AI, ML, IoT, LLMs: TRANSFORMING MAINTENANCE IN THE MARINE INDUSTRY

 7.7 IMPLEMENTING AI, ML, IOT, AND LLMS FOR PREDICTIVE AND PRESCRIPTIVE MAINTENANCE IN THE MARINE INDUSTRY: A COMPREHENSIVE ANALYSIS.

By Aleksandar Pudar

Technical Superintendent and Planned Maintenance Supervisor Reederei Nord BV

Co-founder of "Out of Box Maritime Thinker Blog" and founder of Narenta Gestio Consilium Group.

7.7.1 INTRODUCTION

The marine industry is rapidly undergoing a digital transformation, thanks to advancements in technologies like Artificial Intelligence (AI), Machine Learning (ML), the Internet of Things (IoT), and Large Language Models (LLMs). These technologies are ushering in a new era of maintenance strategies - predictive and prescriptive maintenance, that have the potential to dramatically increase the efficiency, reliability, and lifespan of the machinery onboard a vessel. This analysis aims to explore the implementation of these cutting-edge technologies and methodologies in the context of the marine industry.

7.7.1.1 PRESCRIPTIVE MAINTENANCE

Prescriptive maintenance elevates the capabilities of predictive maintenance by not only forecasting potential issues but also suggesting necessary actions for the maintenance teams. These recommendations can be executed on-site or off-site by in-house teams or service providers working from remote service centres.

7.7.2 SMART TERMINALS

Appropriate measuring devices connect to sensors and data sources like vibration, temperature, and oil quality sensors to a computing server (onboard/office ) or a cloud platform; these devices collect raw data, and the software converts it into intelligent data, adding connectivity, security, and remote configurability for ease and speed of use.

The software is a comprehensive tool for demanding use cases with features like 8 Time Synchronous high-frequency channels with RPM scaling, special means for slowly rotating machinery, and user-friendly configuration; the software is used for asset health monitoring of large machines such as compressors, gearboxes, generators, supply pumps, cargo pumps and turbines, large blowers/fans, and other critical equipment.

7.7.3 THE ANALYTICS SOLUTION

The analytic solution combines embedded analytics and a server (onboard/onshore) or cloud service to provide a comprehensive analytical platform for maximising machine health. High-speed data is analysed locally, converted into smart data, and sent to the cloud for further analysis and storage.

·         Data Collection: The first step is to collect data from the various systems and machinery onboard a vessel; this can be done using sensors and IoT devices installed on critical equipment, such as engines, generators, pumps, and turbines. The collected data typically includes parameters like temperature, vibration, pressure, and RPM. Collecting more sophisticated data, such as acoustic, ultrasonic, or thermal imaging, is possible.

·         Centralised Data Management: After gathering the data, storing and managing it in a central location is important. , A robust data management system must handle substantial amounts of data from multiple sources in real-time to accomplish this. The system should also be capable of cleaning, processing, and structuring the data for analysis.

·         Predictive Analytics: The next step is applying algorithms to the collected data. Machine learning (ML) and artificial intelligence (AI) can be leveraged to build predictive models to forecast equipment failures or performance degradation. Various methods such as regression models, time series analysis, or complex techniques like deep learning can be utilised depending on the nature of the data and problem.

·         Large Language Models (LLMs) in Prescriptive Maintenance: LLMs like GPT-3 or GPT-4 can analyse maintenance logs, reports, and other text data to extract valuable insights or identify patterns. They can generate text based on their learned patterns, auto-generate reports, explain the analytics results, or even provide step-by-step instructions for maintenance tasks. LLMs can also identify anomalies in new data, which could indicate potential problems. Additionally, these tools can be utilised to create smart chatbots or virtual assistants that can engage with users, respond to their inquiries, offer details, or assist them in completing particular tasks. LLMs can also mine knowledge from large amounts of textual data, such as technical manuals, maintenance guides, or historical incident reports. This knowledge can then support decision-making in the prescriptive maintenance process.

·         Integration with CBM: Condition-Based Maintenance (CBM) can be made more effective using predictive and prescriptive analytics based on the real-time condition of equipment rather than predetermined schedules; this would help carry out maintenance activities only when needed, improving efficiency and reducing costs.

·         User Interface and Visualisation: Lastly, the solution should include a user-friendly interface and visualisation tools that allow operators to understand the analytics results, predictions, and prescriptions easily; this should consist of dashboards, alerts, and reports that provide real-time insights into the vessel's condition and performance.

This overall system should be capable of operating in real time, allowing for proactive interventions and adjustments. The main benefits would be enhanced operational efficiency, reduced maintenance costs, improved safety, and increased equipment lifespan. Implementing such an analytics solution would require a multidisciplinary approach involving marine engineering, data science, AI, and software engineering expertise.

7.7.4 USEFUL TOOLS

Cloud solutions may offer several purpose-built tools for data analysis, automation, and anomaly tracking. Tools such as the Trend Revealer help users detect issues early, while the Anomaly Revealer automatically sets value ranges for anomalies. Other tools include the Lubrication Indicator for detecting and managing lubrication-related issues and the Forecasting tool for better planning. Given the complexity of prescriptive maintenance in the marine industry and the involvement of Large Language Models (LLMs), various tools may be required to facilitate this process. The following are some suggestions:

·         Data Collection Tools: IoT devices, sensors, and data loggers can be used for real-time data collection from various machinery and systems.

·         Data Management Tools: For centralised data storage and management, solutions like Apache Hadoop or cloud-based platforms (AWS S3, Google Cloud Storage, Microsoft Azure) can be used. Tools like Pandas (Python library) or Talend could be beneficial for data cleaning and preprocessing.

·         Machine Learning and AI Tools: TensorFlow, PyTorch, and Keras are popular deep learning frameworks for building predictive models. For simpler machine learning models, tools like Scikit-learn could be used.

·         Large Language Models (LLMs): OpenAI's GPT-3 or GPT-4 can be used for natural language processing tasks. APIs provided by OpenAI can be integrated into the system to leverage these models' power.

·         Prescriptive Analytics Tools: Tools like MATLAB or Python's SciPy and NumPy libraries can be used for optimisation and simulation tasks required for prescriptive analytics.

·         Visualisation Tools: Tools like Tableau, Power BI, or Python libraries like Matplotlib and Seaborn can be used for creating interactive dashboards and visualisations to interpret the analytics results easily.

·         Chatbot and Virtual Assistant Tools: Frameworks like Rasa or Microsoft Bot Framework can be used to develop intelligent chatbots or virtual assistants. These can be integrated with the LLMs for natural language understanding and text generation capabilities.

·         Knowledge Extraction Tools: Tools like Elasticsearch or Python's NLTK and Spacy libraries can be used for mining knowledge from large text data.

·         Workflow Orchestration Tools: Given the complex data workflows involved in this system, tools like Apache Airflow or Luigi can help manage and automate these workflows.

·         Integration Tools: Since this system would require integrating various components, tools like Apache Kafka for real-time data streaming and APIs for connecting different software components can be useful.

7.7.5 FEATURES

The following are some key features that functioning smart analysis software should possess for predictive and prescriptive maintenance in the marine industry:

·         Real-Time Monitoring and Alerts: The software should monitor the various machinery systems. Alarms and messages should be triggered if potential issues or anomalies are identified, including abnormal thermal, vibration, or visual conditions.

·         Thermal, Vibration, and Visual Analysis: Advanced thermal, vibration and visual data analysis capabilities should be integrated; this includes using AI/ML models to analyse and predict potential issues based on these data. For example, excessive heat might indicate friction or a malfunction in the cooling system. Abnormal vibration patterns can be early signs of wear and tear or alignment issues. Visual analysis, such as from CCTV footage or specific machinery imaging, can detect visible issues like leaks, breakages, or corrosion.

·         Visualisations: The software should have visualisation capabilities to assist professionals in interpreting data and analytics results; this includes thermal, vibration, and visual data, presented in a way that makes trends and issues straightforward and easy to understand.

·         Third-Party Integration: The software should support integration with other third-party systems, including thermal, vibration, and visual monitoring systems, databases, cloud platforms, and Large Language Models like GPT-3 or GPT-4.

·         Scalability and Remote Device Management: The software should be scalable to handle increasing amounts of data as more machinery and systems are added and provide remote device management capabilities.

·         AI/ML Analysis and Alerts: The software should leverage AI and ML for predictive and prescriptive analytics, including the analysis of thermal, vibration, and visual data.

·         Support for Condition-Based Maintenance (CBM): The software should support CBM by providing real-time condition monitoring and analytics, including thermal, vibration, and visual condition data.

·         Diagnostic Capabilities: Diagnostic capabilities should be integrated to analyse the collected data, including thermal, vibration, and visual data, to diagnose the machinery's condition, predict potential failures, and prescribe necessary maintenance activities.

·         Lube Oil Sampling and Analysis: The software should support the scheduling and recording regular lube oil sampling and incorporate the results into its analysis.

·         Advanced Analytics for Lube Oil Data: The software should be capable of analysing the results of lube oil analysis and integrating it with thermal, vibration, and visual data for comprehensive equipment health assessment.

·         Alerts and Recommendations: Based on the analysis results and predictive models, the software should trigger alerts and provide recommendations for corrective actions.

·         Trend Analysis and Predictive Modelling: The software should analyse the thermal, vibration, lube oil, and visual data, tracking changes over time and predicting future trends.

·         Reporting and Visualisation: Detailed reports and visualisations of the thermal, vibration, lube oil, and visual analysis results should be provided, helping operators understand and interpret the data.

7.7.6 CONCLUSION

In conclusion, prescriptive maintenance, powered by AI/ML, IoT, and LLMs, redefines the marine industry's future. It enables the timely prediction of potential machinery failures and prescribes specific actions to be taken, thereby maximising operational efficiency and reducing downtime. Smart terminals have made it possible to collect and analyse data in real time, which allows for intelligent decision-making based on data. The analytical solution outlined presents a cohesive system that leverages these technologies to maximise machine health. The tools suggested are aimed at easing and streamlining this process, making the complex task of data collection, management, and analysis more manageable.

Moreover, the features listed for the functioning smart analysis software outline the comprehensive capabilities needed to implement predictive and prescriptive maintenance effectively. Indeed, integrating these technologies and methodologies into the marine industry promises unprecedented operational efficiency, safety, and equipment lifespan. As we move forward, it is evident that continuous learning and adaptation to these emerging technologies will be pivotal in maintaining a competitive edge in the ever-evolving marine industry.

 

References and Bibliography‌:

1.        Diego Galar Pascual (2015). Artificial Intelligence Tools. CRC Press.

2.        Medium. (n.d.). Medium. [online] Available at:

3.        https://medium.com/@connect.hashblock/the-ultimate-guide-to-decision-tree-algorithms-2ff42d7cf6c [Accessed 13 Feb. 2023].

4.        Pudar, A. (2023). AI-Based Predictive Maintenance - Start. Out of Box Maritime Thinker. Available at: https://outboxmaritimethinker.blogspot.com/2023/02/ai-based-predictive-maintenance-start.html [Accessed 13 Jun. 2023].

5.        www.rolls-royce.com. (n.d.). Rolls-Royce opens first Ship Intelligence Experience Space. [online] Available at: https://www.rolls-royce.com/media/press-releases/2017/27-11-2017-rr-opens-first-ship-intelligence-experience-space.aspx [Accessed 13 Feb. 2023].

Disclaimer:

Out of Box Maritime Thinker © by Narenta Gestio Consilium Group 2022 and Aleksandar Pudar assumes no responsibility or liability for any errors or omissions in the content of this paper. The information in this paper is provided on an "as is" basis with no guarantees of completeness, accuracy, usefulness, or timeliness or of the results obtained from using this information. The ideas and strategies should never be used without assessing your company's situation or system or consulting a consultancy professional. The content of this paper is intended to be used and must be used for informational purposes only.