Revolutionizing IC Genset Operations with IIoT and AI: A Study on Fuel Savings and Predictive Maintenance

Abstract

In a world increasingly aware of its carbon footprint, the quest for sustainable energy production and consumption has never been more urgent. A key player in this monumental endeavor is fuel conservation, which helps curb greenhouse gas emissions and preserve our planet’s finite resources. In the realm of the Industrial Internet of Things (IIoT) and artificial intelligence (AI) technologies, Caterpillar (CAT) generator set (genset) operations have been revolutionized, unlocking unprecedented fuel savings and reducing environmental harm. Envision a system that not only enhances fuel efficiency but also anticipates maintenance needs with state-of-the-art technology. This standalone IIoT platform crafted with Visual Basic.Net (VB.Net) and the KEPware Object linking and embedding for Process Control (OPC) server gathers, stores, and analyzes data from CAT gensets, painting a comprehensive picture of their inner workings. By leveraging the Modbus Remote Terminal Unit (RTU) protocol, the platform acquires vital parameters such as engine load, temperature, pressure, revolutions per minute (RPM), and fuel consumption measurements, from a radar transmitter. However, the magic does not stop there. Machine Learning.Net (ML.Net) empowers the platform with machine learning capabilities, scrutinizing the generator’s performance over time, identifying patterns and forecasting future behavior. Equipped with these insights, the platform fine tunes its operations, elevates fuel efficiency, and conducts predictive maintenance, minimizing downtime and amplifying overall efficiency. The evidence is compelling: IIoT and AI technologies have the power to yield substantial fuel savings and enhance performance through predictive maintenance. This research offers a tangible solution for industries eager to optimize operations and elevate efficiency by embracing IIoT and AI technologies in CAT genset operations. The future is greener and smarter, and it starts now.

1. Introduction

The power generation industry has been facing significant challenges in recent years, ranging from increasing fuel costs to environmental concerns and the need for sustainable and efficient operations. As the demand for cleaner and more efficient energy sources continues to grow, the industry is under pressure to innovate and optimize its processes. One potential solution to these challenges lies in the integration of Industrial Internet of Things (IIoT) and artificial intelligence (AI) technologies. IIoT and AI have shown promise in various industries and have the potential to contribute significant improvements to the power generation industry, particularly in the operations of internal combustion (IC) generator sets (gensets). However, despite the promising potential, the current literature on the subject lacks comprehensive studies on the potential of IIoT and AI in revolutionizing IC genset operations and reducing fuel consumption. Therefore, this research paper aims to bridge this gap by providing valuable insights into the subject. The main problem addressed in this research is the need for improved fuel efficiency and predictive maintenance in IC genset operations. Integrating IIoT and AI technologies can potentially lead to better monitoring, control, and optimization of gensets, ultimately resulting in reduced fuel consumption and improved sustainability. To explore this potential, in this study, we focus on the development and implementation of a standalone IIoT platform for IC genset operations. We first provide a brief but comprehensive overview of the existing literature on the use of IIoT and AI in the power generation industry and conclusively that the integration of these technologies in IC genset operations has the potential to revolutionize the power generation industry by providing improved fuel efficiency and predictive maintenance. Furthermore, with this research paper, we aim to contribute to the existing literature by detailing the development, implementation, and evaluation of a standalone IIoT platform for IC genset operations. We also analyze the potential benefits and challenges associated with integrating these technologies in the industry. The IIoT platform developed for this study can enable real-time monitoring and analysis of IC genset operations, including fuel consumption, engine performance, and maintenance needs. The platform utilizes various sensors and data analytics tools to collect and analyze data. Additionally, the platform has predictive maintenance capabilities that help to identify potential issues before they become critical problems, significantly reducing downtime and maintenance costs. In this study, we evaluate the performance of the IIoT platform in the real-world setting a power generation facility. The evaluation is focused on the platform’s effectiveness in reducing fuel consumption and improving the overall efficiency of IC genset operations. The results of this study provide valuable insights into the enormous potential of IIoT and AI in revolutionizing IC genset operations and substantially reducing fuel consumption.

2. Literature Review

Sustainability encompasses environmental, social, and economic dimensions and aims to ensure that current generations can meet their needs without compromising the ability of future generations to meet their own needs [1]. Industrial operations offer significant opportunities for sustainability improvements through increased efficiency, productivity, waste reduction, and predictive maintenance [2]. Emerging technologies such as the Industrial Internet of Things (IIoT) and artificial intelligence (AI) enable intelligent and optimized operations through data acquisition, predictive modeling, and automation [3]. Several studies have demonstrated the potential of IIoT and AI to revolutionize industrial operations by improving efficiency, productivity, and quality as measured by standard sustainability criteria [4]. This literature review examines how IIoT and AI can be used to revolutionize generator (genset) operations by optimizing fuel usage and enabling predictive maintenance. IIoT, AI, and other advanced technologies have significant potential to improve sustainability, efficiency, and productivity in health care and transportation [5,6,7,8,9]. However, their full potential can only be achieved by addressing the challenges related to security, privacy, standardization, and interoperability. Research related to the use of these technologies in optimization problems shows promising results and can have significant applications in various fields [10,11,12,13,14,15]. The references reported herein demonstrate the potential for new technologies such as IIoT and AI to facilitate sustainability improvements in industrial operations. AI technologies such as artificial neural networks (ANNs) and genetic algorithms (GAs) can be used to optimize energy and resource consumption [16,17,18,19]. Predictive maintenance enabled by IIoT and machine learning has significant potential to reduce waste and improve productivity [20,21,22,23,24,25,26]. Several studies have reported the successful implementation of AI techniques to optimize fuel usage and improve efficiency in various applications. Garralaga et al. [16] conducted a case study on a power generation system integrating photovoltaics, battery storage, and gensets, achieving reduced fuel consumption and increased overall efficiency. Ziolkowski et al. [17] used ANNs to predict the fuel consumption of vehicles based on technical parameters, providing accurate predictions that could improve vehicle efficiency. Sharma et al. [18] employed a gene expression programming model to predict the performance and emissions of a diesel engine running on linseed biodiesel blends, with the model accurately predicting engine performance and emissions. Li et al. [19] optimized engine efficiency and emissions using ANNs and GAs, resulting in improved engine efficiency and reduced emissions. Predictive maintenance has been shown to reduce downtime and maintenance costs [20]. Thakur et al. [21] discussed the use of IIoT in utility operations, finding improved efficiency and reliability. Björklöf et al. [22] implemented IIoT to improve overall equipment effectiveness and reduce downtime. Kalsoom et al. [23] conducted a systematic review and found that IIoT improved efficiency, productivity, and quality in manufacturing. Sasikumar et al. [24] used AI and IIoT for a sustainable smart industry, improving security and energy efficiency. Zheng et al. [25] applied AI to analyze data from gensets and equipment to predict failures and maintenance needs, with IIoT enabling real-time monitoring and control of gensets for efficient maintenance. Ramesh et al. [26] reported cost savings and improved availability when implementing predictive maintenance systems in remote plants. Katreddi et al. [27] explored AI for predictive maintenance, energy management, and autonomous driving in heavy-duty trucks, finding that AI improved efficiency and safety. Theissler et al. [28] applied machine learning for predictive maintenance in the automotive industry, resulting in improved efficiency and reliability. Achouch et al. [29] provided a summary of predictive maintenance in Industry 4.0, models, and challenges, and found that predictive maintenance improved efficiency and reliability. Artificial neural networks have been applied to forecast engine performance and emissions in various contexts. Fu et al. [30] used ANNs to forecast engine performance and emissions in spark ignition engines, resulting in improved efficiency and performance. Castresana et al. [31] assessed a diesel engine using ANNs and thermodynamic simulation, finding that ANNs improved efficiency and performance. Ahmad et al. [32] proposed using ANNs and GAs to predict energy consumption and optimize management, resulting in reduced energy consumption and improved management. This literature review demonstrates that IIoT and AI can revolutionize IC genset operations by optimizing fuel usage and enabling predictive maintenance. The studies reviewed here show that AI technologies such as artificial neural networks and genetic algorithms can be successfully applied to optimize energy and resource consumption, while predictive maintenance enabled by IIoT and machine learning can significantly reduce waste and improve productivity. These findings establish that IIoT and AI have the potential to facilitate sustainability improvements in industrial operations, including genset operations, by improving fuel savings and predictive maintenance. The current state of research shows promising results in the application of IIoT, AI, and other advanced technologies to optimize resource usage, improve efficiency, and reduce waste in various fields, including health care, transportation, and industrial operations. The full potential of these technologies can be realized by addressing challenges related to simplicity, security, privacy, standardization, and interoperability.

3. Methodology

Figure 1 shows an outline of the methodology employed in this research paper. The aim of this study was to investigate the effectiveness of an AI–IIoT-based engine-monitoring and control system in reducing fuel consumption and enhancing predictive maintenance in a Caterpillar diesel engine.

The research began with the design and development of an AI–IIoT-based engine-monitoring and control system. The system comprised the following components:

• Sensors: A set of sensors was installed on the genset to collect real-time data on engine parameters, such as fuel consumption rate, engine RPM, and temperature;
• Data Acquisition and Communication: An IIoT gateway was employed to acquire data from the sensors and transmit them to a cloud-based server for storage and analysis;
• Data Analysis and AI Algorithms: The collected data were analyzed using machine learning algorithms to identify patterns and correlations between engine parameters and fuel consumption. The AI algorithms were trained to predict fuel consumption rates and maintenance requirements based on the analyzed data;
• Control and Optimization: Based on the insights derived from the AI algorithms, the system was designed to automatically adjust engine settings to optimize fuel consumption and schedule predictive maintenance tasks.

3.1. Experimental Setup

This study was conducted on a Caterpillar diesel genset, which was initially operated using manual controls and periodic maintenance. The AI–IIoT-based engine-monitoring and control system was then implemented, and the genset’s performance was monitored over a period of 2000 h.

3.2. Data Collection

Data on fuel consumption rates and engine parameters, including engine RPM and temperature, were collected before and after the implementation of the proposed system. The data were used to create maps of engine temperature and RPM relative to fuel consumption rates, which were analyzed to assess the effectiveness of the AI–IIoT system in reducing fuel consumption and enhancing predictive maintenance.

3.3. Data Analysis

The collected data were analyzed using statistical methods and machine learning algorithms. The fuel consumption rate (L/kWh) was analyzed to provide a fair comparison of the diesel genset’s performance before and after implementing the AI–IIoT system. The distribution of fuel consumption rates was also examined to assess the consistency and stability of engine performance. Additionally, the relationship between engine parameters, such as engine RPM and temperature, and fuel consumption rates was investigated. Maps of engine temperature and RPM relative to fuel consumption rates were created and analyzed to evaluate the effectiveness of the AI–IIoT system in optimizing engine performance and reducing fuel consumption.

3.4. Evaluation and Validation

The results of the study were evaluated by comparing the fuel consumption rates and engine performance before and after implementing the AI–IIoT system. The effectiveness of the system in reducing fuel consumption and enhancing predictive maintenance was assessed based on the observed improvements in engine performance and the consistency and stability of fuel consumption rates. The study’s findings were validated through a thorough analysis of the collected data and the application of machine learning algorithms. The results were also compared with existing literature on the applications of IIoT and AI in industrial settings to ensure the reliability and relevance of the research. In conclusion, the methodology employed in this research paper aimed to provide a comprehensive and rigorous approach to investigating the potential of AI and IIoT technologies to revolutionize IC genset operations, with a focus on fuel savings and predictive maintenance.

4. Sustainable Design and Configuration

To achieve the research objectives of revolutionizing CAT genset operations with IIoT and AI in a sustainable manner, a systematic design and configuration of the research platform was required. The design involves identifying the inputs, outputs, data collection mechanisms, analytics engines, and user interfaces to build an end-to-end system while considering the environmental, social, and economic aspects of sustainability. The following steps outline the sustainable design and configuration of the research study:

• Identifying the research objectives and research questions to be answered by the study, considering the sustainability benefits, such as reduced emissions, resource conservation, and cost savings;
• Selecting the appropriate industrial devices and equipment to be used in the study, such as an energy-efficient CAT genset, sensors, and other necessary hardware, prioritizing eco-friendly and energy-saving options;
• Installing and configuring the IIoT platform, including the KEPware OPC server and the necessary drivers and libraries, ensuring the platform’s energy efficiency and minimal environmental impact;
• Configuring the wireless Ethernet connection between the industrial devices and the IIoT platform, including setting IP addresses, network settings, and security settings, while minimizing the energy consumption of the communication infrastructure;
• Setting up the data collection and storage system, including the use of databases and cloud storage, prioritizing energy-efficient and environmentally friendly options for data centers;
• Designing the user interface for the platform, including the real-time data display, device control, alarm and event monitoring, historical data, configuration and settings, security and access control, and help and documentation, with a focus on promoting sustainable practices and energy conservation;
• Implementing the AI and machine learning algorithms for predictive maintenance and fuel savings analysis, contributing to resource conservation and emissions reduction;
• Testing and evaluating the performance of the system, including monitoring the fuel savings, predictive maintenance capabilities, and sustainability benefits, such as reduced emissions and resource conservation;
• Deploying the system on a standalone machine or in a cloud-based environment, ensuring proper configuration, connectivity to the industrial devices, and alignment with sustainability objectives.

To practically implement a machine learning model for prediction of potential maintenance issues and optimization of fuel consumption with sustainability in mind, first, data that account for environmental, social, and economic factors are collected and analyzed. In this case, the main feature is the fuel consumption data, specifically the ratio of liters consumed per kilowatt-hour of energy produced (L/kWH). These data can be gathered from the CAT genset’s onboard monitoring system, as shown in Figure 2.

Next, the model is trained with a machine learning framework such as ML.Net. The training process involves feeding the collected data into the model and allowing it to identify patterns and anomalies in the data. This process can be performed using techniques such as supervised, unsupervised, or reinforcement learning. Once the model has been trained, it can then be used to predict potential maintenance issues and optimize fuel consumption. For example, suppose the model detects an anomaly in the liter/KWH data. In that case, it may predict that there is a problem with the genset’s fuel injection system. This prediction can then be used to schedule maintenance and address the issue before it causes significant downtime. The model can also be used to optimize fuel consumption by identifying patterns in the data that indicate where fuel is being wasted. For example, suppose the model detects that fuel consumption is higher than normal during certain times of the day, which could indicate that the genset is being operated with inefficient settings. By addressing these issues and making adjustments, fuel consumption can be reduced, which can lead to cost savings for the organization.

Overall, implementing a machine learning model for prediction of potential maintenance issues and optimization of fuel consumption is a practical and effective way to improve the performance and efficiency of CAT gensets. By using the right tools and techniques, organizations can leverage the power of IIoT and AI to revolutionize their operations and achieve significant cost savings. Finally, implementing a machine learning model for prediction of potential maintenance issues and optimization of fuel consumption is a practical and effective way to improve the performance and efficiency of CAT gensets while promoting sustainability. By following the sustainable design and configuration steps outlined above, connecting the gensets to an IIoT platform, and applying AI techniques, significant cost savings, operational improvements, reduced emissions, and resource conservation can be achieved. The results and impacts of this study can be measured by monitoring the reduction in maintenance costs, the decrease in fuel consumption, the increased availability and uptime of the gensets, and the overall sustainability benefits.

4.1. Standalone IIoT Platform

Our standalone IIoT platform is a self-contained system not dependent on any other platform or service. It includes all the necessary hardware and software components to collect, process, and transmit data from industrial devices, specifically CAT gensets. The platform is built using VB.Net and KEPware OPC, which allows for collection of data from the genset’s ECM and the radar-level transmitter connected to the horizontal cylindrical tank. The data are then processed locally using ML.Net to create an AI model to predict and recommend fuel savings and predictive maintenance adjustments. The platform also can store and visualize the collected data and pare the data to a suitable structure to create an AI model. One of the key advantages of our standalone IIoT platform is that it does not require an Internet connection to operate, which can be beneficial in industrial environments where Internet connectivity is unreliable or unavailable. Additionally, it can process data locally, reducing the amount of data that needs to be transmitted over the network and can help reduce bandwidth requirements. Furthermore, it can perform real-time data processing, which is crucial for applications that require immediate action based on sensor data. The platform also allows for operation in remote locations, as it does not require a cloud connection.

We developed a standalone IIoT platform that was crucial in realizing our objectives. The platform was built using VB.Net and KEPware OPC. It is the backbone for collecting, analyzing, and visualizing data from the genset’s engine control module (ECM) and the radar-level transmitter. One of the primary functions of the standalone IIoT platform is data collection. The platform communicates with the genset’s ECM through the Modbus RTU protocol to collect real-time data such as engine speed, fuel consumption, and temperature. The platform also communicates with the radar-level transmitter to collect data on the liquid level in the cylindrical tank connected to the genset. These data are then stored in a suitable format and made available for further analysis. The platform also plays a significant role in analyzing the collected data. The platform uses ML.Net to create an AI model based on the data passed from the platform. The AI model is then used to predict and recommend adjustments for the genset’s operations. This helps reduce fuel consumption and increase the genset’s efficiency, resulting in cost savings for the user. Another important function of the standalone IIoT platform is data visualization. The platform presents the collected data in an easy-to-understand format, providing users with real-time insights into the genset’s performance. This helps users to identify patterns and trends that can be used to optimize the genset’s operations. In conclusion, the standalone IIoT platform developed in our research plays a crucial role in realizing our objective of revolutionizing CAT genset operations with IIoT and AI. The platform’s data collection, AI-based analysis, and data visualization enabled us to achieve significant fuel savings and improve the genset’s efficiency through predictive maintenance. This demonstrates the potential of IIoT and AI in optimizing the operations of industrial systems, such as gensets.

4.1.1. Implementation of Standalone IIoT Platform

Building a standalone IIoT platform requires a combination of hardware and software components. In this case, our platform was implemented using VB.Net and KEPware OPC. The first step in building the platform was to install and set up the KEPware OPC server on the machine, including the appropriate drivers for the SCADAPack 535E device. This step ensured that the KEPware OPC server could communicate with the SCADAPack 535E device, a crucial part of the platform.

Next, we configured the Nanostation M5 and M2 devices to establish a wireless Ethernet connection to the SCADAPack 535E device. This involved configuring these devices’ IP addresses, network settings, and security settings. This step was critical, as it allowed the data from the SCADAPack 535E device to be transmitted wirelessly to the platform.

We then created a new VB.Net project and added references to the KEPware OPC server and other necessary libraries. This step allowed us to use the KEPware OPC server’s API to establish a connection to the SCADAPack 535E device over the wireless Ethernet connection. This connection allowed us to read and write data from the SCADAPack 535E device, such as reading sensor values or writing commands to control the device.

We used VB.Net to create a user interface for the platform, allowing users to interact with and view the data from the SCADAPack 535E device. This included creating forms to display the data and buttons to send commands to the device. Additionally, we used VB.Net to add any necessary logic or functionality to the platform, such as data processing or alert systems.

We then thoroughly tested the platform to ensure it was functioning as intended, stable, and reliable. This step was critical, as it helped us identify and fix any bugs or issues before the platform was deployed.

Finally, the platform was deployed on a standalone machine, either on-premises or cloud-based. We ensured that it was properly configured to run as a service and connected to the SCADAPack 535E device over the wireless Ethernet connection. This step ensured that the platform was fully operational and ready to monitor and control the CAT genset operations.

4.1.2. IIoT Platform User Interface

Once implemented, our standalone IIoT platform offers a user interface that allows users to interact with and view data from connected industrial devices. The interface includes key features for real-time data display, device control, alarm and event monitoring, historical data, configuration and settings, security and access control, and help and documentation.

In terms of real-time data display, users can view real-time data from industrial devices, such as sensor readings or process values. These data are displayed in a variety of formats, including text, charts, and gauges, to provide an intuitive and user-friendly experience.

Device control is also a key feature of our platform; users are able to control industrial devices by sending commands to start or stop a process. This can be accomplished through various controls, such as buttons, toggle switches, and other interactive elements.

Alarm and event monitoring is also an important aspect of our platform. Users can view any alarms or events generated by the industrial devices, such as a sensor reading going out of range or a process shutting down unexpectedly. These data are displayed in a list or table, with the ability to acknowledge or clear alarms.

In addition to real-time data and monitoring, our platform offers the ability to view historical data from industrial devices. These data can be displayed in various formats, such as graphs, tables, or trend charts, providing a comprehensive view of the device’s performance over time.

Users can also configure and set up industrial devices, such as IP addresses or network settings, through a settings menu or a separate configuration page. Furthermore, security and access control represent another important feature, allowing users to control who has access to the platform and to what level of access. This can be accomplished through a login system and role-based access control.

Help and documentation are available, including a user manual or tutorial videos. Additionally, the platform is meant to be used on mobile devices. In that case, the UI is mobile-responsive, meaning that the layout and functionality adjust to mobile the screen size and resolution of mobile devices.

5. Data Collection and Analysis

Prior to the implementation of our automated data collection and analysis platform, the genset’s performance was reliant on manual controls and sporadic maintenance, leading to suboptimal operation and excessive fuel consumption. However, with the integration of this cutting-edge technology, the genset’s performance was optimized, resulting in significant fuel savings and a remarkable improvement over the previous manual approach. Moreover, our low-cost solution demonstrated the potential for broader applications of IIoT and AI in industrial settings, particularly in the optimization of heavy-duty power generation equipment such as gensets. Despite the extensive exploration of these technologies, their potential to revolutionize the performance of gensets and similar equipment remains largely unexplored, and our study provides a valuable contribution to this area of research. Data collection for our research was conducted using various industrial automation tools and devices. One of the key components used in this process was the SCADAPACK 535E shown in Figure 3, a programmable logic controller (PLC) that was used to interface with the genset’s engine control module (ECM) and collect data on the genset’s performance. The SCADAPACK 535E was also used to interface with a radar-level transmitter, which was used to measure the fuel level in the genset’s fuel tank. We used the Modbus RTU protocol as a communication protocol to interface between the SCADAPACK 535E and the other devices. This allowed us to easily collect and transmit data from the genset’s ECM and the radar-level transmitter to our data collection and analysis software.

In addition to the SCADAPACK 535E and the genset’s ECM and radar-level transmitter, other sensors and devices such as temperature, pressure, and vibration sensors were used to gather more data related to the genset operation. All these data were collected and stored in a database for further analysis.

5.1. Genset Data Collection

The engine control module (ECM) is a central component of the genset’s control system. It is responsible for monitoring parameters such as engine speed, temperature, and fuel consumption and adjusting its performance to ensure optimal operation. The ECM is connected to a wide range of sensors and actuators that provide data and allow it to control various aspects of the genset’s operation. In our study, the genset’s ECM played a crucial role in providing real-time data on the genset’s performance, which were used for fuel savings and predictive maintenance analysis. The ECM was connected to a SCADAPACK 535E programmable controller, which was used to gather data from the genset’s ECM and other sensors, such as the radar-level transmitter. The data collected by the SCADAPACK 535E were then sent to a cloud-based platform for further analysis using AI and IIoT.

The ECM was also responsible for controlling the genset’s operation based on the data and instructions received from the SCADAPACK 535E. For example, the ECM adjusted the fuel injection rate or engine speed to optimize fuel consumption and reduce emissions. This helped us to ensure that the genset’s operation was always optimal, resulting in significant fuel savings. In summary, the genset’s ECM played an important role in our study by providing real-time data on the genset’s performance, which were used for fuel savings and predictive maintenance analysis. It also helped us to optimize the genset’s operation and reduce emissions.

5.2. Fuel Consumption Data

The CAT genset is connected to a horizontal cylindrical tank with a cone on both sides, as shown in Figure 4. A radar-level Rosemont 5408 transmitter from Emerson is used to measure the liquid level in the tank, which is passed to the SCADAPACK 535E device. The Rosemount 5408 is a non-contacting radar-level transmitter that provides high-accuracy measurement. It uses guided wave radar technology to measure the level of liquids, solids, and slurries and interfaces with applications. It has an accuracy of ±3 mm, which allows for precise level control and monitoring. The 5408 has a measurement range of up to 60 m (200 feet) and is suitable for high-temperature and high-pressure applications. It also has diagnostic tools that proactively monitor transmitter health and the level measurement itself. The high accuracy and reliability of the Rosemount 5408 make it suitable for a wide range of applications for which precise level control is critical, such as in reactors, separators, and storage tanks in the oil and gas, chemical, and petrochemical industries.

In the context of a horizontal cylindrical tank with cone ends, the volume of liquid can be calculated by measuring the liquid level (h) and the tank’s dimensions. The formula for the volume of liquid in a cylindrical tank with cone ends is:

• Define the variables you will be using in the problem. In this case, you will need to define the volume of the liquid in the tank (V), the radius of the cylindrical portion of the tank (r), the height of the liquid in the tank (h), the radius of the cones at either end of the tank (the same as the tank radius ) (r), the flow rate of the liquid being added to the tank (F), and the rate at which liquid is being removed from the tank (L);
• Calculate the volume of the liquid in the tank at time t using the following equation to generate the volume chart shown in Figure 5:

  $$\begin{array}{cc}\hfill V\left(t\right)=L\left({cos}^{-1}\right.& \left.\left(\frac{r-h}{r}\right)r^2-(r-h)\sqrt{}\left(2rh-h^2\right)\right)\hfill \\ \hfill & +2\left(\frac{d{r}^2}{3}\left\{\frac{\pi }{2}-2k\sqrt{1-k^2}-{sin}^{-1}k+k^3{cosh}^{-1}\frac{1}{k}\right\}\right)\hfill \end{array}$$

  (1)

  where ${\displaystyle k=1-\frac{h}{r}}$;
• Calculate the volume of the liquid in the tank at time $t+\Delta t$ using the same equation;
• Calculate the change in volume of the liquid in the tank over the time interval using the following equation:

  $$\Delta V=V(t+\Delta t)-V\left(t\right);$$

  (2)
• Use the following equations to calculate $F\left(t\right)$ and $L\left(t\right)$ at each time interval:

  $$F\left(t\right)=\frac{\Delta V}{\Delta t}$$

  (3)

  $$L\left(t\right)=-\frac{\Delta V}{\Delta t};$$

  (4)
• To calculate the total volume of liquid added to and removed from the tank over a given period of time, sum the values of $F\left(t\right)$ and $L\left(t\right)$ over each time interval in the period using the following equations:

  $$V_{added}=\sum F\left(t\right)\Delta t$$

  (5)

  $$V_{removed}=\sum L\left(t\right)\Delta t;$$

  (6)
• Repeat the calculations for each time interval in the period to estimate the total volume of liquid added to and removed from the tank over the time period.

5.3. Data Analysis

Data analysis is crucial in realizing the goal of revolutionizing CAT genset operations with IIoT and AI. Specifically, the focus is on fuel savings and predictive maintenance. The primary metric for measuring fuel savings is the liter per kilowatt-hour (L/kWH) ratio. The following is a flow chart outlining the steps for analyzing these data:

• Collect data on fuel consumption (measured in L/kWH) from the CAT genset using the IIoT system;
• Clean and preprocess the data to ensure accuracy and consistency;
• Use statistical analysis methods such as regression analysis to identify patterns and relationships in the data;
• Use machine learning algorithms such as decision trees and random forests to model the data and predict fuel consumption;
• Compare the predicted fuel consumption to actual fuel consumption to evaluate the accuracy of the model;
• Use the model to identify potential areas for improvement in terms of fuel savings and predictive maintenance;
• Implement changes and continue to monitor fuel consumption to track the effectiveness of these improvements;
• Repeat the analysis process periodically to adapt to any changes in the system and to continue to optimize fuel savings and predictive maintenance.

6. AI Model

Figure 6 shows the general architecture of the AI system used to generate suitable maintenance recommendations. The AI model utilizes a variety of inputs, such as engine RPM, engine hours, coolant temperature, fuel level, fuel consumption rate, battery voltage, alternator voltage, oil pressure, air intake temperature, exhaust temperature, throttle position, engine load percentage, engine speed (frequency), engine oil life, fuel filter life, air filter life, oil filter life, battery state of charge, fuel type and grade, oil filter age, air filter age, fuel filter age, coolant filter age, battery age, belt age, hose age, spark plug age, alternator brush age, starter motor brush age, and engine diagnostic trouble codes (DTCs). These inputs are collected from various sensors, including the engine control module (ECM) and other relevant sources. The AI model then uses machine learning algorithms, such as those provided by the ML.Net library, to analyze the collected data, predict potential maintenance issues, and optimize fuel consumption. The model can identify patterns and anomalies in the data and make recommendations for adjustments to the genset’s operation to improve efficiency and reduce fuel consumption.

6.1. Decision Trees

Decision trees are a type of algorithm that recursively splits the data into subsets based on the attribute values, ultimately leading to a prediction. The main parameters to consider when using decision trees include the splitting criterion (e.g., Gini impurity and information gain), the maximum depth of the tree, and the minimum number of samples required to split a node. Decision trees may be suitable because they can handle both numerical and categorical data, provide intuitive visualizations, and are relatively easy to interpret. However, they can be prone to overfitting, especially when the tree depth is not properly constrained.

6.2. Random Forests

Random forests are an ensemble method that combines multiple decision trees to improve prediction accuracy and reduce overfitting. The main parameters to consider when using random forests include the number of trees in the forest, the maximum depth of each tree, and the minimum number of samples required to split a node. Random forests may be a good option because they can capture complex interactions between variables, are less prone to overfitting than single decision trees, and typically provide high prediction accuracy.

6.3. Gradient Boosting Machines

Gradient boosting machines (GBMs) are another ensemble method that builds multiple decision trees sequentially, with each new tree aiming to correct the errors made by the previous trees. The main parameters to consider when using GBMs include the learning rate, the number of trees, the maximum depth of each tree, and the minimum number of samples required to split a node. GBMs may be suitable for this study because they often provide high prediction accuracy and can handle both numerical and categorical data. However, they can be more sensitive to parameter tuning and may require more computational resources than decision trees or random forests.

6.4. Best Algorithm Selection

The goal of this study was to build an AI model that anticipates potential maintenance issues and enhances fuel efficiency for CAT generators. To achieve this, the ideal algorithm for the AI model needed to be identified. Traditionally, the best algorithm is selected by systematically comparing possible algorithms based on the accuracy of predictions, computing needs, and how easy the algorithm’s results are to understand. However, ML.Net’s AutoML feature was used instead to pick the algorithm and adjust its settings. ML.Net’s AutoML feature was utilized to determine the best algorithm for the AI model. This feature automates algorithm selection by exploring many algorithms and their settings to ultimately choose the highest-performing model based on metrics. The AutoML feature works as a “black box”, meaning the user cannot see how algorithms are selected. This approach removed the need to manually select algorithms and adjust settings, which can take a long time and require extensive AI expertise. It also decreased the risk of choosing a subpar algorithm and enabled the construction of an AI model that improves CAT generator efficiency and reliability, significantly reducing costs for operators. The AI model developed this way can predict potential maintenance issues before they happen and optimize fuel usage, enabling more efficient and cost-effective operation of CAT generators. The AutoML feature in ML.Net allowed us to focus on other aspects of AI model development while still achieving excellent performance.

In summary, the AutoML feature in the ML.Net framework was used to automatically select the best algorithm for the AI model, operating as a “black box” to simplify algorithm selection. This led to an AI model that improves CAT generator efficiency and reliability, considerably cutting costs for operators. Using AutoML eliminated the need to manually select algorithms and adjust settings, which can require extensive time and AI expertise.

6.5. ML.Net

ML.Net is a machine learning framework developed by Microsoft that allows developers to build custom machine learning models using C# or F# without requiring expertise in machine learning. In our research on “Revolutionizing CAT Genset Operations with IIoT and AI: A Study on Fuel Savings and Predictive Maintenance”, ML.Net was used to train a machine learning model using the collected and analyzed sensor data from the genset. The model aims to predict potential maintenance issues and optimize fuel consumption. The training process involves feeding the model input data for training using various algorithms, such as linear regression, decision tree, random forest, and others, then selecting the most suitable algorithm based on the prediction accuracy. Once the model is trained, it can be used to identify patterns and anomalies in the data and make predictions about potential maintenance issues and fuel consumption optimization. For example, the model can predict when a certain component is likely to fail, which can be used to schedule preventative maintenance. Furthermore, it can predict the optimal time to refuel the genset based on the consumption rate. The selection of the best algorithm is based on the prediction accuracy, the available computational resources, the tradeoffs between accuracy and resource consumption, and the interpretability of the model. As mentioned earlier, cross validation and grid search were employed to determine the optimal parameters for each algorithm and make an informed decision about the most suitable algorithm for the AI model. In summary, ML.Net played a crucial role by providing a tool to train the machine learning model that can optimize fuel consumption and predict potential maintenance issues based on the sensor data collected from the genset. The integration of detailed information about candidate algorithms, their parameters, and the process of algorithm selection addresses concerns, providing a comprehensive understanding of the AI model implementation.

6.5.1. Testing and Validation of AI/ML Methods with ML.Net

ML.Net simplifies the testing and validation process for the AI/ML methods employed in this study. The framework automates various aspects of the process, ensuring the effectiveness and reliability of the AI model in predicting potential maintenance issues and optimizing fuel consumption.

6.5.2. Data Splitting

ML.Net provides a built-in method for splitting the dataset into training and testing sets. The TrainTestSplit function can be used to divide the dataset into two parts: a training set and a testing set. By default, 80% of the data is used for training, and the remaining 20% is reserved for testing. This function ensures that the data are split randomly and uniformly.

6.5.3. Model Evaluation Metrics

ML.Net includes a set of evaluation metrics that can be used to assess the accuracy of the AI model’s predictions. These metrics can be calculated using the RegressionMetrics class, which provides methods for computing mean absolute error (MAE), root mean square error (RMSE), and R-squared ($R^2$).

6.5.4. Cross Validation

The ML.Net framework also supports cross validation through the CrossValidate function. This function automates the process of dividing the training dataset into k equal-sized subsets or “folds” and iteratively training and testing the model on these folds. The average performance across all k iterations is calculated to provide a more reliable estimate of the model’s performance.

6.5.5. Algorithm Selection and Hyperparameter Tuning

ML.Net offers a feature called AutoML, which automates the process of selecting the most suitable machine learning algorithm and tuning its hyperparameters. The AutoML class can be used to automatically explore various algorithms and their hyperparameters, selecting the best-performing model based on the evaluation metrics.

6.5.6. Model Validation

Once the AI model has been trained and optimized using ML.Net’s automated features, it can be tested on the reserved testing dataset. The model’s predictions can be compared to the actual performance of the genset during the testing period, and the evaluation metrics can be calculated using the RegressionMetrics class to assess the accuracy and effectiveness of the AI model. In conclusion, ML.Net automates various aspects of the testing and validation process, ensuring the reliability and effectiveness of the AI/ML methods used in this study. The results of this process demonstrate the AI model’s ability to accurately predict potential maintenance issues and optimize fuel consumption, which is the crucial contribution of this article.

7. Results

To assess the effectiveness of the research approach, a comparison of the fuel consumption of a Caterpillar (CAT) genset before and after implementing the IIoT and AI system was conducted. Prior to this study, the genset relied solely on manual controls and periodic maintenance. The introduction of an automated data collection and analysis platform enabled the optimization of the genset’s performance, resulting in a significant reduction in fuel consumption. Over a span of 2000 h, a 12% decrease in the liters per kilowatt-hour ratio was observed, demonstrating a substantial improvement compared to the previous manual approach. This research contributes to the broader applications of IIoT and AI in industrial settings; while these technologies have been extensively explored, their potential for optimization of heavy-duty power generation equipment, such as gensets, remains largely untapped. Our work illustrates how an integrated IIoT and AI solution can drive higher performance, lower emissions, and reduced costs for gensets through continuous monitoring, automated control, and predictive maintenance. The methods and results of this study serve as a model for the development and deployment of IIoT and AI systems to enhance operations and sustainability across various industrial sectors. With further development, the platform we have built could be adapted to other types of industrial engines and equipment. The results of this study are presented in this section. The aim of the study was to investigate the effectiveness of the proposed AI–IIoT-based engine-monitoring and control system in reducing fuel consumption in a Caterpillar diesel engine. In this study, we collected data on fuel consumption rates and engine parameters, including engine RPM and temperature, before and after the implementation of the proposed system.

7.1. Fuel Consumption Rate

The fuel consumption rate (L/kWh) was analyzed to provide a fair comparison of the diesel genset’s performance before and after implementing the AI–IIoT system. Throughout this section, several figures employ a color gradient from blue to red to represent varying levels of fuel consumption rate, measured in liters per kWh. Specifically, data points with lower fuel consumption rates are represented in blue, while data points with higher fuel consumption rates are depicted in red. Intermediate values are represented by colors transitioning between these two extremes. This consistent color-coding provides a visual representation of the range and distribution of fuel consumption rates observed during our study, allowing for an immediate and intuitive understanding of the data presented.

Figure 7 displays a 2D scatter of the fuel consumption rate over 1000 h before implementing the proposed system. The data show that fuel consumption varied widely over this period, with no clear trend or pattern. However, Figure 8 presents the distribution of fuel consumption rate, indicating that the data were normally distributed with a high variance.

After implementing the proposed system, fuel consumption rates were measured over another 1000 h.

Figure 9 shows a 2D scatter of the fuel consumption rate, which indicates that the average decreased by 12% compared to the preimplementation period. Figure 10 demonstrates that the distribution of fuel consumption rates also followed a normal distribution, indicating that the data were normally distributed with a symmetrical bell curve and low variance.

7.2. Engine Parameters

In this study, we also collected data on engine parameters, including engine RPM and temperature, before and after implementing the proposed system. The data were used to create maps of engine temperature and RPM relative to fuel consumption rates.

Figure 11 and Figure 12 show 3D scatter maps of engine temperature and fuel consumption over time before and after the implementation of the proposed system, respectively. The maps suggest a clear improvement in engine performance after implementing the proposed system, with more consistent and stable engine temperature and fuel consumption patterns.

Similarly, Figure 13 and Figure 14 show 3D scatter maps of engine RPM and fuel consumption over time before and after the implementation of the proposed system, respectively. The maps suggest a clear improvement in engine performance after implementing the proposed system, with more consistent and stable engine RPM and fuel consumption patterns.

Finally, Figure 15 and Figure 16 show 3D scatter maps of the fuel consumption maps over engine RPM and temperature before and after implementing the proposed system, respectively. After implementing the proposed system, the maps suggest a more consistent and stable relationship between fuel consumption, engine RPM, and temperature.

The results suggest that the proposed AI–IIoT-based engine performance optimization system effectively improves fuel efficiency and stabilizes engine performance. The system could be further improved by continuously monitoring the real-time maintenance recommendations introduced by the system to maintain smooth performance of diesel engines and feedback mechanisms to adapt to changing engine conditions and external factors.

7.3. Future Research Directions

Future research should focus on developing new models, techniques, and frameworks that can leverage the capabilities of IIoT and AI to further enhance sustainability, efficiency, and productivity in IC genset operations and other industrial contexts. Advanced optimization algorithms, such as hybrid heuristics, metaheuristics, adaptive algorithms, self-adaptive algorithms, and island algorithms, are an essential component of these models and frameworks, and they have the potential to significantly improve the efficiency and sustainability of power generation operations. By integrating these algorithms into IIoT and AI systems, operators can optimize power plant operations, predict and prevent equipment failures, and optimize the allocation of resources. In the power generation industry, these advanced optimization algorithms can help optimize the scheduling of power plant units, the allocation of resources, and the management of maintenance activities. They can also predict potential equipment failures before they occur, allowing operators to take preventive measures to avoid costly downtime and repairs, thus reducing maintenance costs and increasing the lifespan of power generation equipment. Furthermore, these algorithms can be used to optimize the allocation of resources in power generation systems, such as renewable energy sources, leading to a more sustainable and efficient power generation systems. Therefore, future research should further investigate the potential of advanced optimization algorithms and explore their applications in the power generation industry and other industrial contexts. Developing new models, techniques, and frameworks that can leverage the capabilities of IIoT and AI and integrating advanced optimization algorithms into these systems can significantly improve the efficiency, sustainability, and productivity of industrial operations.

8. Conclusions

In the quest to revolutionize and optimize internal combustion (IC) generator set (genset) operations, this research unveils the remarkable potential of fusing Industrial Internet of Things (IIoT) and artificial intelligence (AI) technologies. By crafting an autonomous IIoT platform that harnesses the power of cutting-edge innovations, we have demonstrated the feasibility of significantly enhancing fuel efficiency and foreseeing maintenance requirements. Our findings reveal that the implementation of IIoT and AI solutions paves the way for considerable fuel conservation, elevated performance through predictive maintenance, and pragmatic approaches for industries to refine processes and boost efficiency in IC genset operations. This study embodies inherent sustainability principles and advantages, encompassing energy efficiency, predictive maintenance, environmental impact reduction, and economic benefits. The integration of IIoT and AI technologies bolsters fuel efficiency, championing resource preservation and sustainable energy generation. Predictive maintenance curtails downtime, diminishes waste, and prolongs equipment lifespan, adopting a forward-thinking stance on maintenance that aligns with sustainability. Enhanced fuel efficiency and optimized operations also shrink the carbon footprint, mitigating environmental damage. Furthermore, the proposed technology amplifies profitability, fosters sustainable growth, and trims costs and risks, delivering economic benefits for businesses while promoting sustainability. This research signifies a monumental stride towards realizing a more sustainable and efficient energy generation landscape, lessening environmental impacts, and bolstering financial performance. As advancements and refinements continue to unfold, the proposed AI–IIoT-based engine performance optimization system harbors the potential to emerge as a transformative force within the power generation industry. Ultimately, the encouraging results of this study underscore the substantial sustainable and practical benefits of digital transformation in IC genset operations.