1.0 PROBLEM DESCRIPTION 1.

1 Introduction The efficiency of current situation of cargo screening processes whether at sea and airports are largely unfamiliar as few benchmarks exists against which they could be measured. Some manufacturers provide benchmarks for individual sensors but we found no benchmarks that take a holistic view of the overall screening procedures and no benchmarks that take operator variability into account. Our aim is to develop a decision support tool (cargo-screening system simulator) that will view the right technology and manpower to the right commoditythreat combination in order to maximise detection rates.

The primary goal of cargo screening at seaports and airports is to detect human stowaways, conventional, nuclear, chemical and radiological weapons and other potential threats. This is an extremely difficult task due to the volume of cargo being moved continually through ports between countries. For example in sea freight, 200 million containers are moved through 220 ports around the globe every year.

1.2 Process Flow Diagram

Indonesia Border Control Office and Detention Facilities

Indonesia Pasport Check

Indonesia Screening Facilities

Indonesia Deep Search Facilities

Tickets

Depart at Klang Port

1.3 Belawan Port We have selected the port in Belawan (Indonesia) that links Belawan with Klang (Malaysia). This site is ideal for beginning as the security measures in place focus on detecting only one threat or illegal immigrants as they are called by the Indonesia Border Force.

1.4 The Real System In the real system the detection rates also depend on factors like the time of day (at busy times the operators have less time to apply the sensors and wait for the readings and therefore readings are more likely to produce more errors, operators skills misleading of interpreting the outputs from the sensors and operators fatigue. In Belawan, the cargo screening process is separated into two major zones, the first under the control of the Belawan Chamber of Commerce (BCI), the second under the control of the Malaysia Border Agency. Different types of sensors are used at the various screening facilities and some of them are also in use as mobile devices. The technology and operations used for screening includes Passive MilliMetre Wave scanners (PMMW), Heart Beat Detectors (HBD), , canine sniffers and visual inspection. The process on the Indonesia site starts with a passport check by the Indonesia authorities. Then all lorries are screened for illegal immigrants and suspicious lorries are routed to deep search facilities where they are further inspected by using an alternative method and if suspicion is confirmed then lorries are opened for visual inspection. In some cases (e.g. if it does not interrupt the process flow much, e.g. at non-busy times) lorries are opened directly for a quick visual check after or instead of being screened. If illegal immigrants are found on board a lorry they are removed by the Indonesia police, registered, and released into freedom. The process on the Indonesia site is very similar; the major

difference is that lorries are searched rather then screened and that only a fraction of the lorries going through the system is actually searched. Once the lorries have passed all check points they park at the Berth where mobile squads are operating to check the lorries a last time before they get on their way to Klang.

2.0 PRELIMINARY ANALYSIS OF THE SYSTEM 2.1 What are the overall purpose and characteristics of the system? The overall purpose of the development of a cargo screening process simulator is we attempt to identify and test innovative methods in order to advance the use of simulation for supporting decision making at the strategic and the operational level of the cargo screening process. Wilson(2005) confirms the usefulness of simulation for the analysis and prediction of operational effectiveness, efficiency, and detection rates of existing or proposed security systems. The characteristic of the system is this system will take place at the port which is involving two ports that is port in Klang, Malaysia and port in Belawan, Indonesia. The systems focus more on detecting illegal immigrants.

2.2 What are the physical features of the system? The physical features of the system are it involves two ports that are Klang and Belawan. Port Klang features two port terminals which are privately owned and managed and well-served by excellent infrastructure. In this system, every lorry cargo needs to pass screening process before they proceed to the agent boarder. At that port in the system, there are crew that will do the examination at each screening process in order to detect any illegal immigrants. The screening machine is needed in this system in order to make this procedure went well.

2.3 What specific performance features of this system are you interested in knowing more about? The specific performance features of this system that we interested in knowing more about are the screening procedure at the port. By implementing this system, the screening machine will detect any suspicious lorry that involved with the illegal immigrants conspiracy.

2.4 What elements of the system do you think are important to include in your model in order to address the issues you has raised? The elements of the system that are important to include in the model has three. First is a Detection Rate Matrix (DRM), a simulation model and a resource optimiser. The DRM will provide sensor detection rates as an input for the sensors represented in the simulation model, based on sensor types, commodities, threats, and other indicators. The simulation model will allow carrying out what if analyses for the system under examination. The results of the simulation will be fed into the resource optimiser to create a new set of input parameter values for the simulation.

2.5 What elements of the system performance do you need to consider that illuminate the issues you has raised? The elements of the system performance that we need to consider that illuminate the issues that we raised are screening the cargo will take some time and sometimes it can delay the process if there is some problem at the certain part. Thus, the process will be delay and can cause some loss to certain parties.

2.6 What are the roles of a schematic and process flow diagram (should be built yourself) in your model? The roles of a schematic and process flow diagram in our model act as a guideline to the management side at the port to carry out the process. The output of the simulator will consist of required technology and manpower and an estimation of the system detection rate that can be achieved by implementing the proposed system set-up. Other input data required for the cargo screening process simulator include an annual job list, guideline on how to carry out

jobs, and observations if jobs are carried out in accordance with these guidelines, and a list of existing sensors and staff.

2.7 How accurate would a schematic have to be sufficient for answering the questions that management is asking? In this project, we present a description of the real system and its operations, a conceptual model of it, an implementation of the conceptual model in form of a discrete event simulation model, and finally we show the results of an initial test run with our simulation model. So, the accuracy that would a schematic have to be sufficient for answering the questions that management is asking is totally high. This is because this model will represent the actual process.

2.8 What would you need to include in a process flow diagrams to completely characterize all possible entity pathways through the system? All the resources, entity and process that take part in the system need to be included in process flow diagrams to completely characterize all possible entity pathways through the system.

2.9 What are the primary resources of the system and the tasks or operation they perform? The primary resources of the system are the Boarder Agency Force at Belawan Port. The tasks and operation that this resources need to perform are to make sure all the entity which is the lorry are undergoing the screening process in order to eliminate any illegal immigrants during the cargo process at the port.

2.10 What kind of data do you think you will need to gather to model the system? The data that we need to collect is how many entities that will be involved and the time taken for the process to be complete. All this data can be obtaining from the port and at the same time can be obtaining throughout the secondary resources which are from the internet. On the other hand, we are dealing with the events that can be happen which is smuggling the illegal immigrants. The data can be obtained from the internet.

2.11 What are the assumptions have you made at each stage of modeling? There are a few assumptions that can we made at each of the stage of the modeling. First is during the screening process, there will be an inspection towards the entity by the resources. If the lorry is in the suspicious condition, it will go to the deep screening process. In this stage, the entire suspicious thing will be identified and will be removed. After everything is fine, the lorry can be proceeding to the shipping process. Thus, this model will be eliminating all the illegal immigrants that try to go to the Malaysia. At the end of this process, there is no more illegal immigrants.

3.0 MODEL CONSTRUCTION ISSUE

3.1 The most efficient and reliable way to determine the number of entities that come to the system The way to determine the number of entities that come to the system is cargo screening process. Which is, divided by two prior zones, one is under the control of the Malaysia Government and another one is under Indonesia agency. There are different types of sensor are used and some of them are also in use as mobile equipments, which can be use as a detector to the number of entity that come to the system. There have several operations used for screening as we mentioned earlier.

3.2 A strategy for getting service times that would be accurate and thorough and data-takers that is necessary to take data at each station in the system

Routing elements use a custom made function which routes the entity to the next level upstream element with the shortest queue. This represents the routing activities normally conducted by an officer (resource). On the one hand we are dealing with rare events which impacts on the way we have to do our input sampling and output analysis, and usually some data cannot be obtained from the real system. So, we have to make a lot of guesses. Even if we had the resources to collect the data there are some legal issues regarding the sampling frame which prohibits us to collect some of the required data as we are not allowed to sample an entire population.

3.3 The entities, paths to follow, resources, arrivals, distributions and decision points relevant to this system that would have to be considered in the model

In our study, entities of type lorry, which is in form of soft and hard sided are grab into the simulation at a certain rate by the source items (arrival). A particular lorry will arrive with an additional load (illegal immigrant) on board. We currently modelled a illegal immigrant as a resources. The prior elements in the simulation model are the service run down, which represent the reusable component as passport and ticket booth on the Indonesia compartment. These are used for modelling the situations when we have fixed sensor and fixed targets. The service run down elements are linked via some routing elements. The routing elements use a custom made function which routes the entity to the next level upstream element with the shortest queue. This represents the routing activities normally conducted by an officer. The moving squads are modelled by events that pick one lorry at random and check it. The time it takes to check a lorry is represented by the inter arrival time between two events. This means that the squads are currently modelled as being 100% utilised as long as there are lorry to check. There are two situations, either lorry can be checked only ones, which mean lorries that have been checked already are registered on an ignore list, or lorries can be checked multiple times, which represents the situation where illegal immigrant enter the lorry while these are parking at the berth and therefore the squad would check suspicious lorry again.

3.4 How many variables, random variables and attributes should be in the model? The variables for our study are illegal immigrants, screening sensor, lorry, arrival rate and berth. The attributes for our study are passport, ticket, service time and soon. In the end the big question is if modelling all these details is not necessary for getting useful results.

3.5 Do you need to perform sensitivity analysis to assess the impact of changes in data on the model results?

In order to answer this question, in case if we choose to do so, we can conduct a sensitivity analysis and implement them at least partially and build some smaller simulation models that only represent a small section of the overall real system. So far we have conducted some very basic preliminary tests with our simulation model.

3.6 How to verify that your computer representation represents the model that has developed?

A verification and validation exercise is still to be carried out. Thus, here we briefly report on one of the tests we have conducted. We have set up simulation model using our standard set of parameter values, except for the sensor detection rates, which we have set to the same value for all sensors. As for the results we expect to see a non-linear relationship between sensor detection rates and the average proportions of illegal immigrants detected. This is due to the fact that many lorries will go through several screening procedures and therefore combinatorial effects appears for this relationship, where higher individual sensor detection rates will have a proportionally lower benefit regarding the system detection rate. This first test has already shown the impact of modelling rare events. We observed that the illegal immigrant detection rates vary significantly throughout most of the simulation runtime and seems only to stabilise towards the end. In addition, we noticed some significant differences between runs. Hence, in future we have to assess very carefully the required warm-up period, run length and number of replications.