Project Proposal

Prime Mover (PM) Waiting Time in Yard

Li Zheng Long, Lim Kai Chin https://kaichinlim.netlify.app/
02-25-2021

Introduction

Maritime trade has been the backbone of international trade as they account for approximately 92% of world trade.1 PSA Singapore handles about a fifth of the world’s transhipped containers as the world’s busiest transhipment port.2 For more than four decades, PSA continuously developed and upgraded its container handling infrastructure, pioneered new systems and processes, and streamlined operations to meet the rapid growth in its container terminal business as part of the strive for operational excellence.3

Upon a vessel’s arrival at a berth in the container terminal, containers are discharged from and loaded onto it. A typical discharging operation starts with a quay crane picking up a container from the vessel and placing it onto a PM, which will then transport to a storage yard. At the yard, a yard crane picks up the container from the PM and shifts it to a designated spot. Loading operations involve the transporting of containers in the opposite direction, from the yard to the vessel.4

Therefore, PM productivity is of key interest to PSA as it is the main driver for the time taken to load and unload vessels. PM Productivity is defined as the sum of the total number of containers handled divided by the total hours. The following are the key terms which PSA uses to define PM productivity.

\[ Productivity = Total Containers Handled / Total Time \] \[Total Time = Sum (Est. Travel Time + Est. Wait Time + Non Work Time)\]

Total Time is defined as the time difference between the two operation activities.

Estimated Travel Time is the duration between two locations based on distance matrix with fixed speed limit/hr.

Non-Work Time is the time taken for a change of driver, meal break, and PM breakdown (if any).

Estimated Wait Time = Total Time – Non-Work Time – Est. Travel Time

Project Motivation

The study objective is to seek insight from Prime Mover (PM) Operations records to identify common characteristics exhibited by PM with high and low waiting times, through understanding of PM events and operational data. This, in turn, enables us to pinpoint and identify correlated attributes and embark on further study to improve the overall productivity of PM operations and resource utilisation through active targeting of activities contributing to the PM waiting time.

Review of Existing Work

Current operation performances are tracked using Operation Indicators such as PM productivity and PM waiting time aggregated by duration, from shifts to monthly reports and breakdown by individual terminal and PM.

Previous studies on PM operation efficiency typically focus on crane productivity by work schedule5 , and resource planning & deployment to find the optimal number of PMs and trucks (haulier) to reduce average PM waiting time.6

Proposed Scope and Methodology

This study seeks to create an R Shiny app that allows for users to input parameters to suit their needs.

  1. Interactive dashboard that allows for Operations Managers to understand

    1. Summary statistics for PM events

      1. Histograms for wait time with parameters for each type of container
      2. Time series (across hours) for wait times with variable parameters

      1. Average PM wait time by cntr length by time of day
      2. Average PM wait time by equipment type by time of day
      3. Tree maps for container types and movements
      4. Bar charts of container types
      5. Scatter Plots (Total duration of entry vs waiting time)

    2. Performance of Terminals

      1. Histograms for wait time with parameters for each type of container

    3. Performance of type of containers (Dangerous Goods, Reefer, General Purpose) & size of containers (20-footer, 40-footer, Oversized)

    4. Performance of type of equipment (Quay crane, various types of Yard Cranes)

  2. Confirmatory data analysis

    1. Uncertainty analysis through calculation of median and confidence intervals of time spent based on types and sizes of containers, different terminals, Equipment type, Day/night shifts, different yard activities.

    1. Distribution comparisons between categories of containers(e.g. using box plots, violin plots)
    2. Multivariate analysis using cross tabs
    3. Parallel coordinates analysis

Storyboard and Visualisation Features

The Shiny app will be designed in this manner:

Introduction: To give context and background for the project

EDA - Time Series: To visualise the activities that are carried out across time periods

EDA - Treemaps: To visualise the proportion of different types of parameters (container type, equipment type)

EDA - Bar Charts: To visualise the counts and activities happening in a day

EDA - Histogram: To visualise the distribution of waiting times and driving times of the activities

CDA - ANOVA: To test for different variances and distributions

CDA - Parallel Coordinates: To visualise different patterns in the data

CDA - Violin / Box Plots: To visualise the confidence intervals

About: To provide definitions and FAQs

Data Source

The data source used is provided by PSA Singapore’s PM OPS anonymised dataset that contains PM operation event records in a generic 12 hr shift. There are approximately 65,000 records with 69 variables. The dataset is reduced to the following 17 variables for the analysis:

S/N Variable Name Description
1 [SHIFT_D] Date of shift
2 [TERMINAL_ID] Container Terminal ID
3 [EVENT_C] EQOF = Equipment Offload from PM to Yard Crane EQMT = Equipment Mount to PM from Yard Crane
4 [EVENT_DT] Event Datetime
5 [EVENT_SHIFT_I] Shift indicator D =Day N = Night
6 [MOVE_OP_C] Move Operation Code
7 [LENGTH_Q] Container Length e.g 20 40 45 ft
8 [CNTR_TYPE_C] Container Type RF = Reefer GP = General Purpose DG = Dangerous Goods OH = Over height Container UC = UnContainerized
9 [CNTR_ST_C] Container Status E = Empty , F = Full
10 [DG_I] Dangerous Good Indicator
11 [REEFER_I] Reefer Indicator
12 [UC_I] Uncontainerized Indicator
13 [OVER_SIZE_I] Over Size Container Indicator
14 [EQUIPMENT_TYPE_C] Equipment Type Code (Quay Crane and Type of Yard Cranes)
15 [PM_DISTANCE_Q] Distance travelled from previous location
16 [PM_TRAVEL_TIME_Q] Travel Time
17 [PM_WAIT_TIME_Q] Wait Time

Application Libraries & Packages

Task Breakdown

The areas of responsibility are foreseen to be as follows.

Tasks Team Members Responsible
Introduction and business problem definition KC
Data acquisition and preparation - cleaning & anonymisation ZL
Literature Review KC
Exploratory Data Analysis -Time Series -Scatter Plots (total duration of entry vs waiting time) ZL
Exploratory Data Analysis - Treemaps (container types & terminals) - Pie / Bar charts (activity counts) - Histograms (waiting time, driving time) KC
Confirmatory Data Analysis - Analysis of Variance (ANOVA) - Equal / Unequal variance - Normal / Not normal distribution ZL
Confirmatory Data Analysis - Parallel coordinates analysis to visualise key relationships KC / ZL (Tbc)
Uncertainty - Median, confidence intervals via box plots / violin plots for waiting times - Error bars KC
Insights and recommendations for actions KC
Limitations and conclusion ZL

References

1 https://royalsocietypublishing.org/doi/10.1098/rsos.200386

2 https://seanews.co.uk/features/the-busiest-ports-in-the-world-and-how-they-handle-operations-part-ii-singapore/

3 https://www.globalpsa.com/psa-international/

4 https://www.scs.org.sg/articles/how-arti-cial-intelligence-can-make-our-port-smarter

5 https://www.pomsmeetings.org/confpapers/005/005-0094.doc

6 https://www.win.tue.nl/oowi/final%20project/archive/KoenStaats.pdf