A fault identification method for swash plate plunger pumps based on wavelet transform and convolutional neural network (CNN) involves the detection and classification of faults in pumps using signal processing techniques and machine learning algorithms. Here is an overview of the method: 1. Data Collection: Vibration or sound signals from the swash plate piston pump are collected under normal and fault operating conditions. These signals are used as input data for the fault identification process. 2. Wavelet transform: Use wavelet transform to analyze the collected signals. Applying wavelet transform to decompose the signal into different frequency components can reveal specific fault-related features. By choosing appropriate wavelet basis functions and decomposition levels, fault-related information can be extracted from the signal. 3. Feature extraction: Extract statistical features such as mean, standard deviation, energy, entropy, etc. from the wavelet transformed signal. These signatures capture the unique characteristics of the signals associated with different fault types in swash plate piston pumps. 4. Data preprocessing: Preprocessing the extracted features, normalizing the data, and removing noise or irrelevant information that may hinder the fault identification process. 90-R-100-KT-5-BC-80-P-3-S1-E-03-GBA-35-35-24 90R100KT5BC80P3S1E03GBA353524 90-R-100-KT-5-BC-80-P-3-S1-E-03-GBA-32-32-24 90R100KT5BC80P3S1E03GBA323224 90-R-100-KT-5-BC-80-L-4-S1-F-00-GBA-35-35-24 90R100KT5BC80L4S1F00GBA353524 90-R-100-KT-5-BC-60-S-3-T2-E-00-GBA-42-42-20 90R100KT5BC60S3T2E00GBA424220 90-R-100-KT-5-BB-80-P-3-F1-F-03-GBA-35-35-24 90R100KT5BB80P3F1F03GBA353524 90-R-100-KT-5-AB-80-R-3-C7-F-03-EBC-38-38-24 90R100KT5AB80R3C7F03EBC383824 90-R-100-KT-2-CD-60-P-3-F1-E-03-GBA-35-35-24 90R100KT2CD60P3F1E03GBA353524 90-R-100-KT-2-AC-60-R-4-S1-E-03-GBA-32-32-24 90R100KT2AC60R4S1E03GBA323224 90-R-100-KT-2-AB-60-P-3-S1-E-03-GBA-35-35-24 90R100KT2AB60P3S1E03GBA353524 90-R-100-KT-1-NN-80-R-4-S1-E-03-GBA-35-35-24 90R100KT1NN80R4S1E03GBA353524 5. Convolutional Neural Network (CNN): CNN is trained using preprocessed feature data. The CNN architecture consists of multiple convolutional layers for feature extraction and fully connected layers for classification. The CNN learns to recognize patterns and features in the input data that correspond to specific types of failures in a swash plate piston pump. 6. Training and Validation: The labeled training data, consisting of input feature vectors and corresponding fault labels, is used to train the CNN. The network is optimized by adjusting its weights and biases through a process called backpropagation. The validation data is used to evaluate the performance of the trained CNN and fine-tune its parameters when needed. 7. Fault identification: Once the CNN is trained and validated, it can be used to identify faults in the swash plate piston pump. The new vibration or acoustic signal from the pump is processed using wavelet transform and the extracted features are fed into the trained CNN. The network then classifies the type of fault based on the learned patterns and provides an output indicating the identified fault. 8. Performance evaluation: By comparing the results obtained by CNN with the actual fault situation, the accuracy and effectiveness of the fault identification method are evaluated. Performance metrics such as precision, recall and F1-score can be calculated to evaluate the performance of the method in detecting and classifying different faults. 9. Data augmentation: In order to enhance the performance and robustness of the CNN model, data augmentation techniques can be applied. Data augmentation involves artificially creating additional training samples by introducing changes to the original data, such as random shifts, rotations, or the addition of noise. This helps the model generalize better to different operating conditions and improves its ability to accurately identify faults. 10. Multiple types of faults: The fault identification method can be extended to deal with multiple types of faults of swash plate plunger pumps. By including labeled data of various fault types, a CNN model can be trained to classify and distinguish different fault scenarios, such as worn valves, leaking cylinders, or damaged piston rings. This allows comprehensive troubleshooting of the pump. 11. Real-time monitoring: The fault identification method can be realized in the real-time monitoring system, which continuously monitors the swash plate plunger pump. Vibration or acoustic signals can be acquired continuously from the pump and processed in real time using wavelet transform and CNN models. This enables early detection and timely intervention of potential failures, preventing serious damage and increasing maintenance efficiency. 12. Transfer learning: Transfer learning can be used to leverage pre-trained CNN models on related tasks or datasets. By fine-tuning the pre-trained CNN model with a smaller dataset specific to swash plate piston pump faults, the method can benefit from the generalization ability of the pre-trained model and require less training data for accurate fault identification. 90-R-100-KT-1-NN-80-L-4-S1-F-03-GBA-40-40-20 90R100KT1NN80L4S1F03GBA404020 90-R-100-KT-1-CD-80-S-4-F1-F-00-GBA-35-26-30 90R100KT1CD80S4F1F00GBA352630 90-R-100-KT-1-CD-80-R-4-T2-E-03-GBA-35-35-24 90R100KT1CD80R4T2E03GBA353524 90-R-100-KT-1-CD-80-R-3-S1-E-03-GBA-35-35-24 90R100KT1CD80R3S1E03GBA353524 90-R-100-KT-1-CD-80-R-3-C7-D-00-GBA-42-42-26 90R100KT1CD80R3C7D00GBA424226 90-R-100-KT-1-CD-80-P-3-C7-E-00-FAC-42-42-28 90R100KT1CD80P3C7E00FAC424228 90-R-100-KT-1-CD-80-L-3-S1-F-03-GBA-35-35-24 90R100KT1CD80L3S1F03GBA353524 90-R-100-KT-1-BC-80-P-3-S1-E-00-FAC-42-42-28 90R100KT1BC80P3S1E00FAC424228 90-R-100-KT-1-BC-80-P-3-F1-E-03-GBA-35-35-24 90R100KT1BC80P3F1E03GBA353524 90-R-100-KT-1-BC-80-L-4-S1-F-03-GBA-40-40-20 90R100KT1BC80L4S1F03GBA404020 13. Integration with condition monitoring systems: The fault identification method can be integrated into a wider condition monitoring system for comprehensive pump health assessment. The output of the fault identification model can be combined with other sensor data, such as temperature or pressure measurements, to provide a holistic view of the pump's condition. This integrated approach enables proactive maintenance planning and pump performance optimization. 14. Continuous Model Improvement: Fault identification methods can be further refined and improved through an iterative process. Feedback from field data and actual pump failures can be used to update and enhance the model's performance over time. Continuous learning and model improvement contribute to the accuracy and reliability of fault identification systems. By utilizing the capabilities of wavelet transform and CNN, the fault identification method provides an efficient and automated way to detect and classify faults in swash plate piston pumps. It offers the potential for early fault detection, reduced downtime, minimized maintenance costs and improved overall pump reliability.