Original Article

PREDICTING MACHINE FAILURES AND SYSTEM SECURITY USING MACHINE LEARNING AND DEEP LEARNING ALGORITHMS

 

 

INTRODUCTION

In general, ML techniques must be used in conjunction with other security measures to offer complete security for IoT systems. ML algorithms and methods have been applied in various tasks, including machine translation, regression, clustering, transcription, detection, classification, probability mass function, sampling, and estimation of probability density. Numerous applications utilize ML techniques and algorithms, such as spam identification, image and video recognition, customer segmentation, sentiment analysis, demand forecasting, virtual personal assistants, detection of fraudulent transactions, automation of customer service, authentication, malware detection, and speech recognition4.

In addition, IoT and ML integration can enhance the devices of IoT levels of security, thereby increasing their reliability and accessibility. ML’s advanced data exploration methods play an important role in elevating IoT securityfr om only providing security for communication devices to intelligent systems with a high level of security5. ML-based models have emerged as a response to cyberattacks within the IoT ecosystem, and the combination of Deep Learning (DL) and ML approaches represents a novel and significant development that requires careful consideration. Numerous uses, including wearable smart gadgets, smart homes, healthcare, and Vehicular Area Networks (VANET), necessitate the implementation of robust security measures to safeguard user privacy and personal information. The successful utilization of IoT is evident across multiple sectors of modern life6. By 2025, we expect that the IoT will have an economic effect of $2.70–$6.20 trillion. Research findings indicate that ML and DL techniques are key drivers of automation in knowledge work, thereby contributing to the economic impact. There have been many recent technological advancements that are shaping our world in significant ways. By 2025, we expect an estimated $5.2–$6.7 trillion in annual economic effects from knowledge labor automation7. This research study addresses the vulnerabilities in IoT systems by presenting a novel ML- based security model. The proposed approach aims to address the increasing security concerns associated with the Internet of Things. The study analyzes recent technologies, security, intelligent solutions, and vulnerabilities in IoT-based smart systems that utilize ML as a crucial technology to enhance IoT security. The paper provides a detailed analysis of using ML technologies to improve IoT systems’ security and highlights the benefits and limitations of applying ML in an IoT environment. When compared to current ML-based models, the proposed approach outperforms them in both accuracy and execution time, making it an ideal option for improving the security of IoT systems. The creation of a novel ML-based security model, which can enhance the effectiveness of cybersecurity systems and IoT infrastructure, is the contribution of the study. The proposed model can keep threat knowledge databases up to date, analyze network traffic, and protect IoT systems from newly detected attacks by drawing on prior knowledge of cyber threats.

 

The study comprises five sections: “Related works” section presents a summary of some previous research. “IoT, security, and ML” section introduces the Internet of Things’ security and ML aspects. “The proposed IoT framework architecture” section presents the proposed IoT framework architecture, providing detailed information and focusing on its performance evaluation. “Result evaluation and discussion” section provides an evaluation of the outcomes and compares them with other similar systems. We achieve this by utilizing appropriate datasets, methodologies, and classifiers. “Conclusions and upcoming work” section concludes the discussion and outlines future research directions.

Nowadays, with the rise of the Internet of Things (IoT), a large number of smart applications are being built, taking advantage of connecting several types of devices to the internet. These applications will generate a massive amount of data that need to be processed promptly to generate valuable and actionable information. Edge intelligence (EI) refers to the ability to bring about the execution of machine learning tasks from the remote cloud closer to the IoT/Edge devices, either partially or entirely. Examples of edge devices are smartphones, access points, gateways, smart routers and switches, new generation base stations, and micro data centers.

 

Some edge devices have considerable computing capabilities (although always much smaller than cloud processing centers), but most are characterized by very limited capabilities. Currently, with the increasing development in the area of MEMS (Micro–Electro– Mechanical Systems) devices, there is a tendency to carry out part of the processing within the data producing devices themselves (sensors) Kunst et al. (2019), Tessoni and Amoretti (2022), Vollert et al. (2021), O’Donovan et al. (2019)  . There are certainly several challenges involved in performing processing on resource-limited devices, including the need to adapt complex algorithms and divide the processing among several nodes. Therefore, in Edge Intelligence, it is essential to promote collaboration between devices to compensate for their lower computing capacity. Some synonyms of this concept found in the literature are: distributed learning, edge/fog learning, distributed intelligence, edge/fog intelligence and mobile intelligence Romeo et al. (2020), Boyes et al. (2018), Arena et al. (2022) . The leverage of edge intelligence reduces some drawbacks of running ML tasks entirely in the cloud, such as:

High latency Kaushik and Yadav (2023): offloading intelligence tasks to the edge enables achievement of faster inference, decreasing the inherent delay in data transmission through the network backbone;

·        Security and privacy issues Adhikari et al. (2018),Zhou and Tham (2018): it is possible to train and infer on sensitive data fully at the edge, preventing their risky propagation throughout the network, where they are susceptible to attacks. Moreover, edge intelligence can derive non-sensitive information that could then be submitted to the cloud without further processing;

·        The need for continuous internet connection: in locations where connectivity is poor or intermittent, the ML/DL could still be carried out;

·        Bandwidth degradation: edge computing can perform part of processing tasks on raw data and transmit the produced data to the cloud (filtered/aggregated/pre-processed), thus saving network bandwidth. Transmitting large amounts of data to the cloud burdens the network and impacts the overall Quality of Service (QoS) Liu et al. (2018).

·        Power waste Carbery et al. (2018): unnecessary raw data being transmitted through the internet demands power, decreasing energy efficiency on a large scale. The steps for data processing in ML vary according to the specific technique in use, but generally occur in a well-defined life cycle, which can be represented by a workflow.

Model building is at the heart of any ML technique, but the complete life cycle of a learning process involves a series of steps, from data acquisition and preparation to model deployment into a production environment. When adopting the Edge intelligence paradigm, it is necessary to carefully analyze which steps in the ML life cycle can be successfully executed at the edge of the network. Typical steps that have been investigated for execution at the edge are data collection, pre-processing, training and inference. Considering the aforementioned steps in ML and the specific features of edge nodes, we can identify many challenges to be addressed in the edge intelligence paradigm, such as (i) running ML/DL on devices with limited resources, (ii) ensuring energy efficiency without compromising the inference accuracy; (iii) communication efficiency; (iv) ensuring data privacy and security in all steps; (v) handling failure in edge devices; and (vi) dealing with heterogeneity and low quality of data. In this paper, we present the results of a systematic literature review on current state-of-the-art techniques and strategies developed for distributed machine learning in edge computing. We applied a methodological process to compile a series of papers and discuss how they propose to deal with one or more of the aforementioned challenges.

To align with Industry 4.0, traditional industrial automation approaches are evolving with the integration of new elements. The Internet of Things (IoT) and Cyber-Physical Systems (CPS) play crucial roles in enabling artificial intelligence and facilitating intelligent manufacturing, leading to the creation of innovative products and services Kunst et al. (2019). Companies embracing this approach face increased competition in a dynamic market environment, where simply increasing production capacity does not guarantee success Tessoni and Amoretti (2022). Despite various interpretations of industrial challenges, incorporating domain knowledge into understandable and explainable models remains challenging Vollert et al. (2021). DSS, leveraging machine learning algorithms, aids in product design iterations and facilitates effective policymaking, potentially enabling quicker recovery from failures O’Donovan et al. (2019),Romeo et al. (2020). Furthermore, leveraging vast amounts of data, particularly in predicting machine failures and scheduling maintenance, allows industries to enhance performance and autonomously manage product requirements Boyes et al. (2018). However, many manufacturing organizations struggle to embrace data-driven strategies due to various challenges, particularly during the data preprocessing stage Arena et al. (2022). As data generated within Industry 4.0 proliferate, machine learning algorithms play a vital role in extracting insights for improved understanding Kaushik and Yadav (2023). Nowadays ML can not only be used to diagnose problems, but can also be used to diagnose, prognosticate, and forecast problems Adhikari et al. (2018),Zhou and Tham (2018). In many instances, machines exhibit signs of deterioration and symptoms before they fail. Predictive maintenance (PdM) is a strategy used by engineers to anticipate failures before they occur, relying on sensor-based condition monitoring of machinery and equipment. However, implementing PdM requires substantial data and real-time monitoring, posing challenges such as latency, adaptability, and network bandwidth Liu et al. (2018). Implementing predictive maintenance at various stages of design offers several benefits but also presents challenges. Advantages include increased productivity, reduced system faults Carbery et al. (2018), decreased unplanned downtime, and improved resource efficiency Wang and Wang (2017). Predictive maintenance also enhances maintenance intervention planning optimization Balogh et al. (2018). However, managing data from multiple systems and sources within a facility is challenging, as is obtaining accurate data for predictive modelling Bousdekis et al. (2019),Ferreira et al. (2017). Additionally, implementing machine learning models and artificial intelligence faces challenges such as collecting training data Xu et al. (2018),

 

LITERATURE REVIEW

Maintenance can be broken down into two primary categories reactive maintenance and proactive maintenance. These are the types of maintenance most performed in Industries. After a valued item has had a breakdown, the purpose of reactive maintenance is to return it to working order Ding and Kamaruddin (2015). Through the utilization of preventative and predictive maintenance procedures, the purpose of proactive maintenance is to forestall the occurrence of expensive repairs and the early breakdown of assets. Corrective maintenance does not incur any upfront costs and does not need any prior preparation to be carried out Susto and Beghi (2016). In most cases, machines will experience some level of deterioration before finally breaking down. It is possible to monitor the trend of degradation to rectify any flaws that may exist before they cause any failure or the equipment to break down. Since machine maintenance is only carried out when it is necessary, this tactic results in greater levels of efficiency Susto and Beghi (2016). One such strategy that assists us in predicting failures before they take place is known as predictive maintenance (PdM). For this strategy to work, the asset in question needs to undergo condition monitoring, which employs sensor technologies to look for warning signs of deteriorating performance or impending breakdown. The PdM allows for the decision-making process to be examined from two different vantage points: the diagnosis and the prognosis. According to Jeong et al. Jeong et al. (2007),diagnosis is the process of determining the underlying cause of a problem, whereas prognosis is the process of estimating the likelihood that a failure will occur in the future Lewis and Edwards (1997). Prognosis maintenance policy is further divided into statistical-based maintenance and condition-based maintenance. Industry 4.0 equipped digital model gives maintenance staff the ability to schedule repairs more effectively because it provides real-time equipment information. It is evident that predictive maintenance is gaining more attention due to the recent advancements in data collection through Industry 4.0technologies and data analysis capabilities using evolutionary algorithms, cloud technology, data analytics, machine learning and artificial intelligence. According to U car et al. Ucar et al. (2024) AI is the main component of PdM for next-step autonomy in machines, which can improve the autonomy and adaptability of machines in complex and dynamic working environments. It is comprehensible that predictive maintenance is getting additional consideration because of recent advancements in data accessibility and analytics capabilities brought on by expanding research into ML and AI algorithms. Machine learning is frequently used by researchers to anticipate failure and improve output. Hesser and Markert Hesser and Markert (2019) used an Artificial Neural Network (ANN) model embedded within a CNC-milling machine to monitor tool wear. Models like this one can be applied to older machinery that can be used in Industry 4.0, as well as for research purposes. Kamariotis et al. (2024)found testing and validating AI-based PdM systems face a challenge due to the absence of standard evaluation metrics. This complicates the comparison of system performance and assessment of accuracy and reliability. Various evaluation metrics, including prediction accuracy, mean squared error, and precision and recall, have been suggested by researchers to address this issue. Sampaio et al. (2019) created an ANN model based on vibration measurements for the training dataset. Additionally, they compare the outcomes of ANN to those of Random Forest and Support Vector Machine (SVM), two other ML techniques, and discover that ANN is superior. Binding et al. [29] created Logistic Regression, XG Boost, and Random Forest models to assess the machine’s operational status. By way of choice criteria, Random Forest and XG Boost execute far improved as compared to Logistic Regression, while all algorithms perform better than one another in terms of Receiver Operating Characteristic (ROC). It was developed by Falamarzi et al. [30] to track forecast data and measure variation.SVM models predict curved segment gauge deviation better than ANN models do for straight segment gauge deviation. To recognize various rotary equipment scenarios, Biswal and Sabareesh [31] developed a Deep Neural Network (hereafter referred to as DNN) and Convolutional Neural Network (hereafter referred to as CNN). It can be used to monitor bearings in production lines and enhance the monitoring of online conditions in coastal turbines for wind energy. Data mining canbe used to predict system behaviour based on historical data. Amodel-based approach that heavily relies on analytical models to illustrate how the system operates has a few benefits. In fields with an abundance of data, like industrial maintenance, machine learning maybe used [32]. Actual results, answers based on cloud-based, and new algorithms are all becoming more popular. Quiroz et al. [33] applied the Random Forest technique for fault identification and validity and reliability analysis with turned 98 %diagnostic accuracy rate. Yan and Zhou [34] use TF- IDF (Term Frequency- Inverse Document Frequency) and RF (Radio Frequency) data from aircraft speed and torque sensors to create ML models for defect prediction. After extracting the features with TF-IDF from the unprocessed data after the earlier flights, Random Forest had an accurate optimistic degree of 66.67 percent and a percentage of false positives of0.13 percent. To predict wear and failures, Lee et al. [35] observed the spindle motor and cutting machine using data-driven machine learning modelling. It has been demonstrated that models using SVM and neural networks with artificial intelligence can forecast system health and longevity with high accuracy. According to Palangi et al. [36], recurrent neural network (RNN) and long short-term memory networks (LSTM) algorithms perform well with data that is sequential, time-series data with dependencies that last along time, and information from IoT flow sensors. LSTM and Naïve Bayes models combined, according to the study, may effectively identify trends and produce forecasts. The Naive Bayes anomaly detector was created by the LSTM model. Learning through deep learning with Cox proportional hazard (CoxPHDL) was developed in a research study to address the common problems of data flexibility and filtering that occur when functional maintenance information is analyzed [37]. The main goal was to develop an integrated solution based on dependability analysis and deep learning. In Gensler et al. [38] IoT application that combines solar panel prediction, the Deep Belief Network (DBN).

The idea of security in IoT devices has been recently articulated in studies that analyze the security needs at several layers of architecture, such as the application, cloud, network, data, and physical layers. Layers have examined potential vulnerabilities and attacks against IoT devices, classified IoT attacks, and explained layer-based security requirements8. On the other hand, industrial IoT (IIoT) networks are vulnerable to cyber attacks. Developing IDS is important to secure IIoT networks. The authors presented three DL models, LSTM, CNN, and a hybrid, to identify IIoT network breaches9. The researchers used the UNSW-NB15 and X-IIoTID datasets to identify normal and abnormal data, then compared them to other research using multi-class, and binary classification. The hybrid LSTM + CNN model has the greatest intrusion detection accuracy in both datasets. The researchers also assessed the implemented models’ accuracy in detecting attack types in the datasets9. In Ref.10, the authors introduced the hybrid synchronous-asynchronous privacy-preserving federated technique. The federated paradigm eliminates FL-enabled NG-IoT setup issues and protects all its pieces with Two- Trapdoor Homomorphic Encryption. The server protocol blocks irregular users. The asynchronous hybrid

LEGATO algorithm reduces user dropout. By sharing data, they assist data-poor consumers. In the presented model, security analysis ensures federated correctness, auditing, and PP. Their performance evaluation showed higher functionality, accuracy, and reduced system overheads than peer efforts. For medical devices, the authors of Ref.11 developed an auditable privacy-preserving federated learning (AP2FL) method. By utilizing Trusted Execution Environments (TEEs), AP2FL reduces issues about data leakage during training and aggregation activities on both servers and clients. The authors of this study aggregated user updates and found data similarities for non-IID data using Active Personalized Federated Learning (ActPerFL) and Batch Normalization (BN). In Ref.12, the authors addressed two major consumer IoT threat detection issues. First, the authors addressed FL’s unfixed issue: stringent client validation. They solved this using quantum-centric registration and authentication, ensuring strict client validation in FL. FL client model weight protection is the second problem. They suggested adding additive homomorphic encryption to their model to protect FL participants’ privacy without sacrificing computational speed. This technique obtained an average accuracy of 94.93% on the N-baIoT dataset and 91.93% on the Edge- IIoT set dataset, demonstrating consistent and resilient performance across varied client settings. Utilizing a semi-deep learning approach, Steel Eye was created in Ref.13 to precisely detect and assign responsibility for cyber attacks that occur at the application layer in industrial control systems. The proposed model uses category boosting and a diverse range of variables to provide precise cyber-attack detection and attack attribution. Steel Eye demonstrated superior performance in terms of accuracy, precision, recall, and Fl-score compared to state-of-the-art cyber-attack detection and attribution systems. In Ref.14, researchers developed a fuzzy DL model, an enhanced adaptive neuro-fuzzy inference system (ANFIS), fuzzy matching (FM), and a fuzzy control system to detect network risks. Our fuzzy DL finds robust nonlinear aggregation using the fuzzy Choquet integral. Metaheuristics optimized ANFIS attack detection’s error function. FM verifies transactions to detect block chain fraud and boost efficiency. The first safe, intelligent fuzzy block chain architecture, which evaluates IoT security threats and uncertainties, enables block chain layer decision- making and transaction approval. Tests show that the block chain layer’s throughput and latency can reveal threats to block chain and IoT. Recall, accuracy, precision, and F1-score are important for the intelligent fuzzy layer. In block chain-based IoT networks, the FCS model for threat detection was also shown to be reliable. In Ref.15, the study examined Federated Learning (FL) privacy measurement to determine its efficacy in securing sensitive data during AI and ML model training. While FL promises to safeguard privacy during model training, its proper implementation is crucial. Evaluation of FL privacy measurement metrics and methodologies can identify gaps in existing systems and suggest novel privacy enhancement strategies. Thus, FL needs full research on “privacy measurement and metrics” to thrive. The survey critically assessed FL privacy measurement found research gaps, and suggested further study. The research also included a case study that assessed privacy methods in an FL situation. The research concluded with a plan to improve FL privacy via quantum computing and trusted execution environments.

 

IoT, security, and ML IoT attacks and security vulnerabilities

Critical obstacles standing in the way of future attempts to see IoT fully accepted in society are security flaws and vulnerabilities. Everyday IoT operations are successfully managed by security concerns. In contrast, they have a centralized structure that results in several vulnerable points that may be attacked. For example, unpatched vulnerabilities in IoT devices are a security concern due to outdated software and manual updates. Weak authentication in IoT devices is a significant issue due to easy-to-identify passwords. Attackers commonly target vulnerable Application Programming Interfaces (APIs) in IoT devices using code injections, a man-in-the-middle (MiTM), and Distributed Denial-of-Service (DDoS)16. Unpatched IoT devices pose risks to users, including data theft and physical harm. IoT devices store sensitive data, making them vulnerable to theft. In the medical field, weak security in devices such as heart monitors and pacemakers can impede medical treatment. Figure 1 illustrates the types of IoT attacks (threats)17. Unsecured IoT devices can be taken over and used in botnets, leading to cyber attacks such as DDoS, spam, and phishing. The Mirai software in 2016 encouraged criminals to develop extensive botnets for IoT devices, leading to unprecedented attacks. Malware can easily exploit weak security safeguards in IoT devices18. Because there are so many connected devices, it may be difficult to ensure IoT device security. Users must follow fundamental security practices, such as changing default passwords and prohibiting unauthorized remote access19. Manufacturers and vendors must invest in securing IoT tool managers by proactively notifying users about outdated software, enforcing strong password management, disabling remote access for unnecessary functions, establishing strict API access control, and protecting command-and-control (C&C) servers from attacks.

 

IoT applications’ support security issues

Security is a major requirement for almost all IoT applications. IoT applications are expanding quickly and have impacted current industries. Even though operators supported some applications with the current technologies of networks, others required greater security support from the IoT-based technologies they use20. The IoT has several uses, including home automation and smart buildings and cities. Security measures can enhance home security, but unauthorized users may damage the owner’s property. Smart applications can threaten people’s privacy, even if they are meant to raise their standard of living. Governments are encouraging the creation of intelligent cities, but the safety of citizens’ personal information may be at risk21,22. Retail extensively uses the IoT to improve warehouse restocking and create smart shopping applications. Augmented reality applications enable offline retailers to try online shopping. However, security issues have plagued IoT apps implemented by retail businesses, leading to financial losses for both clients and companies.

 

PROPOSED METHODOLOGY

This study employs a quantitative, experimental, and comparative research design to evaluate the efficiency of Machine Learning (ML) and Deep Learning (DL) models in predicting machine failures within an Industry 4.0 environment. The design emphasizes supervised learning techniques, model benchmarking, hyperparameter optimization, and rigorous performance validation using real-world industrial data.

 

Two major analytical components structure the research: ML-based classification models for machine failure prediction and LSTM-based deep learning models for time-series behaviour analysis. Both components are tested on a highly imbalanced predictive maintenance dataset sourced from the UCI Repository. The experimental design allows systematic comparison of models under identical conditions, ensuring fairness in evaluating accuracy, detection capability, and robustness. This approach supports developing reliable predictive maintenance solutions capable of offering early warnings and improving industrial decision-making.

 

Data Source and Description

The study utilizes the Predictive Maintenance Dataset from the UCI Machine Learning Repository, which contains comprehensive machine-level operational data, including thermal readings, rotational speed, torque, voltage levels, tool wear, and system load conditions. The dataset reflects real industrial behaviour, combining normal operating records with rare failure occurrences, resulting in a significantly imbalanced class distribution. It includes several categories of machine failures, such as tool wear, heat-related issues, overstrain, power failures, and random breakdowns. Key variables consist of sensor-generated numerical features and operational parameters, along with binary and multi-class failure labels. This rich, multi-dimensional dataset provides an ideal testing ground for evaluating predictive maintenance algorithms, enabling the study to assess how well ML and DL models capture early deterioration signals and accurately classify impending failures.

 

Data Preprocessing

A structured data preprocessing workflow was implemented to ensure reliability and analytical accuracy. The dataset was first examined for missing values and noise, though no missing entries were found due to the dataset’s curated nature. Minor inconsistencies were addressed through outlier detection and stability checks on sensor readings. Since machine failures occur infrequently, the dataset suffered from severe class imbalance, which was corrected using the SMOTE oversampling technique to synthetically generate minority failure samples. Feature engineering steps included scaling and normalizing numerical attributes, encoding categorical variables, and applying a stratified train–test split to preserve class proportions. These preprocessing techniques enhanced data quality, improved model learning, and ensured that both ML and DL models were trained on balanced, representative, and properly formatted inputs.

 

Machine Learning Models Used

The study implemented and compared seven machine learning algorithms to identify the most effective model for predicting machine failures. These included Random Forest, Logistic Regression, Support Vector Machine, Naïve Bayes, Decision Tree, Gradient Boosting methods, and the XGBoost Classifier, which emerged as the primary performer. Each algorithm was trained using standardized procedures and evaluated using identical metrics to ensure fairness. To enhance model performance, Random Search CV was employed for hyperparameter tuning due to its efficiency and superior exploration of parameter space compared to Grid Search.

1)     Random Forest (RF)

Random Forest is an ensemble learning method that builds multiple decision trees and aggregates their predictions to improve accuracy and reduce overfitting. Each tree is trained on a randomly sampled subset of data and features, making the model robust against noise and variance. RF is highly effective for classification tasks, handles nonlinear relationships well, and performs strongly in predictive maintenance scenarios where sensor data contain complex patterns.

2)     Logistic Regression (LR)

Logistic Regression is a simple yet powerful statistical classification model that estimates the probability of a binary outcome using a logistic function. It is widely used due to its interpretability and low computational cost. LR works best with linearly separable data and provides insights into feature importance. Although less effective for complex nonlinear patterns, it serves as a strong baseline for machine failure prediction tasks.

3)     Support Vector Machine (SVM)

Support Vector Machine is a supervised learning model that identifies an optimal hyperplane separating classes with maximum margin. SVM performs exceptionally well in high-dimensional spaces and can model nonlinear relationships through kernel functions. Its robustness to overfitting makes it suitable for predictive maintenance datasets. Although computationally expensive for large datasets, SVM delivers high accuracy when distinguishing between normal and failure states in machine data.

4)     Naïve Bayes (NB)

Naïve Bayes is a probabilistic classifier based on Bayes’ theorem, assuming independence among features. It is efficient, fast, and performs well even with small datasets or noisy data. NB is particularly useful in early-stage prediction tasks due to its low training time. Although the independence assumption may not always hold, it remains effective for machine failure classification where quick approximations of failure likelihood are required.

5)     Decision Tree (DT)

Decision Tree is a rule-based classification model that splits data into branches using feature thresholds, forming an interpretable tree structure. It models nonlinear patterns and interactions without requiring complex data preprocessing. DT is easy to visualize and understand, making it useful for industrial applications. However, it is prone to overfitting when used alone, which is why it often serves as a base model for ensemble techniques.

6)     Gradient Boosting Methods

Gradient Boosting builds an ensemble of weak learners—typically decision trees—in sequential stages, where each new model corrects the errors of the previous ones. It optimizes predictive performance by minimizing loss functions using gradient descent. This approach captures complex relationships and delivers highly accurate results. Gradient boosting is well-suited for predictive maintenance tasks that require sensitivity to subtle failure patterns in sensor readings.

7)     XGBoost Classifier (Primary ML Model)

XGBoost (Extreme Gradient Boosting) is an advanced and optimized boosting algorithm designed for high speed, scalability, and superior performance. It incorporates regularization, parallel processing, and optimized tree-building techniques, making it highly effective for structured industrial datasets. XGBoost handles missing values, nonlinear relationships, and class imbalance efficiently. Its exceptional accuracy, reliability, and fast execution make it the leading model for machine failure prediction in this study.

 

Deep Learning Framework

The study incorporates a deep learning framework to capture temporal behavior and sequential dependencies inherent in machine operating data. The primary model used is Long Short-Term Memory (LSTM), a recurrent neural network architecture specifically designed to learn long-term patterns from time-series sensor readings. LSTM effectively identifies gradual deterioration trends, cyclical fluctuations, and hidden temporal relationships that traditional ML models may overlook, making it highly suitable for modeling machine failure progression. To benchmark its performance, the study also employs a baseline Artificial Neural Network (ANN) built on a standard multi-layer perceptron structure. While ANN can learn complex feature interactions, it does not retain sequential memory, resulting in lower accuracy compared to LSTM. However, the study notes that ANN performance can improve with deeper architectures and optimized hyperparameters. Together, these deep learning models provide a comprehensive evaluation of temporal predictive capabilities in industrial failure prediction scenarios.

 

Model Training and Validation

Model training and validation were carried out using a structured approach to ensure accuracy, fairness, and reliability in the performance comparison of all machine learning models. After the dataset underwent preprocessing and balancing through the SMOTE technique, it was divided into two subsets: 70% for training and 30% for testing. This split was applied after oversampling to prevent data leakage and maintain the integrity of model evaluation. The training set was used to fit each model and optimize parameters, while the testing set served as an independent dataset to assess generalization performance. This ensured that the models learned from a balanced dataset but were evaluated on realistic, unseen distributions representative of actual industrial conditions.

To measure predictive performance, multiple evaluation metrics were applied, capturing both statistical accuracy and failure detection capability. Accuracy provided an overall measure of correct predictions, while Precision, Recall, and F1 Score offered deeper insights into the model’s sensitivity to rare failure events. ROC-AUC Score evaluated the models’ ability to distinguish between failure and non-failure classes across thresholds, and the Confusion Matrix revealed specific misclassification patterns. Together, these metrics ensured a comprehensive assessment of each algorithm’s effectiveness in predictive maintenance scenarios.

 

RESULTS AND DISCUSSION

The primary goal of the classification model is to predict whether the machine will break down within the allotted time. It is possible to forecast the residual usable life of the machine by utilizing the regression model; nevertheless, the value of the prediction shifts with the deterioration of the asset. They can go down significantly or up significantly. However, if many assets need to be tracked, it will not be possible to do so on an individual basis for each asset. As a result, a classification model has been developed that provides certain early warnings with precision and within a predetermined time frame. The performance metrics of several machine learning models are displayed in Table 1. After obtaining the results for the various ML models, the hyper parameters of these models will be modified to attain a greater degree of precision and accuracy. As the tuning of hyper parameters was discussed in the previous section and it is known that the "Random Search CV" method is superior to the other commonly employed parameter optimization techniques, we can move on to the next step. The effects of hyper parameter tuning and optimization on the performance metrics of multiple machine learning models are displayed in Table 2. Based on the data presented in Table 2, it can be concluded that the XG Boost classifier is the most effective of all the algorithms used in the modeling process. It is primarily because it has the highest AUC score.

Table 1

Table 1 Evaluation Metrics of a Classification Model.
Model	Accuracy	Precision	Recall	F1-Score	AUC-Score
Logistic Regression	0.6453	0.4419	0.9625	0.6057	0.5331
KNN	0.7168	0.6709	0.6831	0.6769	0.7007
SVC	0.7293	0.7809	0.8027	0.7916	0.7138
Decision Tree (Gini)	0.6998	0.7713	0.7668	0.7690	0.7237
Decision Tree (Entropy)	0.6998	0.7713	0.7668	0.7690	0.7237
Bagging Classifier	0.8239	0.8859	0.5799	0.7010	0.7842
Adaptive Boosting	0.7786	0.8551	0.8657	0.8604	0.7814
Adaptive Boosting	0.7671	0.8520	0.8557	0.8538	0.7817
Random Forest Classifier	0.7964	0.9054	0.9234	0.9143	0.8202
XG Boost Classifier	0.8480	0.9460	0.9466	0.9463	0.8544

 

Table 2

Table 2 Evaluation Metrics of Classification Model After Hyper Parameters Tuning Optimization
Model	Accuracy	Precision	Recall	F1-Score	AUC-Score
Logistic Regression	0.6692	0.4693	0.9976	0.6383	0.5532
KNN	0.7407	0.6983	0.7182	0.7081	0.7208
SVC	0.7532	0.8083	0.8378	0.8228	0.7339
Decision Tree (Gini)	0.7237	0.7987	0.8019	0.8003	0.7438
Decision Tree (Entropy)	0.7237	0.7987	0.8019	0.8003	0.7438
Bagging Classifier	0.8478	0.9133	0.6150	0.7353	0.8043
Adaptive Boosting	0.8025	0.8825	0.9008	0.8916	0.8015
Gradient Boosting	0.7910	0.8794	0.8908	0.8850	0.8018
Random Forest	0.8203	0.9328	0.9585	0.9455	0.8403
XG Boost Classifier	0.8719	0.9734	0.9817	0.9776	0.8745

 

And the highest accuracy among all the models. The ROC curve for an XG Boost model illustrates its ability to discriminate between positive and negative classes across various threshold values, serving as a pivotal tool for evaluating its classification performance. The ROC-curve corresponding to the XG Boost model is depicted in Fig. 7, and the confusion matrix for the XG Boost model applied to the validation dataset is shown in Table 3.

 

CONCLUSION AND FUTURE WORK

Predictive maintenance appears as an essential method for accomplishing the goals of Industrial Revolution 4.0 in the worldwide industrial sector. The classification models created in this study greatly improve maintenance planning by providing early indications of asset breakdown. Critical tracking of multiple assets at the same time is now possible using these models. However, implementing predictive maintenance presents obstacles, such as addressing dataset imbalance and optimizing hyperparameters for greater performance. The growing applications of machine learning in all domains of science and technology attempts better prediction through applying wellknown models and comparing their performances. Though the models applied in this study are well known machine learning and deep learning models, the focus of this work was on trying to explore the models that ensure accurate prediction with an imbalance dataset. The researchers can apply the same set of algorithms for predicting machine failure in any type of industrial setting. Additionally, this study has developed a data-driven model for predicting machine failure and compared results from different machine learning algorithms. Practitioners and policymakers can take note that these algorithms require availability of large amount of error-free data to predict failure accurately. Many industries do not invest in data collection through IOT and sensor devices and face frequent breakdowns and interruption in their production lines. Analytics of collected manufacturing and operations data can help in saving huge amounts of time and money of manufacturing and service industries. Additionally, while a relevant problem, this paper does not focus on exploring why some algorithms and techniques are individually better than others. Still, the techniques considered represent the most used approaches, and the algorithms used for the meta-learner also provide a decent insight into the performance that can be expected. The SMOTE approach effectively resolved the imbalance in the dataset, allowing for the comparison of different machine-learning methods. The results show that the XG Boost Classifier outperforms other ML algorithms in terms of performance measures, even after hyperparameter adjustment. This emphasizes the significance of balancing datasets and fine- tuning model parameters to ensure accurate predictions. Furthermore, the use of deep learning models, notably Long Short-Term Memory (LSTM), outperforms typical machine learning approaches. Although the accuracy of Artificial Neural Networks (ANN) initially fell short, layer-based optimization has the potential to improve. Despite its efficiency, predictive maintenance remains very expensive due to the use of advanced monitoring technology. Future efforts should focus on creating low-cost sensor technologies to reduce monitoring costs. Furthermore, future work will focus on optimizing hyperparameter tweaking in machine learning models. Furthermore, investigating a range of deep learning algorithms other than LSTM could improve the modelling process and provide more reliable forecasting capabilities. In conclusion, while predictive maintenance provides significant benefits for maintenance planning and efficiency, overcoming problems such as dataset imbalance and cost considerations is critical. Continued R&D efforts will help to advance and widely adopt predictive maintenance solutions in the industry. Predicting machine failures through predictive models, one of the goals of Industry 4.0, provides timely warnings of potential equipment failures and helps organizations to schedule maintenance activities. The results of this study demonstrate the effectiveness of machine learning and deep learning algorithms in predictive maintenance, enabling proactive maintenance interventions and resource optimization. It also contributes to the growing body of research on machine learning applications in industrial settings, advancing theoretical understanding and paving the way for further refinement of predictive maintenance methodologies.

  

ACKNOWLEDGMENTS

None.

 

REFERENCES

 Adhikari, P., Rao, H. G., and Buderath, M. (2018). Machine Learning-Based Data Driven Diagnostics and Prognostics Framework for Aircraft Predictive Maintenance. In Proceedings of the 10th International Symposium on NDT in Aerospace (24–26). Dresden, Germany.  

Arena, S., Florian, E., Zennaro, I., Orrù, P. F., and Sgarbossa, F. (2022). A Novel Decision Support System for Managing Predictive Maintenance Strategies Based on Machine Learning Approaches. Safety Science, 146, 43–54. https://doi.org/10.1016/j.ssci.2021.105529   

Balogh, Z., Gatial, E., Barbosa, J., Leitão, P., and Matejka, T. (2018). Reference Architecture for a Collaborative Predictive Platform for Smart Maintenance In Manufacturing. In 22nd International Conference On Intelligent Engineering Systems (Ines) (Pp. 299–304). Ieee. Https://Doi.Org/10.1109/Ines.2018.8523969   

Bousdekis, A., Mentzas, G., Hribernik, K., Lewandowski, M., von Stietencron, M., and Thoben, K.-D. (2019). A unified architecture for proactive maintenance in manufacturing enterprises. In Enterprise Interoperability VIII (307–317). Springer. https://doi.org/10.1007/978-3-030-13693-2_26   

Boyes, H., Hallaq, B., Cunningham, J., and Watson, T. (2018). The Industrial Internet of Things (IIoT): An Analysis Framework. Computers in Industry, 101, 1–12. https://doi.org/10.1016/j.compind.2018.04.015   

Carbery, C. M., Woods, R., and Marshall, A. H. (2018). A Bayesian Network-Based Learning System For Modelling Faults In Large-Scale Manufacturing. In Ieee International Conference On Industrial Technology (Icit) (1357–1362). Ieee. Https://Doi.Org/10.1109/Icit.2018.8352377  

Da Cunha Mattos, T., Santoro, F. M., Revoredo, K., and Nunes, V. T. (2014). A Formal Representation for Context-Aware Business Processes. Computers In Industry, 65(8), 1193–1214. Https://Doi.Org/10.1016/J.Compind.2014.07.005

Ding, S. H., and Kamaruddin, S. (2015). Maintenance Policy Optimization—Literature Review and Directions. International Journal of Advanced Manufacturing Technology, 76, 1263–1283. https://doi.org/10.1007/s00170-014-6341-2   

Ferreira, L. L., Albano, M., Silva, J., Martinho, D., Marreiros, G., Di Orio, G., and Ferreira, H. (2017). A Pilot for Proactive Maintenance in Industry 4.0. In 2017 IEEE 13th International Workshop on Factory Communication Systems (WFCS) (1–9). IEEE. https://doi.org/10.1109/WFCS.2017.7991952   

Hesser, D. F., and Markert, B. (2019).

Jeong, I. J., Leon, V. J., and Villalobos, J. R. (2007). Integrated Decision-Support System for Diagnosis, Maintenance Planning, and Scheduling of Manufacturing Systems. International Journal of Production Research, 45(2), 267–285. https://doi.org/10.1080/00207540600678896   

Kaushik, A., and Yadav, D. K. (2023). Analysing Failure Prediction for a Manufacturing Firm Using Machine Learning Algorithms. In Advanced Engineering Optimization Through Intelligent Techniques: Select Proceedings of AEOTIT 2022 (pp. 457–463). Springer Nature Singapore. https://doi.org/10.1007/978-981-19-9285-8_44   

Kunst, R., Avila, L., Binotto, A., Pignaton, E., Bampi, S., and Rochol, J. (2019). Improving Devices Communication in Industry 4.0 Wireless Networks. Engineering Applications of Artificial Intelligence, 83, 1–12.  

Lewis, S. A., and Edwards, T. G. (1997). Smart Sensors and System Health Management Tools for Avionics and Mechanical Systems. In 16th DASC. AIAA/IEEE Digital Avionics Systems Conference. Reflections to the Future. Proceedings 2 (8-5). IEEE. https://doi.org/10.1109/DASC.1997.637283   

Liu, Z., Jin, C., Jin, W., Lee, J., Zhang, Z., Peng, C., and Xu, G. (2018). Industrial AI Enabled Prognostics for High-Speed Railway Systems. In 2018 IEEE International Conference on Prognostics and Health Management (ICPHM) (1–8). IEEE. https://doi.org/10.1109/ICPHM.2018.8448431   

O’Donovan, P., Gallagher, C., Leahy, K., and O’Sullivan, D. T. (2019). A Comparison of Fog and Cloud Computing Cyber-Physical Interfaces for Industry 4.0 Real-Time Embedded Machine Learning Engineering Applications. Computers in Industry, 110, 12–35. https://doi.org/10.1016/j.compind.2019.04.016   

Romeo, L., Loncarski, J., Paolanti, M., Bocchini, G., Mancini, A., and Frontoni, E. (2020). Machine Learning-Based Design Support System for the Prediction of Heterogeneous Machine Parameters in Industry 4.0. Expert Systems with Applications, 140, 112869. https://doi.org/10.1016/j.eswa.2019.112869   

Schmidt, B., and Wang, L. (2018). Cloud-Enhanced Predictive Maintenance. International Journal of Advanced Manufacturing Technology, 99, 5–13. https://doi.org/10.1007/s00170-016-8983-8   

Schmidt, B., and Wang, L. (2018). Predictive Maintenance of Machine Tool Linear Axes: A Case from Manufacturing Industry. Procedia Manufacturing, 17, 118–125. https://doi.org/10.1016/j.promfg.2018.10.022   

Susto, G. A., and Beghi, A. (2016). Dealing with Time-Series Data in Predictive Maintenance Problems. In 2016 IEEE 21st International Conference on Emerging Technologies and Factory Automation (ETFA) (1–4). IEEE. https://doi.org/10.1109/ETFA.2016.7733659   

Tessoni, V., and Amoretti, M. (2022). Advanced Statistical and Machine Learning Methods for Multi-Step Multivariate Time Series Forecasting in Predictive Maintenance. Procedia Computer Science, 200, 748–757. https://doi.org/10.1016/j.procs.2022.01.273   

Ucar, A., Karakose, M., and Kırımça, N. (2024). Artificial Intelligence for Predictive Maintenance Applications: Key Components, Trustworthiness, and Future Trends. Applied Sciences, 14(2), 898. https://doi.org/10.3390/app14020898   

Vollert, S., Atzmueller, M., and Theissler, A. (2021). Interpretable Machine Learning: A Brief Survey from the Predictive Maintenance Perspective. In 2021 26th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA) (1–8). IEEE. https://doi.org/10.1109/ETFA45728.2021.9613467   

Wang, K., and Wang, Y. (2017). How AI Affects the Future Predictive Maintenance: A Primer of Deep Learning. In International Workshop of Advanced Manufacturing and Automation 32 (1–9). Springer. https://doi.org/10.1007/978-981-10-5768-7_1   

Xu, Y., Sun, Y., Liu, X., and Zheng, Y. (2018). A Digital-Twin-Assisted Fault Diagnosis Using Deep Transfer Learning. IEEE Access, 7, 990–999. https://doi.org/10.1109/ACCESS.2018.2890566   

Zhou, C., and Tham, C. K. (2018). Graphel: A Graph-Based Ensemble Learning Method for Distributed Diagnostics and Prognostics in the Industrial Internet of Things. In 2018 IEEE 24th International Conference on Parallel and Distributed Systems (ICPADS) (903–909). IEEE. https://doi.org/10.1109/PADSW.2018.8644943