CTO of Softengi with 30 years of experience in software development, business applications implementation and digital strategy creation.
As cutting-edge technologies continue to advance and the business world becomes highly digitized, asset-intensive industries — such as construction, oil and gas, manufacturing, agriculture and mining — are accelerating safety innovations as never before, enhancing their Environment, Health and Safety (EHS) processes and performances. A new world of safety, powered by technologies such as artificial intelligence (AI), opens up a great number of opportunities for all asset-intensive companies, among which are improved asset performances, quality management and environmental and employee safety.
According to Gartner, EHS applications are developed to assist companies in managing employee health, decreasing safety risks and managing their environmental footprint. The EHS applications are diverse, covering various areas, such as occupational health/industrial hygiene, incident management, product safety and compliance, environmental monitoring and operational risk management.
Of all the above areas, incident management applications are the most complex in development. As incidents and field accidents highly depend on the asset location and environment, incident prediction applications have to operate at the place of risk occurrence. However, designing and architecting open-air solutions and software applications that operate on open ground is very challenging. These have to be adaptable to various environmental conditions and comply with a number of technological limitations, managing both internal and external risks.
The Stages Of Development Of An AI-Powered Open-Air Applications
To create an EHS open-air solution, various technologies need to be applied, such as AI, IoT, computer vision and cloud and edge computing. These ensure digitized safety processes on-site, managing different kinds of health and well-being risks, predicting high potential (HIPo) incidents and forecasting fatigue risk levels.
1. Data Collection
From a technical standpoint, AI-based software runs through several development stages. First, the relevant data has to be collected beginning with the identification of the diverse data and its collection from both internal and external sources — for example, data gained from IoT sensors, video content, mobile devices and more. In the case of AI-powered EHS applications, it is necessary to build a database, feeding it with the company’s internal data and then enriching it with external information about weather, geolocation and air pressure or information from social media. For the EHS segment, it is highly important that the collected data is diverse enough to enable the prediction capabilities of AI applications to adapt to a variety of possible scenarios.
2. Model Training
After the database is built, a training model needs to be defined. At this stage, we speak about defining neural networks, selecting an appropriate machine learning algorithm and “feeding” it with huge amounts of diverse data. In essence, the appropriate identification algorithms and methods are selected and then applied in the AI system, depending on the predefined risk parameters.
Training large neural networks requires massive amounts of labeled training data. To maximize the application’s ability to make accurate predictions, it is recommended to use not only supervised learning methods but also semi-supervised and reinforcement learning approaches as well. For instance, a voice checklist can be applied to train the data on the field thanks to its practicality — employees’ hands are often busy and they can interact with the AI software using their voice. As a result, data can constantly be trained in the field without much effort.
3. Predictive Analysis
The last stage of AI application development is predictive analysis. To enable the rapid processing of real-time data, it is most effective to apply both cloud and edge computing. In the AI applications part, the received data is processed at the edge, bringing the calculations closer to the risk points while the rest of the data is sent to the cloud for further model training. In this case, edge computing is efficient if a user needs prompt instructions and guidance from supervisors or real-time insights into business performances. For instance, if a worker on a construction site violated the requirements and did not wear a helmet when being in a dangerous spot, the AI system would transmit the anomaly data to the edge nodes and the system would recognize the risk immediately, sending an instant notification to a supervisor.
Safe Working Environments With AI
In 2018, more than 5,000 workers in the U.S. died at their workplace. In the same year, the expenses on occupational injuries and accidental deaths in the U.S. amounted to almost $171 billion according to the National Safety Council. To protect employees’ health and well-being, companies have to transform workspaces and take care of their employees. When combined with other related technologies, AI is able to augment the human workforce, allowing EHS professionals to proactively manage various EHS risks and create a safe workplace environment for employees.