Fault Diagnosis Of Reciprocating Compressor Research Articles

Deep Fusion of Intrinsic Vibration Information and Grassmann Manifold-based Similarity for Fault Identification of Reciprocating Compressor

This paper introduces a new method based on deep belief networks (DBN) to integrate intrinsic vibration information and assess the similarity of subspaces established on the Grassmann manifold for intelligent fault diagnosis of a reciprocating compressor (RC). Initially, raw vibration signals undergo empirical mode decomposition (EMD) to break them down into multiple intrinsic mode functions (IMFs). This operation can reveal inherent vibration patterns of fault and other components hidden in the original signals. Subsequently, features are refined from all the IMFs and concatenated into a high-dimensional representative vector, offering localized and comprehensive insights into RC operation. Through DBN, the fault-sensitive information are further refined from the features to enhance their performance in fault identification. Finally, similarities among subspaces on the Grassmann manifold are computed to match fault types. The efficacy of the method is validated using field data. Comparative analysis with traditional approaches for feature dimension reduction, feature extraction, and Euclidean distance-based fault identification underscores the effectiveness and superiority of the proposed method in RC fault diagnosis. Conflict of Interest Statement The authors declare no conflicts of interest.

Feasibility exploration of strain-based indicator diagram reconstruction for reciprocating compressor fault diagnosis

Reciprocating compressor is an indispensable and important production equipment in petrochemical industry. However, due to the large number of vulnerable components, thermal performance failures occur frequently. Indicator diagram has significant advantages in the diagnosis of those thermal performance failures and is therefore regarded as a reliable fault diagnosis technique for reciprocating compressor. Nevertheless, the indicator diagram needs to extract the pressure and piston position information in the cylinder, which has greatly hindered its practical application. The traditional way of extracting the pressure information may damage the cylinder wall, and the obtaining of the piston position requires key-phase sensor, both of which pose potential safety hazards. In this paper, a novel idea of obtaining cylinder pressure and piston position at the same time through cylinder strain is developed, and the feasibility of the proposed idea is verified by numerical simulation. During the simulation, this work employs ANSYS tool to extract the strain from the outer wall of the cylinder as a way of revealing the relationship between pressure and strain, and piston position and strain. Subsequently, the strain-based indicator diagram is reconstructed through the change of the strain at a certain point on the outer wall of the cylinder in a working cycle. The simulation results show that the cylinder strain can reflect cylinder internal pressure and piston position obviously, and it is feasible to construct an indicator diagram based on cylinder strain.

A new method for reciprocating compressor fault diagnosis based on indicator diagram feature extraction

The damage of vulnerable components inside the cylinder of reciprocating compressor, including the valve, piston ring, packing and piston ring, will cause the unexpected shutdown of the compressor unit. The indicator diagram which reflects the thermodynamic process in the cylinder is suitable for fault diagnosis of vulnerable components. However, most of the published fault diagnosis methods based on indicator diagram are aimed at the fault diagnosis of gas valve. In addition, the extracted features lack physical meaning in most fault diagnosis methods using machine learning algorithm, which is not conducive to be widely applied in practical engineering. In this study, features with definite physical meaning, including average suction pressure, average discharge pressure, area of indicator diagram and centroid coordinates of indicator diagram, are extracted from indicator diagram, and the threshold database of features under normal states and various fault states is established according to the contrast experiment. The results of the experiment show that the thresholds of the extracted parameters are obviously different under normal states and various fault states. During fault diagnosis, several groups of indicator diagrams of the compressor to be diagnosed are collected at first. After feature extraction, the extracted features are compared with the thresholds under different compressor states to obtain the average numbers of features within the threshold range under different compressor states to determine the compressor states. The accuracy of the method for judging whether the compressor is faulty or normal could reach 98.3%. Furthermore, the accuracy of identifying individual faulty components and multiple faulty components could reach 86.86%. The reason for the low overall diagnostic accuracy is that certain faults have similar effects on the features extracted from indicator diagram. The proposed method is believed as an excellent fault diagnosis method for the vulnerable components inside the cylinder of reciprocating compressor.

Applications of Machine Learning to Reciprocating Compressor Fault Diagnosis: A Review

Operating condition detection and fault diagnosis are very important for reliable operation of reciprocating compressors. Machine learning is one of the most powerful tools in this field. However, there are very few comprehensive reviews which summarize the current research of machine learning in monitoring reciprocating compressor operating condition and fault diagnosis. In this paper, the recent application of machine learning techniques in reciprocating compressor fault diagnosis is reviewed. The advantages and challenges in the detection process, based on three main monitoring parameters in practical applications, are discussed. Future research direction and development are proposed.

Intelligent Fault Diagnosis of a Reciprocating Compressor Using Mode Isolation Convolutional Deep Belief Networks

Due to the complex transfer paths of vibration signals, and a large number of vibration excitations, fault diagnosis of reciprocating compressors (RCs) has become one of the most challenging problems in the field of health monitoring. Focusing on fault diagnosis, a novel method, which will be referred to in this paper as mode isolation convolutional deep belief network (MI-CDBN), is proposed from the perspective of transfer path analysis, and multimodal data isolation. First, sparse filtering is applied to compress vibration signals and to reduce the computing cost. Second, the MI-CDBN is used to isolate multimodal data of different transfer paths and to calculate features using unsupervised learning. Finally, a multiclass logistic regression is employed to identify the fault types of the RC. Vibration signals from practical industries are used to validate the proposed method. The obtained results demonstrate that the proposed method has an improved performance compared to many other state-of-the-art methods widely used in the fault diagnosis of RCs.

Reciprocating compressor fault diagnosis using an optimized convolutional deep belief network

This article proposes an optimized convolutional deep belief network for fault diagnosis of reciprocating compressors. Sparse filtering is first used to compress raw signal into compact time series by refining the most representative information and to reduce the computational burden. Then, the proposed convolutional deep belief network is adopted to learn the unsupervised features of the compressed signal without the need of feature extraction by human effort. To improve the generalization ability of the network, an optimized probabilistic pooling out is proposed in this article to replace the standard one in the pooling layer of the convolutional deep belief network. Finally, the unsupervised features calculated by the optimized convolutional deep belief network are fed as the input of the softmax regression classifier for fault identification. Four types of vibration signals reflecting different operating conditions are collected from the industry to validate the effectiveness of the proposed method. The obtained results demonstrate that the proposed convolutional deep belief network method can achieve a higher classification accuracy rate of up to 91% for fault diagnosis than the traditional methods and accomplish the fault diagnosis of reciprocating compressor effectively.

Reciprocating Compressor Fault Diagnosis Technology Based on Multi-source Information Fusion

Reciprocating Compressor Fault Diagnosis Technology Based on Multi-source Information Fusion