Condition Monitoring of Three Phase Induction Motor using Artificial Neural Networks
As induction motor is a workhorse for any industrial or mechanical system, monitoring its condition continuously and early prediction of various kinds of faults that may hamper its performance becomes extremely important to pre-empt losses. The proposed research work aims to develop an Artific
2025-06-28 16:30:54 - Adil Khan
Condition Monitoring of Three Phase Induction Motor using Artificial Neural Networks
Project Area of Specialization Artificial IntelligenceProject SummaryAs induction motor is a workhorse for any industrial or mechanical system, monitoring its condition continuously and early prediction of various kinds of faults that may hamper its performance becomes extremely important to pre-empt losses. The proposed research work aims to develop an Artificial Intelligence based test algorithm that would be able to discern a faulty 3-phase induction motor using the bearing fault data. Condition monitoring is a technique or process of observing the operating characteristics of machine in such a way that any change in the trend of target parameters could be used to predict the need for maintenance before any serious worsening occurs. Condition monitoring implies the continuous evaluation of the health of equipment. Among the various strategies that can be followed to assess operating conditions of induction motors, condition monitoring based on Artificial Neural Networks (ANN) is the emergent one. The project proposes to develop a test rig that would compare a healthy 3-phase induction motor with a faulty one. The test rig would diagnose bearing fault through Motor Current Signature Analysis (MCSA) along with Back Propagation algorithm of ANN. The project requires a fast processing and computation for testing of ANN models. Also, a huge dataset would be used to train the model. All this need to be done on two 0.5 HP 3-phase squirrel cage induction motors by using Raspberry Pi 4, SCT-013-005 Split-Core AC Current Sensor and ADS-1115 Analog to Digital Converters (ADCs). First, data acquisition would be done through current sensors and then by using Python programming language on Raspberry Pi that data will train the ANN model. After training the model, the test rig would be able to diagnose a faulty bearing induction motor by the comparison of healthy induction motor.
The ultimate goal of the project is to design a test rig capable enough to distinguish a faulty motor among the healthy motors with high accuracy and minimum cost.
The system block diagram is proposed in figure:
The project aims to develop a test rig that would compare a healthy 3-phase squirrel cage induction motor with a faulty one with comparably low cost. The key project objectives are stated here:
- To obtain data acquisition, where current sensors are required to take measurements.
- To apply signal processing method such as Wavelet Transformation and Statistical analysis for features extraction.
- To extract features for Neural Network algorithm.
- To develop an Artificial Neural Network algorithm.
- To design a test rig for detecting the bearing faults.
As the project is diagnosing bearing fault in motor, the motor would be induced with an unhealthy bearing for the sake of data acquisition. For fetching of data, the motor would then be operated under unhealthy conditions and each of its phase current (3-phases) would be monitored through SCT-013-005 Split-Core AC Current Sensor. The current sensors output needs to be digital as Raspberry Pi doesn’t support analogue inputs. For overcoming this, ADS-1115 Analog to Digital Converters would be interfaced with Raspberry Pi through I2C mode. Then these signals would be passed through some signal processing techniques i.e. Wavelet Transformation, Fast Fourier Transform (FFT) to analyse the change in its frequency spectrum. These spectrums would serve as dataset for training of ANN model. ANN algorithms like Back Propagation or Convolution Neural Networks would be developed using Python programming language through its APIs like Numpy to distinguish between a healthy and faulty motor.
An implementation block diagram is shown below:
The more we learn about this topic, the more we appreciate the benefits of it. If we talk about applications of condition monitoring, there are a bundle of applications like:
- Allows one to predict possible malfunction and to schedule the motor shutdown if a prefixed rate is exceeded.
- Fault diagnosis with full automation.
- Its compact architecture configuration makes it suitable for real time detection and monitoring.
- It has capability to provide efficient training of the classifier with limited size of training dataset.
- The most appealing feature of ANN is that it can be used for online diagnostics.
- Its accuracy is not limited to any certain point as it increases with each sample.
- It is cost effective and practical in real time hardware implementation.
The more we learn about this topic, the more we appreciate the benefits of it. If we talk about applications of condition monitoring, there are a bundle of applications like:
- Allows one to predict possible malfunction and to schedule the motor shutdown if a prefixed rate is exceeded.
- Fault diagnosis with full automation.
- Its compact architecture configuration makes it suitable for real time detection and monitoring.
- It has capability to provide efficient training of the classifier with limited size of training dataset.
- The most appealing feature of ANN is that it can be used for online diagnostics.
- Its accuracy is not limited to any certain point as it increases with each sample.
- It is cost effective and practical in real time hardware implementation.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 46095 | |||
| Raspberry Pi 4 Model B (4 GB RAM) | Equipment | 1 | 14650 | 14650 |
| SCT-013-005 Current Sensor | Equipment | 7 | 835 | 5845 |
| ADS1115 16-Bit ADC | Equipment | 4 | 475 | 1900 |
| 0.5 HP 3-Phase Induction Motor | Equipment | 2 | 11540 | 23080 |
| Breadboard | Miscellaneous | 2 | 160 | 320 |
| Connecting Wires | Miscellaneous | 30 | 10 | 300 |