# Extruder Machine Gear Fault Detection Using Autoencoder LSTM via Sensor Fusion Approach

^{*}

## Abstract

**:**

## 1. Introduction

- A DWT for enhanced vibration signal analysis in plastic extruder gearbox fault diagnosis: By incorporating a DWT strategy, the aim is to extract invaluable insights from the vibration signals entrenched in noise. This technique seeks to bolster the efficacy of diagnosing faults within the plastic extruder gearboxes.
- An effective statistical time-frequency domain feature extraction and correlation filter-based selection technique: A method for feature extraction is presented in our investigation, which is particularly effective in the time-frequency domain. Furthermore, a feature selection process based on correlation filters, a technique commonly utilized in feature engineering, is incorporated. This process aims to enhance significant and crucial characteristics, thereby improving the overall performance of the model.
- A multi-sensor fusion using the FastICA technique: Our strategy includes a multi-sensor fusion paradigm aided by the (FastICA) technique. The proposed technique harmoniously blends selected information from multiple separate sensor datasets. This fusion not only condenses data to a single-dimensional array but also preserves the unique characteristics of each source.
- An AE-LSTM outlier detection using a fused multi-sensor dataset approach: We achieved an outlier detection by leveraging an AE-LSTM, which is enabled by a fusion of multi-sensor data techniques. This comprehensive methodology results in a strong framework ready for defect detection in the context of a plastic extruder gearbox.
- A proposed framework validation and proposed global evaluation metrics: A set of global evaluation indicators are provided to validate our suggested approach. These evaluations highlight the framework’s efficiency and efficacy, demonstrating its ability to manage the complexities of defect detection within plastic extruder gears.

## 2. Materials and Methods

#### 2.1. DWT for Denoising/Decomposition Overview

#### 2.2. FastICA for Dimension Reduction

#### 2.3. Correlation Coefficients

#### 2.4. Autoencoder

#### 2.5. Long Short-Term Memory (LSTM)

#### 2.6. The Proposed Outlier Detection Model

- Data acquisition: Both vibration and thermal data were collected in order to construct an appropriate model for monitoring extruder gear performance. The incorporation of several data sources is prompted by the fact that vibration signals obtained from gearboxes are prone to noise contamination, making it difficult to extract valuable insights on their own. A more comprehensive and useful picture can be built by adding additional data, such as temperature measurements. Vibration data are critical for detecting anomalies or inconsistencies in the operation of the gear. However, because of the existence of noise, it is frequently impossible to distinguish important patterns or trends purely from vibration signals. This is when the extra thermal data come into play. By combining vibration and thermal data, it is possible to identify hidden links and correlations between the performance of the gear and the accompanying temperature fluctuations. The use of both vibration and thermal data seeks to improve the accuracy and usability of the model built to monitor the extruder gear. This method allows for a more comprehensive study, allowing for the detection of potential problems such as high friction, overheating, or abnormal operating circumstances. Finally, by combining multiple data sources, a more robust and efficient model may be constructed, providing useful insights for optimizing extruder gear performance, maintenance, and dependability.
- Signal processing and feature extraction: The second key aspect of the model revolves around signal processing, with the aim of extracting valuable information from gearbox vibration data while minimizing the inherent noise. The DWT was used as a method for deconstructing, filtering, and pre-processing the vibration signals to achieve this. The DWT extracted time-frequency statistical information from both the original signal and each vibration signal decomposition level. A full analysis of the vibration data was performed by performing decomposition at various levels, collecting variances across different scales and frequencies. Thermal data, on the other hand, as a time-varying signal, did not go through decomposition. Instead, from the raw temperature data, time statistical features were extracted. The goal of this method was to capture the temporal patterns and trends revealed by temperature readings. The study aims to improve the quality and usability of the information gained by applying the DWT to vibration signal processing and extracting time statistical features from temperature data. This methodology allowed us to identify key trends, correlations, and anomalies in the vibration and temperature data, allowing us to gain a more comprehensive understanding of the extruder gear’s behavior and performance.
- Feature selection: To obtain an effective diagnosis in the setting of anomaly detection, discriminant traits are required. A correlation filter technique was used to guarantee that the features extracted had enough discriminative power. This technique ensures that only features with a correlation percentage of 70% or above are deemed closely connected. By removing characteristics that do not match this correlation threshold, the resulting feature set is tailored to include informative and discriminating features, improving the accuracy and effectiveness of the diagnosis process.
- Signal fusion: The integration of data from numerous sources while keeping their different characteristics is a critical step in our suggested model’s signal data fusion. FastICA was used as the signal-processing method in our study for this reason. FastICA aided us in the merging of data from several sources, allowing us to mix and extract important information while preserving the distinctive qualities of each data source. We accomplished effective signal integration using FastICA, allowing for a thorough analysis that captures the synergistic effects and correlations across the various data sources in our investigation.
- Diagnosis/outlier detection: The entire model’s procedures are built with the goal of detecting faults, specifically through outlier detection. The model’s structure is deliberately constructed to accomplish this aim. As for the AI tool of choice in our investigation, an AE-LSTM was employed. Details concerning the implementation and operation of the AE-LSTM have been discussed earlier in this section. The overarching goal is to use this AI tool to discover issues by finding anomalies in data, allowing for prompt diagnosis and intervention.

#### 2.6.1. Model Hyper-Parameter Function

#### 2.6.2. Model Global Performance Evaluation Metrics Overview

#### 2.7. Data Collection and Pre-Processing

#### 2.7.1. Feature Extraction

#### 2.7.2. Feature Selection and Sensor Fusion

## 3. Results and Discussion

#### 3.1. Proposed System Training and Validation

#### 3.2. Model’s Outlier Detection Evaluation

#### 3.3. Model’s Evaluation Metrics Validation

#### Model Validation with LSTM

## 4. Discussion

## 5. Conclusions

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Conflicts of Interest

## Abbreviations

RUL | Remaining Useful Life |

FN | False Negative |

TN | True Negative |

TP | True Positive |

FDI | False Detection and Isolation |

IoT | Internet of Things |

PHM | Prognostics and Health Management |

LSTM | Long Short-Term Memory |

ANN | Artificial Neural Network |

ML | Machine Learning |

DL | Deep Learning |

FNNN | Feed-forward neural networks |

DNN | Deep neural networks |

CNN | Convolutional neural networks |

DBN | Deep belief networks |

DWT | Discrete Wavelet Transform |

LLE | Local Linear Embedding |

PCA | Principal Component Analysis |

ICA | Independent Component Analysis |

AE | Autoencoder |

AI | Artificial Intelligence |

FastICA | Fast Independent Component Analysis |

MSE | Mean Square Error |

CBM | Condition Based Monitoring |

## References

- Kumar, S.; Tiwari, P.; Zymbler, M. Internet of Things is a revolutionary approach for future technology enhancement: A review. J. Big Data
**2019**, 6, 111. [Google Scholar] [CrossRef] - Zhou, I.; Makhdoom, I.; Shariati, N.; Raza, M.A.; Keshavarz, R.; Lipman, J.; Abolhasan, M.; Jamalipour, A. Internet of Things 2.0: Concepts, Applications, and Future Directions. IEEE Access
**2021**, 9, 70961–71012. [Google Scholar] [CrossRef] - Do, J.S.; Kareem, A.B.; Hur, J.-W. LSTM-Autoencoder for Vibration Anomaly Detection in Vertical Carousel Storage and Retrieval System (VCSRS). Sensors
**2023**, 23, 1009. [Google Scholar] [CrossRef] - Wu, G.; Yan, T.; Yang, G.; Chai, H.; Cao, C. A Review on Rolling Bearing Fault Signal Detection Methods Based on Different Sensors. Sensors
**2022**, 22, 8330. [Google Scholar] [CrossRef] - Singh, V.; Mathur, J.; Bhatia, A. A comprehensive review: Fault detection, diagnostics, prognostics, and fault modeling in HVAC systems. Int. J. Refrig.
**2022**, 144, 283–295. [Google Scholar] [CrossRef] - Ghobakhloo, M.; Iranmanesh, M.; Tseng, M.-L.; Grybauskas, A.; Stefanini, A.; Amran, A. Behind the definition of Industry 5.0: A systematic review of technologies, principles, components, and values. J. Ind. Prod. Eng.
**2023**, 40, 432–447. [Google Scholar] [CrossRef] - Okwuosa, C.N.; Akpudo, U.E.; Hur, J.-W. A Cost-Efficient MCSA-Based Fault Diagnostic Framework for SCIM at Low-Load Conditions. Algorithms
**2022**, 15, 212. [Google Scholar] [CrossRef] - Okwuosa, C.N.; Hur, J.-W. A Filter-Based Feature-Engineering-Assisted SVC Fault Classification for SCIM at Minor-Load Conditions. Energies
**2022**, 15, 7597. [Google Scholar] [CrossRef] - Gundewar, S.K.; Kane, P.V. Condition Monitoring and Fault Diagnosis of Induction Motor. J. Vib. Eng. Technol.
**2021**, 9, 643–674. [Google Scholar] [CrossRef] - Wang, H.; Zhou, C.; Hu, B.; Liu, Z. Tooth wear prediction of crowned helical gears in point contact. Proc. Inst. Mech. Eng. Part J J. Eng. Tribol.
**2020**, 6, 947–963. [Google Scholar] [CrossRef] - Zhou, J.; Sun, W.; Wang, Z. Vibration and noise characteristics of a gear reducer under different operation conditions. J. Low Freq. Noise Vib. Act. Control
**2019**, 2, 574–591. [Google Scholar] [CrossRef] - Amarnath, M.; Krishna, I.R.P. Local fault detection in helical gears via vibration and acoustic signals using EMD based statistical parameter analysis. Measurement
**2014**, 58, 154–164. [Google Scholar] [CrossRef] - Karabacak, Y.E.; Özmen, N.G.; Gümüşel, L. Intelligent worm gearbox fault diagnosis under various working conditions using vibration, sound and thermal features. Appl. Acoust.
**2022**, 58, 108463. [Google Scholar] [CrossRef] - Tang, X.; Xu, Y.; Sun, X.; Liu, Y.; Jia, Y.; Gu, F.; Ball, A.D. Intelligent fault diagnosis of helical gearboxes with compressive sensing based non-contact measurements. Appl. ISA Trans.
**2023**, 133, 559–574. [Google Scholar] [CrossRef] - Roda-Casanova, V.; Gonzalez-Perez, I. Investigation of the effect of contact pattern design on the mechanical and thermal behaviors of plastic-steel helical gear drives. Mech. Mach. Theory
**2021**, 164, 104401. [Google Scholar] [CrossRef] - Homayouni, H.; Ghosh, S.; Ray, I.; Gondalia, S.; Duggan, J.; Kahn, M.G. An Autocorrelation-based LSTM-Autoencoder for Anomaly Detection on Time-Series Data. In Proceedings of the 2020 IEEE International Conference on Big Data (Big Data), Atlanta, GA, USA, 10–13 December 2020; p. 5077. [Google Scholar] [CrossRef]
- Mallak, A.; Fathi, M. Sensor and Component Fault Detection and Diagnosis for Hydraulic Machinery Integrating LSTM Autoencoder Detector and Diagnostic Classifiers. Sensors
**2021**, 21, 433. [Google Scholar] [CrossRef] [PubMed] - Anwani, N.; Rajendran, B. Training Multi-layer Spiking Neural Networks using NormAD based Spatio-Temporal Error Backpropagation. Neurocomputing
**2020**, 380, 67–77. [Google Scholar] [CrossRef] - Kholkin, V.; Druzhina, O.; Vatnik, V.; Kulagin, M.; Karimov, T.; Butusov, D. Comparing Reservoir Artificial and Spiking Neural Networks in Machine Fault Detection Tasks. Big Data Cogn. Comput.
**2023**, 7, 110. [Google Scholar] [CrossRef] - Morando, S.; Pera, M.C.; Yousfi Steiner, N.; Jemei, S.; Hissel, D.; Larger, L. Reservoir Computing Optimisation for PEM Fuel Cell Fault Diagnostic. In Proceedings of the 2017 IEEE Vehicle Power and Propulsion Conference (VPPC), Belfort, France, 11–14 December 2017; pp. 1–7. [Google Scholar] [CrossRef]
- Kishore, K.; Sharma, A.; Mukhopadhyay, G. Failure Analysis of a Gearbox of a Conveyor Belt. J. Fail. Anal. Preven
**2020**, 20, 1237–1243. [Google Scholar] [CrossRef] - Chen, Y.; Li, J.; Zang, L.; Liu, Y.; Bi, W.; Yang, X. Dynamic Simulation and Experimental Identification for Fatigue Pitting Helical Gear Fault. J. Mech. Eng.
**2021**, 57, 61–70. [Google Scholar] [CrossRef] - Nejad, A.R.; Gao, Z.; Moan, T. Fatigue Reliability-based Inspection and Maintenance Planning of Gearbox Components in Wind Turbine Drivetrains. Energy Procedia
**2014**, 53, 248–257. [Google Scholar] [CrossRef] - Asi, O. Fatigue failure of a helical gear in a gearbox. Eng. Fail. Anal.
**2006**, 13, 1116–1125. [Google Scholar] [CrossRef] - Kale, A.S. Bending Fatigue Failure in Gear Tooth. IJERT
**2013**, 2, 2278-0181. [Google Scholar] - Zhang, S.; Zhou, J.; Wang, E.; Zhang, H.; Gu, M.; Pirttikangas, S. State of the art on vibration signal processing towards data-driven gear fault diagnosis. Inst. Eng. Technol.
**2022**, 4, 249–266. [Google Scholar] [CrossRef] - Poletto, J.C.; Fernandes, C.M.C.G.; Barros, L.Y.; Neis, P.D.; Pondicherry, K.; Fauconnier, D.; Seabra, J.H.O.; De Baets, P.; Ferreira, N.F. Identification of gear wear damage using topography analysis. Wear
**2023**, 522, 204837. [Google Scholar] [CrossRef] - Vanraj; Dhami, S.S.; Pabla, B.S. Gear fault classification using Vibration and Acoustic Sensor Fusion: A Case Study. In Proceedings of the 2018 Condition Monitoring and Diagnosis (CMD), Perth, WA, Australia, 23–26 September 2018; pp. 1–6. [Google Scholar] [CrossRef]
- Jaen-Cuellar, A.Y.; Trejo-Hernández, M.; Osornio-Rios, R.A.; Antonino-Daviu, J.A. Gear Wear Detection Based on Statistic Features and Heuristic Scheme by Using Data Fusion of Current and Vibration Signals. Energies
**2023**, 16, 948. [Google Scholar] [CrossRef] - Zhang, Y.; Baxter, K. Deep Transfer Multi-sensor Fusion for Gearbox Diagnostics. Int. Res. J. Mod. Eng. Technol. Sci.
**2020**, 2, 15. [Google Scholar] - Kumar, T.P.; Saimurugan, M.; Haran, R.B.H.; Siddharth, S.; Ramachandran, K.I. A multi-sensor information fusion for fault diagnosis of gearbox utilizing discrete wavelet features. Meas. Sci. Technol.
**2019**, 30, 085101. [Google Scholar] [CrossRef] - Lee, M.-S.; Shifat, T.A.; Hur, J.-W. Kalman Filter Assisted Deep Feature Learning for RUL Prediction of Hydraulic Gear Pump. IEEE Sens. J.
**2022**, 22, 11088–11097. [Google Scholar] [CrossRef] - Yuan, W. Study on Noise Elimination of Mechanical Vibration Signal Based on Improved Wavelet. In Proceedings of the 12th International Conference on Measuring Technology and Mechatronics Automation (ICMTMA), Phuket, Thailand, 28–29 February 2020; pp. 141–143. [Google Scholar] [CrossRef]
- Chen, X.; Yang, Y.; Cui, Z.; Shen, J. Wavelet Denoising for the Vibration Signals of Wind Turbines Based on Variational Mode Decomposition and Multiscale Permutation Entropy. IEEE Access
**2020**, 8, 40347–40356. [Google Scholar] [CrossRef] - Chen, G.; Xie, W.; Zhao, Y. Wavelet-based denoising: A brief review. In Proceedings of the 2013 Fourth International Conference on Intelligent Control and Information Processing (ICICIP), Beijing, China, 9–11 June 2013; pp. 570–574. [Google Scholar] [CrossRef]
- Chatterjee, S.; Thakur, R.S.; Yadav, R.N.; Gupta, L.; Raghuvanshi, D.K. Review of noise removal techniques in ECG signals. IET Signal Process
**2020**, 9, 569–590. [Google Scholar] [CrossRef] - Sun, C.; He, Z.; Cao, H.; Zhang, Z.; Chen, X.; Jian, M.Z. A non probabilistic metric derived from condition information for operational reliability assessment of aero-engines. IEEE Trans. Rel.
**2015**, 64, 167–181. [Google Scholar] [CrossRef] - Roemer, M.J.; Kacprzynski, G.J.; Schoeller, M.H. Improved diagnostic and prognostic assessments using health management information fusion. In Proceedings of the 2001 IEEE Autotestcon Proceedings. IEEE Systems Readiness Technology Conference. (Cat. No.01CH37237), Valley Forge, PA, USA, 20–23 August 2001; pp. 365–377. [Google Scholar] [CrossRef]
- Duan, L.; Zhao, F.; Wang, J.; Wang, N.; Zhang, J. An integrated cumulative transformation and feature fusion approach for bearing degradation prognostics. Shock Vib.
**2018**, 2018, 9067184. [Google Scholar] [CrossRef] - Bagheriye, L.; Ali, G.; Kerkhoff, H.G. Life-time prognostics of dependable VLSI-SoCs using machine-learning. In Proceedings of the 2020 IEEE 26th International Symposium on On-Line Testing and Robust System Design (IOLTS), Napoli, Italy, 13–15 July 2020; pp. 1–4. [Google Scholar] [CrossRef]
- Lee, T.-W.; Girolami, M.; Sejnowski, T.J. Independent component analysis using an extended infomax algorithm for mixed subgaussian and supergaussian sources. Neural Comput.
**1999**, 11, 417–441. [Google Scholar] [PubMed] - Weidong, J. Fault diagnosis of gearbox by FastICA and residual mutual information based feature extraction. Proc. Int. Conf. Inf. Autom.
**2009**, 13, 928–932. [Google Scholar] [CrossRef] - Akpudo, U.E.; Hur, J.-W. An Automated Sensor Fusion Approach for the RUL Prediction of Electromagnetic Pumps. IEEE Sens. J.
**2021**, 9, 38920–38933. [Google Scholar] [CrossRef] - Hyvärinen, A.; Oja, E. Independent component analysis: Algorithms and applications. Neural Netw.
**2000**, 13, 411–430. [Google Scholar] [CrossRef] - Okwuosa, C.N.; Hur, J.-w. An Intelligent Hybrid Feature Selection Approach for SCIM Inter-Turn Fault Classification at Minor Load Conditions Using Supervised Learning. IEEE Access
**2023**, 11, 89907–89920. [Google Scholar] [CrossRef] - Kang, J.; Kim, C.-S.; Kang, J.W.; Gwak, J. Anomaly Detection of the Brake Operating Unit on Metro Vehicles Using a One-Class LSTM Autoencoder. Appl. Sci.
**2021**, 11, 9290. [Google Scholar] [CrossRef] - Nicholaus, I.T.; Park, J.R.; Jung, K.; Lee, J.S.; Kang, D.-K. Anomaly Detection of Water Level Using Deep Autoencoder. Sensors
**2021**, 21, 6679. [Google Scholar] [CrossRef] - Givnan, S.; Chalmers, C.; Fergus, P.; Ortega-Martorell, S.; Whalley, T. Anomaly Detection Using Autoencoder Reconstruction upon Industrial Motors. Sensors
**2022**, 22, 3166. [Google Scholar] [CrossRef] [PubMed] - Jin, J.; Chung, Y.; Park, J. Development of a Flowmeter Using Vibration Interaction between Gauge Plate and External Flow Analyzed by LSTM. Sensors
**2020**, 20, 5922. [Google Scholar] [CrossRef] [PubMed] - Lee, J.-H.; Hong, J.-K. Comparative Performance Analysis of Vibration Prediction Using RNN Techniques. Electronics
**2022**, 11, 3619. [Google Scholar] [CrossRef] - Huang, J.; Mo, J.; Zhang, J.; Ma, X. A Fiber Vibration Signal Recognition Method Based on CNN-CBAM-LSTM. Appl. Sci.
**2022**, 12, 8478. [Google Scholar] [CrossRef] - Gu, K.; Zhang, Y.; Liu, X.; Li, H.; Ren, M. DWT-LSTM-Based Fault Diagnosis of Rolling Bearings with Multi-Sensors. Electronics
**2021**, 10, 1009. [Google Scholar] [CrossRef] - Kwon, J.E.; Shifat, T.A.; Kareem, A.B.; Hur, J.-W. RUL Prediction of Switched Mode Power Supply Using a Kalman Filter Assisted Deep Neural Network. Processes
**2022**, 10, 55. [Google Scholar] [CrossRef] - Curreri, F.; Patanè, L.; Xibilia, M.G. RNN- and LSTM-Based Soft Sensors Transferability for an Industrial Process. Sensors
**2021**, 21, 823. [Google Scholar] [CrossRef] - Wang, X.; Huang, T.; Zhu, K.; Zhao, X. LSTM-Based Broad Learning System for Remaining Useful Life Prediction. Mathematics
**2022**, 10, 2066. [Google Scholar] [CrossRef] - Zaman, S.M.K.; Liang, X. An Effective Induction Motor Fault Diagnosis Approach Using Graph-Based Semi-Supervised Learning. IEEE Access
**2021**, 9, 7471–7482. [Google Scholar] [CrossRef] - Amruthnath, N.; Gupta, T. A research study on unsupervised machine learning algorithms for early fault detection in predictive maintenance. In Proceedings of the 2018 5th International Conference on Industrial Engineering and Applications (ICIEA), Singapore, 26–28 April 2018; pp. 355–361. [Google Scholar] [CrossRef]
- Jabbar, H.K.; Khan, R.Z. Methods to Avoid Over-Fitting and Under-Fitting in Supervised Machine Learning (Comparative Study). Comput. Sci. Commun. Instrum. Devices
**2014**. [Google Scholar] [CrossRef]

**Figure 4.**Pictorial View of two extruder gearbox: (

**a**) healthy plastic extruder gearbox overview, (

**b**) faulty plastic extruder gearbox overview.

**Figure 5.**Raw motor time-domain signal from both extruder gearboxes: (

**a**) vibration signals, (

**b**) thermal signals.

**Figure 6.**Discrete wavelet transform for: (

**a**) healthy plastic extruder gearbox decomposition, (

**b**) faulty plastic extruder gearbox decomposition.

**Figure 9.**Correlation plot for: (

**a**) filter-based correlation selected temperature features, (

**b**) filter-based correlation selected vibration features.

**Figure 11.**Visualization plot for the set threshold with reconstruction error for the proposed model’s outlier detection.

Model Architecture | Description |
---|---|

Number of Classes | 2 |

Number of Layers | 7 |

Batch Size | 32 |

Number of Epochs | 100 |

Dropout Rate | 0.001 |

Optimizer | Adam |

Activation Function | Relu |

Loss Function | MSE |

Validation Split | 0.2 |

Domain | Feature Name | Definition |
---|---|---|

Time-Domain | Median | ${\left(\right)}^{\frac{n+1}{2}}$ sample |

Mean | $\overline{x}=\frac{1}{n}\left(\right)open="("\; close=")">{\sum}_{i=1}^{n}{x}_{i}$ | |

Root Mean Square | ${X}_{rms}=\sqrt{\frac{{\sum}_{i=1}^{n}{\left(\right)}^{{x}_{i}}2}{}n}$ | |

Standard Deviation | $\sigma =\sqrt{\frac{\sum {\left(\right)}^{{x}_{i}}2}{}n}$ | |

Variance | ${S}^{2}=\frac{\sum {\left(\right)}^{{x}_{i}}2}{}n$ | |

Standard Error | $SE=\frac{\sigma}{\sqrt{n}}$ | |

Kurtosis | ${X}_{kurt}=\frac{1}{N}\Sigma \left(\right)open="("\; close=")">\frac{{\left(\right)}^{{x}_{i}}}{3}\sigma $ | |

Skewness | ${X}_{\mathrm{skew}}=E\left[\left(\right),\frac{{\left(\right)}^{{x}_{i}}}{3},\sigma \right]$ | |

Max | ${X}_{max}=max\left(\right)open="("\; close=")">{x}_{i}$ | |

Min | ${X}_{max}=min\left(\right)open="("\; close=")">{x}_{i}$ | |

Crest Factor | ${X}_{CF}=\frac{{x}_{max}}{{x}_{rms}}$ | |

Peak-to-peak | ${X}_{p-p}={x}_{max}-{x}_{min}$ | |

Peak factor | ${x}_{PF}=\frac{{x}_{max}}{\sqrt{{x}_{s}}}$ | |

Wave Factor | ${x}_{WF}=\frac{\sqrt{\frac{1}{n}{\sum}_{i=1}^{n}{\left(\right)}^{{x}_{i}}2}}{}\frac{1}{n}{\sum}_{i=1}^{n}\left(\right)open="|"\; close="|">{x}_{i}$ | |

Clearance factor | ${x}_{CF}=\frac{{x}_{max}}{mean\left|2x\right|}$ | |

Impulse factor | ${X}_{IF}=\frac{{x}_{max}}{\frac{1}{N}{\sum}_{i=1}^{N}\left(\right)open="|"\; close="|">{x}_{i}}$ | |

Frequency-Domain | Maximum Frequency | ${X}_{MF}={x}_{max}(\frac{1}{N}{\sum}_{i=1}^{N}{\left(\right)}^{{x}_{i}}2)$ |

Variance Frequency | ${X}_{VF}={x}_{var}(\frac{1}{N}{\sum}_{i=1}^{N}{\left(\right)}^{{x}_{i}}2)$ | |

Spectral Kurtosis | $SK=\frac{2{\sum}_{k=0}^{{\overline{B}}_{L}/2-1}\left(\right|X(k,n)|-{\mu \left|X\right|)}^{4}}{{B}_{L}\xb7{\sigma}_{\left|X\right|}^{4}}-3$ | |

Spectral Skewness | $SS=\frac{2{\sum}_{k=0}^{{\overline{B}}_{L}/2-1}\left(\right|X(k,n)|-{\mu \left|X\right|)}^{4}}{{B}_{L}\xb7{\sigma}_{\left|X\right|}^{3}}$ |

Definition | Accuracy (%) | Precision (%) | Recall (%) | F1-Score (%) |
---|---|---|---|---|

AE-LSTM | 94.42 | 100 | 93.67 | 91.72 |

Model Architecture | Description |
---|---|

Number of Classes | 2 |

Number of Layers | 4 |

Batch Size | 64 |

Number of Epochs | 20 |

Optimizer | Adam |

Activation Function | Sigmoid |

Loss Function | MSE |

Validation Split | 0.2 |

Definition | Accuracy (%) | Precision (%) | Recall (%) | F1-Score (%) |
---|---|---|---|---|

LSTM | 89.67 | 98 | 88.01 | 86.31 |

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |

© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Lee, J.-H.; Okwuosa, C.N.; Hur, J.-W.
Extruder Machine Gear Fault Detection Using Autoencoder LSTM via Sensor Fusion Approach. *Inventions* **2023**, *8*, 140.
https://doi.org/10.3390/inventions8060140

**AMA Style**

Lee J-H, Okwuosa CN, Hur J-W.
Extruder Machine Gear Fault Detection Using Autoencoder LSTM via Sensor Fusion Approach. *Inventions*. 2023; 8(6):140.
https://doi.org/10.3390/inventions8060140

**Chicago/Turabian Style**

Lee, Joon-Hyuk, Chibuzo Nwabufo Okwuosa, and Jang-Wook Hur.
2023. "Extruder Machine Gear Fault Detection Using Autoencoder LSTM via Sensor Fusion Approach" *Inventions* 8, no. 6: 140.
https://doi.org/10.3390/inventions8060140