Fault diagnosis in harvey machinery using machine learning techniques

In industries, fault diagnosis of different rotating equipment plays an important role as it ensures protection, reliability and avoids failure and loss of any energy supply. Early detection is a crucial factor in diagnosing defects, reducing both time and expense and preventing hazardous conditions

Project Title

Fault diagnosis in harvey machinery using machine learning techniques

Project Area of Specialization

Artificial Intelligence

Project Summary

In industries, fault diagnosis of different rotating equipment plays an important role as it ensures protection, reliability and avoids failure and loss of any energy supply. Early detection is a crucial factor in diagnosing defects, reducing both time and expense and preventing hazardous conditions. Intelligent fault diagnostic experiments with machine learning techniques are being performed. In addition, in the form of fault detection, a thorough analysis of various machine learning techniques that are implemented into different rotating equipment is often carried out. Finally, to spread the quality of fault diagnosis, numerous algorithms are suggested and the possible research ideas of implementing machine learning methods are simplified on separate rotating machines. Machinery fault diagnosis has advanced over the past decades with the advancement of machineries in terms of size. To guarantee their designed functions and efficiency over their lifespan, high-value machinery requires condition monitoring and fault diagnosis. Machinery Research In recent years, fault diagnostics have evolved rapidly. A short debate on possible developments and problems of machinery fault diagnosis continues with a study of the underlying principle of machinery fault diagnosis and its realistic applications in engineering. In the field of process monitoring, machinery fault detection is becoming increasingly necessary due to higher standards for mechanical systems that have higher efficiency, protection, and reliability. Science and technology developments have led to mechanical structures being developed, such as those used in wind turbines, airplanes, high-speed trains and machine tools. Some of these tasks include the monitoring of system working conditions, the detection of when devices or parts have an abnormal condition or defect, the estimation of the original cause of abnormal conditions or faults, the evaluation of its occurrence level, and the estimation of the remaining useful life or patterns of abnormal conditions. One of the primary methods for continuous maintenance is the detection of machinery faults, which can help deter irregular case development, minimize offline duration, predict residual life, and reduce loss of productivity. In exchange, this will help prevent large system breakdowns and catastrophes. Due to complicated and extreme circumstances, such as heavy load, high temperature, and high speed, the main components of mechanical machinery will eventually produce multiple faults of varying degrees. Machinery fault detection techniques include regularly sampled measurements from a variety of sensors to observe a mechanical device over a period of time.

Project Objectives

In machinery classification, rotating machinery is a common class. The root cause of the faults in rotating equipment is also a defective rolling part bearing. Over time and continuous use of machines, the rotors suffer abrasions and fatigue which compromises their performance. They are losing their best over time. Difficult work is to diagnose faults and repair the rotary machines. By analyzing and giving a solution before a fault occurs, we can minimize the cost, time and lifetime of machines. Although it is a very difficult task. Some algorithms are used to diagnose failures in rotary machines using “Machine Learning Techniques”.
The main aim of this research was to develop novel signal processing methods to enable automated diagnosis of rotating machinery. The work program comprised a synthesis of techniques from the fields of pattern recognition and SVM technique, with application to condition monitoring. Further detailed objectives of the research are.
Novel techniques in the field of signal processing were developed and applied in feature extraction for machinery faults diagnosis, which enabled features to be extracted effectively and clearly to reduce the dependency of fault diagnosis on well-trained technicians. In particular, wavelet packets based methods were invest?gated, improved, and applied from the perspective of
•    Attain?ng better time-frequency resolution, which assists in avoiding missing information and increasing information accuracy.
•    Wavelets that are suitable for vibration analysis. These wavelets were selected based on their resemblance to real characteristics of vibration signals of rotating machinery.
Fault diagnosis technique is developed to try to, reduce the dependency on human interpretation for the task of fault diagnosis. Broadly, the SVM algorithm is used to classify faults in bearings.
We are going to solve a real-time problem based on industrial need, which will cover two main real-time problems, monitor the health of machine and Fault Diagnosis and Predictive maintenance. This project is an Industrial idea and assumed to be a system based project as explained. That is why it must have an easy to use GUI. It is a business idea based on industry needs. The de-noising of vibration signal of faulty rolling bearing contributes to extracting the fault frequency of the faulty rolling bearing, to achieve the fault identification and diagnosis of rolling bearing.

Project Implementation Method

Project Implementation:
Using Machine Learning methods, a way to detect faults in rotary machines. For fault analysis of the rotational mismatch in the rotor, a support vector machine(SVM) algorithm is suggested. Help vector machines (SVMs) have recently been one of the most popular methods of classification in vibration analysis techniques. Using support vector machines, the Axis unbalance fault is classified. The experimental data is taken from a rigid-shaft rotor rotary machine platform and adjustable bearings, an experimental setup for the vibration analysis research. Several cases of imbalance defects have been successfully detected. Using Machine Learning methods, we apply a method of diagnosing defects in rotary machines. Using vibration study, the rotor unbalance fault in an induction motor has been established. For fault diagnosis of the rotational unbalance in the rotor, the SVM algorithm was suggested. Several cases of imbalance defects have been successfully detected. In the case of a minimal number of samples, the SVM algorithm has functional relevance for machine learning. An actual laboratory setup was taken from the vibration study and reports of performance were collected. The SVM, used for classification and regression analysis, is a machine learning algorithm supervised or not with related learning algorithms that interpret the data and identify patterns. In the case with a minimal number of samples, the SVM takes the optimal solution. The SVM algorithm converts Sample Space (SS) by nonlinear transformation into High Dimensional Feature Space (HDFS). In addition to performing a linear classification, SVMs will use what is called the kernel trick to effectively perform a non-linear classification, indirectly mapping their inputs into high-dimensional feature spaces.
Wavelet Packet is most commonly used for mother wavelet collection, and the high frequency transient signal band providing rich information on bearing defects cannot be easily separated. Using Wavelet packet transform (WPT) vibration signal characteristics, the Artificial Neural Network and Support Vector Machine (SVM) are commonly used to model bearing defects. The vibration database used in this study taken from online available bearing data center website of Case Western Reserve University (CWRU)
 

Benefits of the Project

Project Benefits:
Benefit of this project is that “Cost”, “Time”, “Machinery life”, “Danger of life” are reduced.  We use sensor for detection for the faulty Machines. Engineers and the employees must know before the faulty machine by using sensors. At present, the problems related to basic research on the diagnosis of machinery failure can be summarized as follows in "eight more and less":

• More to study fault action, less to breakdown mechanism;
• More to study revolving machinery, less to reciprocating machinery;
• More to study general machinery, less to advanced machinery,
• More to study single process, less to detailed diagnosis.

Therefore, in these five ways, breakthroughs related to the fundamental science of machinery fault diagnosis must be realized: breakthroughs from behavioral research to process analysis, from qualitative to quantitative research, from single to community fault research, from serious to poor fault research, and from part level to fault research at system level. In the analysis, only sparse information about the perception and diagnosis of mechanical defects is obtainable based on the principle of "what you see is what you get". The process of malfunction is the root cause of the flaw in nature's reflection. Further science study on failure mechanisms is, therefore, necessary. Due to the lack of previous samples, the conventional method of diagnosis can neglect the mechanical faults of new equipment. Many novel, large-scale, and high-speed mechanical equipment are being developed and widely deployed in practical fields, such as wind power equipment, commercial gas turbines, and the rapid advancement of science and technology. Railway locomotives, the transmission of aircraft control and the shield tunneling system. As regards the mechanical, electrical and hydraulic systems in these novel rotaries and reciprocating mechanical machines, it is also important to study and examine the fault mechanism and the evolutionary mechanisms under special operating conditions. For starters, we need to construct mathematical and mechanical models and experimental platforms for traditional misalignment faults, as well as research the signs of failure and features of the frequency continuum.
 

Technical Details of Final Deliverable

Technical details of Final Deliverable:
     Over the past decades, machinery fault diagnosis has advanced with the evolution of machinery in terms of sophistication and size. To guarantee their designed functions and efficiency over their lifespan, high-value machinery requires condition monitoring and fault diagnosis. Machinery Research In recent years, fault diagnostics have evolved rapidly. Fault mechanism, signal acquisition and sensor procedure, signal processing, and smart diagnostics. The main point is that “Cost”, “Time”, “Machinery life”, “Danger of life” are reduced.  We use sensor for detection for the faulty Machines.  We used SVM to implement this project. We have download dataset from different sources (websites), this dataset is in the notepad form. We converted into our required format .csv then we analyze using different techniques like making some charts and visual form of data, clean the data, feature extraction on dataset. After that we visualize the faulty data and using SVM algorithm and done comparative analysis to and training models to predict the life of machinery. Doing all of the analysis we make a product that describe the life, health, time duration of heavy machines. Furthermore, programming and establishing the standard database of fault diagnosis should be encouraged.
 

Final Deliverable of the Project

Software System

Core Industry

Education

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Quality Education, Industry, Innovation and Infrastructure

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com