SAFETY FORESIGHT

1. INTRODUCTION

Preface

Jakarta Port is a key hub for maritime activities, with a significant increase in shipping volume, cargo handling, and other related operations. The reported figure for the movement of international containers in 2022 reached 3.48 million TEUs, constituting 94.28 percent of the set target of 3.66 million TEUs. This represented a 2.04% increase compared to the 3.41 million TEUs recorded in 2021.

Simultaneously, domestic container activity throughout 2022 amounted to 7.67 million TEUs, attaining 96.1% of the targeted 7.98 million TEUs. The flow of domestic containers experienced a modest growth of 0.65 percent compared to the 7.62 million TEUs recorded in 2021.

The activity surge raises concerns about the potential rise in incidents, such as accidents, security breaches, and environmental hazards. Besides that, every port facility must comply with international and national maritime regulations in expediting the operations of the Port. To overcome these two challenges, we can use historical data to identify trends and analyze patterns that can inform forecasting models, providing valuable insights into the factors influencing incident occurrence and recurrence.

These insights will serve as a strategic foundation, allowing us to systematically minimize the occurrence of incidents by allocate resources for the emergency response team efficiently, reduce response time, and enhance the overall effectiveness of emergency measures.

Goals

Indeed, proactive forecasting of incident occurrences plays a crucial role in empowering port facilities to prepare for and mitigate risks effectively. The Health Safety Environment (HSE) Department, positioned at the forefront of risk management, stands to benefit significantly from this practice. By leveraging forecasting models and insights derived from historical data, the HSS Department can anticipate potential incidents, strategically enhancing safety protocols. This involves fine-tuning and reinforcing existing safety measures to address specific high-risk areas identified through historical patterns.

In summary, prioritizing the minimization of incidents not only prevents immediate disruptions but also safeguards the long-term sustainability and success of port facilities. Through the use of forecasting models, port facilities can proactively prevent unforeseen events that might otherwise result in substantial costs.

2. DATA PREPARATION

We seek to leverage a dataset covering incidents at one of ports at North Jakarta from 2019 to 2023, justifying the collected data’s credibility through its source validity and collection methodology. The dataset itself is the digitalization of manual report from The Health Safety Environment (HSE) Department. The reliability of this data is pivotal to ensuring the accuracy and meaningfulness of the upcoming time series model.

This dataset aligns with business needs, offering an overview of accidents at one of North Jakarta’s port within a specific timeframe. It is a valuable resource for analyzing accident trends, identifying causative factors, and crafting effective prevention strategies. These insights can significantly enhance the port’s safety system and optimize risk management. Preserving data privacy through the masking process allows us to create a time series model that delivers precise results without compromising the security of personal information.

Importing Library

# use for data preparation
library(readxl)                     
library(dplyr)

# use for graphic chart
library(ggplot2)
library(plotly)
library(glue)
library(zoo)

# use for wordcloud
library(stringr)
library(tm)
library(SnowballC)
library(wordcloud)

Importing Dataset

This dataset is the Summary of Incident happened in one of North Jakarta’s port. the Dataset row explaining report of each incident, for example at when is the incident occured, this incident can be classify into what type, who are involved in this incident and short description about the incident chronology.

Our focus is to utilize the incident number from 2019 to 2023 as our target variable. Upon thorough examination of the dataset, we have assumed a correlation between the number of incidents and prevailing weather conditions. Factors such as rainy weather, strong winds, and high temperatures may significantly influence the likelihood of incidents occurring. To bolster our analysis, we have incorporated additional datasets detailing the average temperature, average wind speed, and average rainfall specifically in the North Jakarta Area for the aforementioned period. These supplementary datasets have been sourced from the BMKG Database, accessible at https://dataonline.bmkg.go.id/.

rawdata <- read_excel("data_input/RAW_DATA.xlsx")
bmkg <- read_excel("data_input/BMKG.xlsx")

head(rawdata)

#> # A tibble: 6 × 21
#>   Datetime            Tanggal Bulan   Tahun   Jam Hari   Department Tipe_Insiden
#>   <dttm>                <dbl> <chr>   <dbl> <dbl> <chr>  <chr>      <chr>       
#> 1 2019-01-05 00:00:00       5 January  2019    14 Satur… Safety     Near Miss   
#> 2 2019-01-09 00:00:00       9 January  2019     8 Wedne… Safety     Near Miss   
#> 3 2019-01-11 00:00:00      11 January  2019    20 Friday Safety     Near Miss   
#> 4 2019-01-14 00:00:00      14 January  2019    19 Monday Safety     Near Miss   
#> 5 2019-01-14 00:00:00      14 January  2019    23 Monday Safety     Near Miss   
#> 6 2019-01-16 00:00:00      16 January  2019     1 Wedne… Safety     Near Miss   
#> # ℹ 13 more variables: Jenis_Alat <chr>, Nomor_Alat <chr>, Lokasi <chr>,
#> #   Penyebab <chr>, Akibat <chr>, Nama_Terlibat <chr>, Nama_Perusahaan <chr>,
#> #   Status <chr>, Group <chr>, Shift <chr>, Nama_Korban <chr>,
#> #   Nomor_Container <chr>, Deskripsi_Kejadian <chr>

head(bmkg)

#> # A tibble: 6 × 7
#>   Tanggal     Date Month    Year Temp_avg Curah_Hujan Wind_avg
#>   <chr>      <dbl> <chr>   <dbl>    <dbl>       <dbl>    <dbl>
#> 1 01-01-2019     1 January  2019     25.9         9.8        4
#> 2 02-01-2019     2 January  2019     28.2         2          5
#> 3 03-01-2019     3 January  2019     28.7         0.7        5
#> 4 04-01-2019     4 January  2019     29.9        NA          3
#> 5 05-01-2019     5 January  2019     29.2        NA          2
#> 6 06-01-2019     6 January  2019     29.6        NA          1

3. EXPLORATORY DATA ANALYSIS

As we delve deeper, we encounter insights—nuggets of wisdom embedded in the data’s fabric. The exploratory phase allows us to pose questions and receive answers, fostering a dynamic dialogue with the information. Here below is the map of the question we already enlisted, which we think is crucial to be answered. We would like to see which word is frequently used in our manual report using NLP.

Graphic Chart

Time-Related

Certainly, starting with a detailed analysis of the monthly incident data over the past 5 years is a sound approach to gaining insights. By breaking down the number of incidents for each month annually, we can identify patterns, trends, and potential factors influencing variations in incident rates over time. This analysis will serve as a foundational step in understanding the dynamics of incidents throughout the years and can guide further exploration into the underlying causes and correlations.

📈 Insight :
1. From the Incident Trend Analysis plot we conclude that the trend has no pattern.
2. The number of incident is range from 2 to 16.

# calculating correlation coefficient between Total_Incident and Avg_Temp
cor(eda_monthly$Total_Incident,eda_monthly$Avg_Temp)

#> [1] 0.1086488

# calculating correlation coefficient between Total_Incident and Avg_Rain
cor(eda_monthly$Total_Incident,eda_monthly$Avg_Rain)

#> [1] -0.02655847

# calculating correlation coefficient between Total_Incident and Avg_Wind
cor(eda_monthly$Total_Incident,eda_monthly$Avg_Wind)

#> [1] -0.01183738

📈 Insight :
1. From calculating the correlation coefficient between Total_Incident and weather condition columns (Average Temperature, Average Rainfall, and Average Wind Speed), we can see that the correlation coefficient is very low. This means that the number of Total_Incident is not much affected by the weather condition (no strong correlation between the target column and predictors column).
2. Due to this information, we are confident in using the Time Series Model to forecast the total number of incidents.

Next, we delve into the specific insight on the number of occurrences each year.

📈 Insight :
1. There was no significant change in the total number of incidents over the 5 year period. The total number of incidents remains above 100.
2. From 2019 to 2021 the total number of incidents are increasing but then after 2021 the number slowly began to decrease until 2023.
3. These number of incidents may result from a misalignment between safety strategies and their intended objectives.
4. We recommend to do re-assessment and re-evaluation of safety strategies, for leading to targeted improvements and sustained incident reduction.

Next, we examine monthly incident patterns for a more granular understanding of whether specific months consistently exhibit higher or lower incident rates.

📈 Insight :
1. The highest total number of incidents occurred in May, followed by August and April.
2. In contrast, October recorded the lowest total number of incidents.

Next, we analyze incident patterns at the hourly level to reach a detailed examination of when these incidents are most likely to occur throughout the day.

📈 Insight :
1. The peak occurrence of incidents was recorded during the 22:00 - 22:59 time frame.
2. The peak occurrence happens near shift changes hour (23.00), this may indicate that during the handover phase, there is potential for miscommunication, gaps in awareness, and temporary lapses in vigilance.
3. The predominance of incidents during nighttime is evident, as several time frames within that period have recorded numbers surpassing the average.
4. The higher incidents occurring during nighttime can be influenced by several factors such as Reduced Visibility, Fatigue, Reduction in Supervision and Monitoring, and Reduced Staffing.

Incident Type Related

Exploring incidents based on type is a crucial step to gaining a comprehensive understanding of the nature and characteristics of the incidents. Here, we would like to compare the number of incidents by type.

📈 Insight :
1. It appears that, in each year, the predominant incident type is consistently identified as “Near Miss.” This pattern suggests that incidents characterized as near misses, where an accident nearly occurred but did not result in injury, are the most frequently recorded incidents across the years.
2. Based on Heinrich’s Safety Pyramid theory which suggests that for every major accident (fatality), there are a larger number of minor incidents and near misses. The ratio number may vary or differ in every industry or situation, but the concept is to prevent and address the lower level of incidents that could reduce the likelihood of more severe accidents.

Equipment Related

Analyzing incidents by the type of equipment involved is a crucial step in identifying potential safety vulnerabilities and implementing targeted preventive measures.

📈 Insight :
1. The observation that most incidents are related to RTGC (Rubber-Tired Gantry Crane) equipment is significant information. Addressing safety concerns associated with RTGC equipment becomes a priority, and specific actions can be taken to mitigate risks.
2. The prevalence of incidents involving RTGC (Rubber-Tired Gantry Crane) and QCC (Quay Crane Container) is striking, with these two equipment types surpassing others in frequency. Given their considerable size and elevated inherent risks, we strongly emphasize the need for implementing rigorous and all-encompassing risk mitigation measures precisely designed for these equipment types. This proactive approach is crucial to minimizing the likelihood of incident occurrences and ensuring a safer operational environment.

Wordcloud

We would like to analyze our reported incidents categorized by type comprehensively. We aim to discern the frequently used words within the manual reports associated with each incident type. This exploration will provide valuable insights into the specific language and key terms prevalent in our incident documentation, helping us enhance our understanding and response strategies.

Fatality Report

📈 Insight :
1. It’s crucial to note that if the most frequently used word in reporting a fatality incident is “TRUK,” this information offers valuable insights into the context and circumstances surrounding such incidents. This finding could potentially guide safety measures, training programs, or operational adjustments to address the factors associated with incidents where the term “TRUK” is prevalent.

First Aid/Injury Report

📈 Insight :
1. The predominant terms in reporting first aid or injury incidents are “tkbm” (stevedore), “kontainer” (container), and “menyebabkan” (caused). This discovery leads us to the conclusion that a significant number of individuals receiving first aid or experiencing injuries are related to stevedore activities, particularly in connection with containers. Understanding these key terms helps pinpoint areas for targeted safety measures and underscores the importance of addressing risks associated with stevedore operations.

Dangerous Occurrence Report

📈 Insight :
1. The most frequently mentioned term in dangerous occurrence reports is “API” (fire). To enhance countermeasures, we propose heightened awareness among individuals working in proximity to high-risk areas for fire or explosion. They must comprehend the potential hazards and risks associated with their tasks. Additionally, considering their proficiency in handling fires is vital, as time plays a critical role in effective fire management.

Nearmiss Report

📈 Insight :
1. The prevalent terms in reporting Near Miss Incidents are “kontainer” (container) and “RTGC.” These findings raise awareness about potential risks and hazards associated with RTG or RTGC operations. The connection to the term “kontainer” suggests that near-miss incidents may be most likely to occur during the operation of RTGC, particularly during the delivery of containers. In light of this information, we recommend a thorough assessment of RTGC operation procedures or an evaluation of operators to ensure safety protocols are robust and effective.

4. LAYOUT DASHBOARD