Lee, & Ip, 2005).
literature related to food DC, food safety control, CBR, and RFID is Per Morris and Young (2000) stated that quality has only been uti-
examined in detail. Section 3 discusses the development and oper- lized in areas of production and that people have always underes-
ation mechanism of RFID-FOAS using RFID and CBR engine. In Sec- timated its importance in other supply chain elements. Nowadays,
tion 4, a case study on one Hong Kong-based logistics company is quality is associated to food safety by customers and is regarded as
presented to illustrate the use of RFID-FOAS in a real life food DC an important element for food industry development (Hughes &
using Case Advisor for CBR system construction. The results and Merton, 1996; Morris & Young, 2000). With regard to food quality
benefits of launching RFID-FOAS are discussed in Section 5. Finally, management, implementing HACCP is suggested to be the first step
the conclusion is presented in Section 6. in the attainment of safety management. HACCP is regarded as a
fundamental step that provides an initial building block and core
element for setting up the future status of the quality assurance
2. Review of related studies
system (Holt & Henson, 2000). As a process-based quality assur-
ance tool, the core concerns of HACCP are placed on the front line
This section begins with a review of literature on food DC,
operation of work stations. This helps ensure that the inventories
which addressed the importance of implementing safety controls
of the food company has taken all preventive measures to attain
in the receiving operations. Hence, studies on current food safety
quality and safety (Westenbrink, Oseredczuk, Castanheira, & Roe,
management and difficulties involved are presented. Discussion
2009; Wruck & Jensen, 1998). As mentioned in the Standard
of the use of expert system and Radio Frequency Identification
E2590 (2009), the HACCP system consists of seven principles.
(RFID) technology in decision making and data capturing, follows.
These are conducting a hazard analysis, determining the critical
The purpose of this review is to provide an understanding of exist-
control point, establishing critical limit(s), establishing a system
ing studies and the rationale for the choice of system and tools in
to monitor critical control points, establishing corrective monitor-
the current study.
ing of critical control points that are not under control, establishing
procedures for verification of the HACCP system performance, and
2.1. Background of food DC establishing documentation of all procedures and records. Compa-
nies have always found the implementation of HACCP complicated
Inventory in DC requires a significant length of long time; and difficult; thus, they fail to establish it. In an effort to monitor
hence, these operations appear to be critical in influencing cus- food quality, as well as guide companies in fulfilling HACCP, the
tomer service and operation lead times (Faber, Koster, & Velde, government has emphasized the development of a food safety plan
2002; Johnston, Taylor, & Visweswaramurthy, 1999). Moreover, based on the principle of HACCP system. In this endeavor, studying
concerns related to controlling DC operations, especially in receiv- necessary operation processes is required. Hence, specific hazards
ing operations are increasing because people believe DCs can de- and consequences must be correctly identified. The scope of HACCP
rive benefits from business operations. For example, the controls ranges from the nature of food safety to incentives of launching
in operations may be tied to the exploitation of growth in demand systems on food safety control (Henson et al., 1999). Experts are re-
of fast supply chains by providing a rapid inventory turnover quested to judge the severity of the hazards and the relationship of
(Christopher, 1998; Aghazadeh, 2004; Baker, 2004). Overall, this the hazards with the critical control points. Currently, all the justi-
leads to increased customer satisfaction. Hence, aside from con- fication processes are based on the knowledge of the workers
trolling physical operations like pick and pack, food DC requires (Westenbrink et al., 2009). On the other hand, customers have
S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548 2535
become more reliant on DC to assist them in performing value- uthu, 2007). In addition, during data capturing, no direct physical
added activities (Hsiao, Kemp, van der Vorst, & Omta, 2010; contact is required. This significantly reduces the actual operation
McPhee & Wheeler, 2006). Therefore, diversified operations further time for product identification and tracking. Basically, the software
increased the complexity and difficulties in the arrangement of and hardware of the RFID technology include a middleware, RFID
safety plans. Quality is intrinsically difficult to define and can re- tags, and readers (Kwok & Wu, 2009). RFID highly increases the
sult in confusion when applied (Morris & Young, 2000). Hence, if convenience in inventory management and inventory-information
the decision-making process solely relies on human beings, the gathering; hence, it is currently widely adopted in handling inven-
quality and the reliability of the results are in question. To speed tories, particularly, inventories of food items (Bottani & Rizzi, 2008;
up the decision-making process and to improve the decision qual- Jedermann, Behrens, Westphal, & Lang, 2006; Ngai, Suk, & Lo,
ity properly, the assistance of technology or a new system is 2008; Tan & Chang, in press).
needed. With the above studies, proposing a real-time based system
that incorporates the CBR engine to aid food DC in launching a
2.3. Current application of expert system in food management safety plan during the receiving operation is important.
devices. The process of data storing in the database is shown in 3.3. Tier 3 – data management
Fig. 6. When a list of tag IDs is captured by RFID readers installed
in the workplace, IMM decodes the electronic signals and trans- With the inventory and facility information in the DC and data-
forms the signals into human-readable information, such as order base captured, a decision tree for the retrieval of historical cases is
number, item number, receiving location, and receiving time. constructed based on the nature of the food and the warehouse
Hence, data is stored into the centralized database for retrieval at operation, as illustrated in Fig. 7.
a later time. The database contains fixed information of customer Similar past cases are retrieved to manage newly arrived case.
requirements on food handling and value-added activities perfor- With the help of the CBR engine, the safety plan operation, i.e., food
mance, the nature of operation, and food characteristics. Statistical acceptance, value-added activities, and food storage assignment,
data transmitted by electronic signals, such as inventory arrival can be settled in a systematic way that guarantees food property
information, are stored at the same time. Only a certain proportion is taken care of and that facilitates the safety control arrangement.
of the inventory information is stored in the RFID tag; hence, when Before the actual adoption of the system, operation specifications
the IMM receives the electronic signal, the database is searched to are required to be determined based on the current manual deci-
retrieve the required corresponding information. The collected sion-making mechanism, food nature, and characteristics. Past
information is sent to the CBR engine for further processing. practices are organized in terms of cases for retrieval. Detailed
S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548 2537
operations of the CBR engine consist of four activities: (i) case re- mainly related to the: (a) physical nature of food items, (b) chem-
trieval, (ii) case reusal, (iii) case revision, and (iv) case retention. ical nature of food items, and (c) customer requirements, as shown
in Fig. 8.
After the generation of key attribute list, CBR engines begin to
3.3.1. Case retrieval
browse the case library to retrieve the potential cases. As shown
The case retrieval mechanism of the CBR engine is illustrated in
in Fig. 9, numerous historical cases are stored in a tree-like struc-
Fig. 8. When a case arrives with an order number from the IMM,
tural hierarchy, which applies a method known as the inductive
the system begins to browse the database to request relevant
indexing approach. The inductive indexing approach is an ap-
information. In this process, information on a list of key attributes
proach for clustering cases with different degrees of similarity.
is generated. These key attributes consist of information used to
Only cases in the selected cluster are considered for case reusal
decide the required receiving operations. These attributes are
Fig. 9. Case retrieval mechanism of the CBR engine importance and relationship among attributes.
in later processes. As the number of potential cases is significantly According to Equation (1), the total similarity value is calculated
minimized, the processes of selecting relevant cases are acceler- with support of the pre-defined weightings, and the degree of sim-
ated. In the tree structure, the most important attribute is pre- ilarity among cases can be found automatically. The weightings of
sented in Level 1, and the importance decreases as the level different attributes are defined with reference to the manual selec-
increases. Storage type is considered as the most important attri- tion mechanism of human beings. Hence, based on the weightings,
bute of the operation assignment. In Level 2, cases are grouped the importance and similarities of the attributes, as well as the list
according to the special request assigned by the customer because of ranked retrieved cases, are generated. The case ranked with the
the involved operation or safety control is highly influenced by the highest similarity value is regarded as the most significant base for
requested service. Thus, the value-added activity performance the final solution generation. Other cases with comparatively low-
influences the control points and measurement of the safety plan. er similarity value act as references.
With different food properties and nature, cases are grouped into Pn
different clusters. i¼1 wi simðfiI ; fiR Þ
Pn ð1Þ
When the system browses the case library to retrieve historical i¼1 wi
cases by following the tree structure path, the CBR engine can rec-
where
tify the new case with data from the stored historical cases. Hence,
a group of potential cases in the resulting cluster are chosen and
wi = weight of attribute I;
presented in a list format for case reusal.
min jfiI j; fiR
simðfiI ; fiR Þ ¼ ;
3.3.2. Case reusal maxðjfiI j; jfiR jÞ
With the formulation of the potential case list, the nearest
neighbor retrieval algorithm is adopted to further resolve the most
appropriate case. Food operation handling regulations and rules fiI ; fiR = the value of fi in the input.
are studied; the knowledge gained is then converted into different
pre-defined weightings (wi) to illustrate the importance and rela- 3.3.3. Case revision
tionship among attributes, as shown in the entity relationship dia- When the case with the highest similarity value is suggested by
gram in Fig. 10. the CBR engine, comparison with the suggested case leads to its
Weighting is added to (fi), which represent the corresponding revision and a newly inputted case is preformed. Modification,
attributes of the case. In addition, similarity values of different op- such as editing and combinations are made based on the value of
tions among attributes is defined and inputted into the system. the cases to ensure that the solution fits_the actual needs.
2540 S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548
Fig. 10. Entity relationship diagram that illustrates the importance and relationship among attributes.
(ii) The time allotted in deciding safety control actions is pro- be mounted on the dock door when receiving arriving RFID singles
longed. Hence, the lead time for handling orders is as shown in Fig. 13. Once the readers are mounted, inventories
lengthened. with RFID tags pass the gateway, and the readers record the event,
then pass the signal to the system. Afterwards, RFID tags are at-
4.2. Case illustration – generating a safety plan for the receiving tached to the inventories and pallets of goods. Hence, numerous
process of raw materials and ingredients for sushi Web-based monitors are set up at work stations to display infor-
mation, as shown in Fig. 14.
Raw materials and ingredient for sushi is one of the frequently
handled inventories of the company. Inventories are imported 4.2.2. Step 2: system construction of CBR engine
worldwide and distributed to different local stores in Hong Kong. Preparation is necessary before constructing the CBR engine in
Generally, inventories can be categorized into two types: the direct the Data Management Tier using the Case Advisor. Case Advisor
distribution type, in which no value-added service is needed, and is software developed by Sententia Software at Simon Frazier Uni-
the value-added type, in which value-added production is involved versity, Canada (Wastson, 1997). In Fig. 15, three main components
before storage and distribution. With such features, the company are present: (i) CaseAdvisor Authoring Module, (ii) CaseAdvisor
suffers from the problem of insufficient guidelines on safety con- Problem Resolution Module, and (iii) CaseAdvisor Webserver
trol when handling different value-added services. As entirely dif- Module.
ferent control points are required in different food inventories, the
required decision-making time required is inappropriately long. 4.2.2.1. CaseAdvisor Authoring Module. In the Authoring Module, the
Hence, the unsystematic decision-making process leads to opera- historical safety plan of the operating case and the critical point
tion errors. Raw materials and ingredients for sushi are sensitive assignment decision tree of the CBR engine are constructed in a
to time and environmental conditions; thus, cautious control in portal-like environment. Details of the case and DC operation are
all stages of the operation is required to prevent spoilage and con- inputted to the case library. New case information is created in
tamination. Any delay in operation or inappropriate handling af- Case Advisor by entering the case name, case description, and
fects the product safety and the suitability of latter delivery. the safety plan solution. Attributes that identify questions are con-
Therefore, RFID-FOAS is adopted, using salmon to demonstrate stituted and matched with the corresponding cases. Different oper-
the generation of a specific and customized operating procedure ating attributes, such as storage type, required value-added
for a period of 2 months starting in March 2010. The system em- service, food type, and SKU Dimension, are set by the developer
braced the support of RFID real-time data capturing technology to identify the relationship between the case information and the
and knowledge supported CBR engine using Case Advisor. Fig. 12 required control action. Weightings of the operating attribute that
shows the seven implementation steps of RFID-FOAS. indicate the importance are adjusted. Hence, a decision tree is con-
structed to group similar cases in a systematic way.
4.2.1. Step 1: physical set up of RFID device
4.2.2.2. CaseAdvisor Problem Resolution Module. Identification and
To gather real-time inventory information and to support the
case review of problems are carried out in the Problem Resolution
application of the system, different RFID devices must be set up
Module. The Problem Resolution Module is a case retrieval model
in the DC and in work stations. First, RFID readers are required to
that is used to search and extract potential historical cases. In this
module, operating attributes and their values are inputted. Subse-
quently, the corresponding case questions are displayed. These
questions are used to help users filter historical safety plan records.
Potential case solutions are retracted by the system in accordance
with the operation and critical point assignment decision tree built
in the Authoring Module. Netscape is used to display the resulting
solution.
information is stored in RFID tags; hence, the data retrieval process 4.2.5. Step 5: rank similar cases for case selection
aims to search for other relevant attributes from the database. Re- In this demonstration of the process of receiving salmon, six po-
cords which are retrieved include customer name, list of items, and tential historical cases are found in the resulted cluster. The case
positions of items. These details are received and displayed on the information is then retrieved and passed for case ranking. Cases
web-based platform. are ranked by calculating the similarity value through the CBR en-
gine with reference to Equation (1), which is shown in the System
4.2.4. Step 4: retrieve similar cases for reuse Architecture section. There are different attributes for calculating
Historical cases are stored in the case library with the corre- the similarity value such as, storage type, customer requirement,
sponding attributes. Based on the information on the attributes, SKU measuring unit, food type, supplier location, and SKU dimen-
similar historical cases are retrieved from the case library through sion. Different pre-defined weighting (wi) is defined in Fig. 10, in
the inductive indexing approach. A decision tree is constructed to the System Architecture section.
demonstrate the selection mechanism and to illustrate the case As illustrated in Fig. 20, the potential cases carry different scores
relationship, as shown in Fig. 18. The company designs the food that represent the difference in similarity value among cases. With
safety plan based on four types of order nature. Therefore, there reference to the Total Similarity Value (TSV), cases with the highest
are four levels in the decision tree: Level 1 – Storage Type; Level similarity values are selected as the most preferable cases to be
2 – Customer Requirement; Level 3 – Operation Specification; adopted. In this scenario, a case named ‘‘Salmon 0325’’ with a score
and Level4 – Inventory Specification. Potential cases are arranged of 95 is selected. Consideration is formulated for this case in
into different groups according to the values of attributes, as arranging the operating procedure of the safety plan.
shown in Fig. 18. Then, the system matches the specification of In the facilitation of the analysis of the system mechanism,
the product nature with cases stored in the case library. demonstration of the similarity value calculation between the
To help arrange cases into appropriate groups, different ques- new case and ‘‘Salmon 0325’’ is shown. In Table 1, ‘‘Salon 0325
tions related to the case structure and the decision attributes are consists of six attributes.’’ Reference on the entity relationship dia-
presented in Fig. 19. By addressing the questions and following gram (Fig. 10) is made; the predefined weighting and similarity va-
the searching path, a particular cluster of potential historical cases lue of different attributes are found.
is identified and retrieved. Therefore, by using Equation (1)
2544 S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548
Table 1
Comparison of ‘‘Salmon 0325’’ and a new case.
Pn
simðfiI ; fiR Þ 50 1 þ 30 1 þ 30 1 þ 10 1 þ 10 0:5 þ 5 0:6
i¼1 wi in Fig. 21. Hence, when the salmon is delivered to the customers,
Pn ¼
i¼1 wi
50 þ 30 þ 30 þ 10 þ 10 þ 5 their feedback on the service and product quality are collected
¼ 0:95 and recorded. The customer feedback is used to evaluate the oper-
ation performance, especially in controlling the outcome of the
The total similarity value between the new case and ‘‘Salmon 0325’’ safety plans. If complaint on the salmon quality is received, a more
is obtained. The value is 0.95, which is the same as the system dis- sincere consideration in future case reusal may be required. Thus,
play. Similarly, the total similarity values among other cases and the this reflects the customer satisfaction level. As the number of case
new case are obtained through the same way. reusal cycles increase, the quality of case information improves.
4.2.6. Step 6: case revision and safety plan generation 5. Results and discussion
The case ranked with the highest similarity value is considered
as the most significant base for generating solutions. With refer- To study and verify the significance and contribution of RFID-
ence to the case information of the selected case, modification FOAS in the context of managing the inventory quality in DC to de-
may further improve the suitability of the adoption. Modification velop safety plans, a trial run was undertaken in one of the Hong
can be made by adding the missed operation steps or by removing Kong-based logistics companies. Preliminary work focused on the
any redundant operations to fit the customer needs. Required facil- safety plan operation procedure generation during the receiving
ities and work stations are adjusted automatically along with the operation in DC. The performance and contributions are presented
changes. Furthermore, the operator can refer to the ‘‘Customer and described in this section to demonstrate the significance of the
Feedback’’ section of the historical case to evaluate whether mod- system. The assistance provided by RFID-FOAS included real time
ification of operations should be made to improve the customer inventory arrival data capturing and knowledge base advices for
satisfaction level. decision making. Hence, a number of benefits are identified by
the DC manager during the interview after the trial run. The bene-
4.2.7. Step 7: store new case and customer feedback fits are as follows:
After the revision and case modification, a new case solution to
receive operations is generated. Each case is assigned with a unique The system provides DC operators with solid and detailed
case number. Accordingly, case details with updated information guidelines on inventory handling with regard to the required
are sent to the case library. At the same time, case details are sent operation, critical control points, and corrective actions for cus-
to the Web-based platform through the LAN network, as illustrated tomized safety control.
2546 S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548
The system reduces the labor force involved in decisions regard- and value-added activities, the operations that need to be com-
ing the operation flows and control actions required. The deci- menced before the transfer to the rack are unclear. With the help
sion maker is no longer a human being but an automated of RFID-FOAS, the data retrieving and decision-making operations
computer system. are automated by the IMM and CBR engines. A systematic safety
The system provides consistent control actions in quality assur- plan is suggested. Therefore, human errors involved are signifi-
ance for the same category of food inventory with the same nat- cantly reduced. Hence, with the resulting safety plan displayed
ure and property. Previously, handling is decided based on the on the Web-based user interface, clear guidelines on the operation
experience and tacit knowledge of expert workers and thus, and control action can be illustrated to the operators easily. In
any labor turnover or absence would affect quality of the Table 2, order fulfillment has been improved by 33.85%.
results.
The system assists the reduction in the receiving operation 5.2. Improvement in timeframe for resource assignment
timeframe because the system automates the data capturing
and decision-making process. With the real time data capturing ability of RFID-FOAS, food ar-
The system provides clear handling of history and records of the rival is notified in real time. The relevant inventory information is
operations and control actions, which are stored in the company retrieved by the IMM and sent to the CBR engine automatically. No
database in terms of historical cases. It provides traceable manual data input or filter is required, which significantly reduces
inventory handling records, which are required for the compli- the required timeframe for inventory notification. In Table 3, the
ance of quality assurance scheme like ISO and TQM. The system time required to notify the inventory arrival and data entry is re-
helps the company establish the initial building blocks for qual- duced by over 96% and the significance increases with order size.
ity assurance system. In addition, a noteworthy time saving in the decision making is
achieved. As different operations are required to perform on differ-
Aside from the interview with the company, different compari- ent inventories, a lengthy manual decision time is required for hu-
sons are made to measure and visualize the system performance man beings. RFID-FOAS replaces the human decision process with
before and after the implementation. Based on the quantitative an expert knowledge system, and the suggested safety plan is gen-
measurement, three areas of major improvement are identified. erated automatically in a short timeframe. In Table 3, the decision-
making process for receiving operation is improved by 85.17%.
5.1. Improvement in operation management
5.3. Improvement in customer satisfaction and quality
Before the implementation of RFID-FOAS, manual checking of
the inventory document such as inbound order, receiving list, Before the launch of RFID-FOAS, safety control is decided based
and customer invoice, is required to locate which customer has or- on the experience and tacit knowledge of the workers. Any work-
dered the goods and what kinds of quality control operations are force turnover or absence would affect the decision quality. There-
required. Since the same group of food inventories may be deliv- fore, the inconsistent decision results in the occurrence of
ered to different retailers after the storage period and different substandard goods. Hence, faults in operations lead to returning
retailers may have different requirements on quality checking of goods or service charge deduction. On the other hand, the long
S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548 2547
Table 2 sion-making process. Hence, the real time data capturing nature of
Improvement in operation management. RFID technology further improves the efficiency and timeframe re-
Before After (with Percentage of quested for the actions.
RFID-FOAS) improvement A comprehensive safety control is required in the food supply
(%) chain to satisfy the demands of customers. DC, where food inven-
Operation management tory spends a large proportion of time and performs value-added
Order fulfillment (free of error and 0.65 0.87 33.85 activities, requires urgent development of a safety plan. However,
fault handling/total handling)
the rise in customization of services has made the development a
complicated task. We proposed a real-time food management sys-
tem that integrates RFID technology and CBR technique for DC ser-
Table 3 vice providers in launching a food safety plan, which aims to
Time reduction in receiving operation handling. manage the receiving operation and improve the operation
Before After (with RFID- Percentage of efficiency and customer satisfaction.
(manual) FOAS) improvement (%)
Time
Inventory notification (data entry) Acknowledgment
Single order 1 min 2s 96.67
Ten orders 12 min 3s 99.58
The authors wish to thank the Research Office of The Hong Kong
Twenty orders 27 min 4s 99.75
Decision process for receiving operation Polytechnic University for supporting the project (Project Code:
Operation 5 min RPBK).
involved
Critical control 6 min
point
Corrective 3 min
References
action
Total 14 min 2 min 85.17 Aghazadeh, S. M. (2004). Improving logistics operations across the food industry
supply chain. International Journal of contemporary Hospitality Management,
16(4), 263–268.
Arvanitoyannis, I., & Mavropoulos, A. (2000). Implementation of the hazard analysis
critical control point (HACCP) system to Kasseri/Kefalotiri and Anevato cheese
production lines. Food Control, 11(1), 31–40.
Table 4 Baker, P. (2004). Aligning distribution center operations to supply chain strategy.
Improvement in customer satisfaction and quality. The International Journal of Logistics Management, 15(1), 111–123.
Bakhrankova, K. (2010). Decision support system for continuous production.
Implementation of RFID-FOAS Before After Percentage of Industrial Management & Data Systems, 110(4), 591–610.
(with improvement Bottani, E., & Rizzi, A. (2008). Economic assessment of the impact of RFID
RFID- (%) technology and EPC system on the fast-moving consumer goods supply chain.
FOAS) International Journal of Production Economics, 112, 548–569.
Chi, R. T., Chen, M., & Kiang, M. Y. (1993). Generalized case-based reasoning system
Customer service and quality for portfolio management. Expert Systems with Applications, 6, 67–76.
Customer complaints – slow response, 13 4 69.23 Chou, J. S. (2009). Web-based CBR system applied to early cost budgeting for
fault in operation (number of pavement maintenance project. Expert Systems with Applications, 36,
customer complaints/month) 2947–2960.
Defective inventory (out of order 0.15 0.06 60 Chow, H. K. H., Choy, K. L., & Lee, W. B. (2007). Knowledge management approach in
inventory/total inventory) build-to-order supply chains. Industrial Management & Data Systems, 107(6),
882–919.
Chow, H. K. H., Choy, K. L., Lee, W. B., & Lau, K. C. (2006). Design of a RFID case-based
resource management system for warehouse operations. Expert Systems with
timeframe required in inventory information gathering and deci- Applications, 30, 561–576.
Choy, K. L., Lee, W. B., & Lo, V. (2003). Design of a case based intelligent supplier
sion making, results in a low turnover of inventory and signifi- relationship management system – The integration of supplier rating system
cantly affects customer satisfaction and company goodwill. With and product coding system. Expert Systems with Applications, 25, 87–100.
the help of RFID-FOAS, specification of safety plan is suggested Christopher, M. (1998). Logistics and supply chain, management (2nd ed.). Harlow:
Pearson Education.
for every single order. Fault and errors are reduced significantly.
Cotterill, J. V., Chaudhry, M. Q., Matthews, W., & Wakkins, R. W. (2008). In silico
It is important to note that a faster decision process is achieved assessment of toxicity of heat-generated food contaminants. Food and Chemical
and the operating efficiency and turnover of inventory is improved. Toxicology, 46, 1905–1918.
In Table 4, the customer compliant and defective inventory level is Darder, M., Valera, A., Nieto, E., Colilla, M., Fernandez, C. J., Remedios, R. A., et al.
(2009). Multisensor device based on Case-Based Reasoning (CBR) for
improved by 69.23% and 60%, respectively. Moreover, the adopted monitoring nutrient solutions in fertigation. Sensors and Actuators B: Chemical,
after-service feedback further assists the evaluation process of ser- 135, 530–536.
vice improvement. Erkoc, M., Iakovouand, E. T., & Spaulding, A. E. (2005). Multi-stage onboard
inventory management policies for food and beverage items in cruise liner
operations. Journal of Food Engineering, 70, 269–279.
Faber, N., Koster, R. B. M., & Velde, S. L. (2002). Linking warehouse complexity to
6. Conclusions warehouse planning and control structure: An exploratory study of the use of
warehouse management information systems. International Journal of Physical
We illustrated the system for safety plan development in food Distribution & Logistics Management, 32(5), 381–395.
Garcia, A., Chang, Y. S., & Valverde, R. (2006). Impact of new identification and
DC receiving operation with the support of RFID technology and tracking technologies on a distribution centre. Computers & Industrial
CBR technique, which is named RFID-FOAS. The proposed system Engineering, 51, 542–552.
assists and automates the safety control actions of deciding the Han, J., Trienekens, J. H., & Omta, S. W. F. (2011). Relationship and quality
management in the Chinese pork supply chain. International Journal of
operation required and the critical control point and corrective ac-
Production Economics, 134(2), 312–321.
tion involved in safety control. RFID-FOAS contains a knowledge- Han, S. G., Lee, S. G., & Jo, G. S. (2005). Case-based tutoring systems for procedural
based CBR engine and a user interface for matching the attributes problem solving on the www. Expert Systems with Application, 29, 573–582.
of the food inventory with historical cases. The developed system Henson, S., Holt, G., & Northen, J. (1999). Costs and benefits of implementing HACCP
in the UK dairy processing sector. Food Control, 10, 99–106.
aims to help solve or reduce the difficulties in safety plan develop- HKSAR. (2002). How to implement a food safety plan. Risk communication Section,
ment using a knowledge-based expert system to facilitate the deci- Food and Environmental Hygiene Department.
2548 S.I. Lao et al. / Expert Systems with Applications 39 (2012) 2532–2548
Ho, G. T. S., Lau, H. C. W., Lee, C. K. M., & Ip, A. W. H. (2005). An intelligent forward Perrot, N., Agioux, L., Ioannou, I., Mauris, G., Corrieu, G., & Trystram, G. (2004).
quality enhancement system to achieve product customization. Industrial Decision support system design using the operator skill to control cheese
Management & Data Systems, 105(3), 384–406. ripening – Application of the fuzzy symbolic approach. Journal of Food
Holt, G., & Henson, S. (2000). Quality assurance management in small meat Engineering, 64, 321–333.
manufacturers. Food Control, 11, 319–326. Piramuthu, S. (2007). Protocols for RFID tag/reader authentication. Decision Support
Houghton, J. R., Rowe, G., Frewer, L. J., Kleef, E. V., Chryssochoidis, G., Kehagia, O., Systems, 43, 897–914.
et al. (2008). The quality of food risk management in Europe: Perspectives and Poon, T. C., Choy, K. L., Chow, H. K. H., Lau, H. C. W., Chan, F. T. S., & Ho, K. C. (2009). A
priorities. Food Policy, 33, 13–26. RFID case-based logistics resource management system for managing order-
Hsiao, H. I., Kemp, R. G. M., van der Vorst, J. G. A. J., & Omta, S. W. F. (2010). A pick operations in warehouses. Expert Systems with Applications, 36, 8277–8301.
classification of logistics outsourcing levels and their impact on service Rong, A., Akkerman, R., & Grunow, M. (2011). An optimization approach for
performance: Evidence from the food processing industry. International managing fresh food quality throughout the supply chain. International Journal
Journal of Production Economics, 124, 75–86. of Production Economics, 131(1), 421–429.
Hughes, D., & Merton, I. (1996). ‘‘Partnership in produce’’: The J. Sainsbury approach Ross, A., & Droge, C. (2002). An Integrated benchmarking approach to distribution
to managing the fresh produce supply chain. Supply Chain Management: An center performance using DEA modeling. Journal of Operations Management, 20,
International Journal, 1(2), 4–6. 19–32.
Iijima, M., Komatsu, S., & Katoh, S. (1996). Hybrid just-in-time logistics systems and Shin, K. S., & Han, I. (2001). A case-based approach using inductive indexing for
information networks for effective management in perishable food industries. corporate bond rating. Decision Support System, 32, 41–52.
International Journal of Production Economics, 44, 97–103. Standard E2590. (2009). Standard guide for conducting hazard analysis-critical control
ISO/TS 22004. (2005). Food safety management systems – Guidance on the application point (HACCP) evaluations. United States: ASTM International.
of ISO 22000:2005. International Organization for Standardization. Suh, M. S., Jhee, W. C., Ko, Y. O., & Lee, A. (1998). A case-based expert system
Jedermann, R., Behrens, C., Westphal, D., & Lang, W. (2006). Applying autonomous approach for quality design. Expert Systems with Applications, 15, 181–190.
sensor systems in logistics – Combining sensor networks, RFIDs and software Tan, T. H., & Chang, C. S. (2010). Development and evaluation of an RFID-based e-
agents. Sensors and Actuators A – Physical, 132, 370–375. restaurant system for customer-centric service. Expert Systems with Applications,
Johnston, D. A., Taylor, G. D., & Visweswaramurthy, G. (1999). Highly constrained 37(9), 6482–6492.
multi-facilities warehouse management system using GIS platform. Integrated Trienekens, J., & Zuurbier, P. (2007). Quality and safety standards in the food
Manufacturing Systems, 10(4), 221–232. industry, developments and challenges. International Journal of Production
Kuo, J. C., & Chen, M. C. (2010). Developing an advanced multi-temperature joint Economics, 113, 107–122.
distribution system for food cold chain. Food Control, 21, 559–566. Unnevehr, L. J., & Jensen, H. H. (1999). The economic implications of using HACCP as
Kwok, S. K., & Wu, K. K. W. (2009). RFID-based intra-supply chain in textile industry. a food safety regulation standard. Food Policy, 24, 625–635.
Industrial Management & Data System, 109(9), 1166–1178. van Donk, D. P. (2001). Make to stock or make to order: The decoupling point in the
Lababidi, H. M. S., & Baker, C. G. J. (2003). Web-based expert system for food dryer food processing industries. International Journal of Production Economics, 69,
selection. Computers and Chemical Engineering, 27, 997–1009. 297–306.
Liao, H. T., Enke, D., & Wiebe, H. (2004). An expert advisory system for the ISO 9001 Wang, C., Luxhoj, F. T., & Fohansen, F. (2004). Applying a knowledge management
quality system. Expert Systems with Applications, 27, 313–322. tool for manufacturing vision (MV) development. Industrial Management & Data
Lyu, J., Jr., Chang, S., & Chen, T. (2009). Integrating RFID with quality assurance Systems, 104(9), 735–743.
system – Framework and applications. Expert Systems with Applications, 36, Wastson, I. (1997). Applying case-based reasoning: Techniques for enterprise systems.
10877–10882. San Francisco, CA: Morgan Kaufmann Publishers, Inc..
McPhee, W., & Wheeler, D. (2006). Making the case for the added-value chain. Westenbrink, S., Oseredczuk, M., Castanheira, I., & Roe, M. (2009). Food composition
Strategy & Leadership, 34(4), 39–46. databases: EuroFIR approach to develop tools to assure the quality of the data
Morris, C., & Young, C. (2000). ‘Seed to shelf’, ‘teat to table’, ‘barley to beer’ and compilation process. Food Chemistry, 113, 759–767.
‘womb to tomb’: Discourses of food quality and quality assurance schemes in Wruck, K. H., & Jensen, M. C. (1998). The two key principles behind effective TQM
the UK. Journal of Rural Studies, 16, 103–115. programs. European Financial Management, 4(3), 401–423.
Ngai, E. W. T., Suk, F. F. C., & Lo, S. Y. Y. (2008). Development of an RFID-based sushi Yang, J., Wong, C. W. Y., Lai, K. H., & Ntoko, A. N. (2009). The antecedents of dyadic
management system: The case of a conveyor-belt sushi restaurant. International quality performance and its effect on buyer-supplier relationship improvement.
Journal of Production Economics, 112, 630–645. International Journal of Production Economics, 120, 243–251.
Orriss, G. D., & Whitehead, A. J. (2000). Hazard analysis and critical control point Ziggers, G. W., & Trienekens, J. (1999). Quality assurance in food and agribusiness
(HACCP) as a part of an overall quality assurance system in international food supply chains: Developing successful partnership. International Journal of
trade. Food Control, 11, 345–351. Production Economics, 60–61, 271–279.
Peris, M. (2002). Present and future of expert systems in food analysis. Analytica
Chemica Acta, 454, 1–11.