Rectifying Unsafe Acts and Ensuring Holistic Safety in Oil Refineries with AI

7 mins read

The Oil & Gas (“O&G”) industry has always focused on robust processes to ensure safe operating environments. That said, we continue to hear of incidents across the industry that have devastating consequences for people asset and environment. Consistent effort are therefore made to push the envelope on more stringent safety standards and simplify their implementation through adoption of modern-day solutions. This focus continues to be relevant, considering that Oil & Natural Gas industries are expected to remain dominant players in the energy mix for most countries and regions globally.

The topic of workplace safety in operating refineries is potentially important for you especially if you are operating within the upstream, mid-stream or downstream ecosystem. We have received numerous queries on our previous blog with a healthy exchange of ideas and keen interest in knowing more. We believe you have read the blog but just in case you have missed it you can read about the refinery safety hazards here.

In this blog we address how adoption of AI/ML tools engineered by Detect are propelling workplace practices to their next frontier.

Table of Contents

Reverse Engineering Safety Incidents

Usually, many aspects of near misses, LTI or observations are recorded daily. But for simplicity, we can broadly categorize them as unsafe acts, unsafe conditions, process gaps and external factors. An unsafe act is a violation of an accepted safety procedure that can cause an accident, while an unsafe condition is a hazardous physical condition or circumstance that could directly lead to an accident.

It is important to mention here that the contributing factors seldom exist in silos and often interact with each other often, to precipitate into a safety event. For example, an inspection or work carried out at height using an unsecured work platform is an “unsafe condition”, while not securing the harnesses while working at height is an “unsafe act”. As imaginable, a sudden shower of rain or an unexpected thunderstorm (external factor) can exacerbate hazards associated with these conditions when they exist in tandem.

Breakdown of workplace accident causes: 88% unsafe acts, 10% unsafe conditions, 2% external factors.
Understanding the root causes of workplace accidents.

But do you know what else exacerbates the danger even further?

The use of traditional and reactive methods for safety, that involve use of discrete information on a piece meal basis can further complicate the situation by not providing 360 degree information to the respondent. Worse still, they leave ample room for error in high-intensity workplaces with asymmetric information leading to lapses in the moment that matter.

If industries such as refineries are to be guarded against all possible threats and hazards, they need to adopt a more proactive approach. Information silos need to be pulled down to ensure, all stakeholders have access to information they need, enabling them to make a change on the go.

Rectifying Unsafe Acts with AI

Following the recent advancements in data analytics and AI technology, it is finally possible to store, analyze, and infer gigantic volumes of data to uncover critical gaps within an organization’s HSE programs. Once located, these gaps need to be filled to effectively check unsafe acts and eliminate unsafe conditions.

To achieve this it needs to be ensured that the AI tool of choice is fit for purpose. A tool that highlights events behind leading safety indicators without creating noise. A large number of duplicate alarms or false positives may dis-incentivize the users, leading to slow adoption or natural rejection of tools.

It is also important that the tool of choice discourages its use as a policing tool. It should rather be generating insights and breaking information silos to extend human capabilities. End of the Day, it’s the experts who make the change and as such the tool should enable them with coherent data points to make the change without subjecting them to information overload.

Any event is not the outcome of one lapse, it can be but most often it is not. Therefore, a truly intelligent tool not only provides relevance but also comprehensive information, which helps the user develop an understanding of the context and therefore the appropriate response.

By design, such a tool promotes a proactive culture, by helping users to move from a reactive to a responsive approach. To understand the point better, let’s look at some use-case based examples.

Responding Effectively: Unsafe Acts

Amputation Hazards

If workplace safety for a refinery were an alphabet set, PPE detection would spell ‘A’. It is well understood that lack of proper adherence to PPE regulations can lead to injuries. However, context matters.

Today there is no dearth of AI models that can continuously monitor and identify a lack of PPE safety compliance in real time. Any tool can quickly identify missing jackets, bands, gloves, or hard hats. However, if such tool generates thousands of notifications, most of which are about events with missing PPE in secure zone, users will get disengaged with the tool. As such it is highly likely that the user might miss notification of an event where a person is not wearing a glove while handling a load subjecting the person to an amputation hazard.

A truly intelligent tool would be able to distinguish between contextually relevant information and noise.

Line of Fire

Similarly, there are tools that claim to notify on line-of-fire hazards by detecting proximity. However, they often end up generating hundreds of notifications which are not relevant in the context of the situation. The definition of Line of Fire does not hold true in its entirety when the machine is under maintenance or is shut down. A smart tool would know how to distinguish between the two scenarios and would enable HSE Experts to understand human factors around genuine line-of-fire issues.

Unsafe Position

Even fatigue, officially defined as a reduced state of mental or physical alertness, causes an approximately 62% higher risk of accidents. A review of various industrial vehicle statistics shows that fatigue contributes to approximately 5-25% of accidents.

Fatigue can lead to serious accidents. Don't ignore the signs. Your safety depends on it

In operating assets, the lagging factors around fatigue are often different and so are the ways of identification. Fatigued work crews tend to exhibit poor posture or rely on methods that require less effort.

Given the context, an attempt to retrofit solutions trained on data sets not relevant to the refinery context, would provide highly erroneous alarms. It might alert a person bending on ground however might miss to detect situations where an individual selects an unsafe accessway due to his fatigued state.

Improper Asset Usage Detection

Improper Machinery/Equipment usage has consistently fared in OSHA’s Top 10 since the very dawn of mechanization. Asset-related accidents are so dangerous because they can occur without warning.

Equipment operators receive extensive training on safe procedures and troubleshooting. Even trained operators can make mistakes if their decision-making is impaired. A smart tool can alert on anomalous operating patterns, prompting fast and decisive remediations.

AI's Role in Shaping a Safer Future for Refineries

To sum up, Refineries are riddled with safety hazards attributable to bunch of conditions, that often interact with each other. AI empowers experts to analyze data, detect anomalies, and make informed decisions to preemptively address these elements of risk.

That brings us to a critical juncture in our journey to understanding the impact of AI in transforming industrial safety. A tech tool should enable human performance by extending and augmenting human cognitive capabilities. It is therefore absolutely essential to identify and adopt tools that fit the purpose. It is widely recognized by every industry that people are a part of the solution and not the problem. Any solution that promotes consequence management will end up fostering a reactive culture and will therefore fail to deliver a holistic cultural improvement.

Today we looked at how AI can help in preemptively identifying and rectifying hazards. In our next piece, we will explore how AI can enhance our processes by enabling a continuously learning document management system. Till then, feel free to check out our flagship AI product designed to impart safety and productivity to industrial workplaces.

Are you ready to experience T-Pulse?

Meet highest compliance, monitor all occupational risks, and get recommended actions to achieve global safety benchmark.

Share this blog post via