Introduction to Predictive Maintenance Solution

1.0 INTRODUCTION

Maintenance of any machine is the process of keeping it in good condition. Everything is bound to depreciate with time, so maintenance is required to preserve machine health.

Maintenance costs take up a major portion of total operating costs in most factories. Organization spend millions of dollars on ineffective maintenance techniques. Not having a scheduled plan to decide on when maintenance should be performed to prevent equipment malfunctioning is the main reason behind these extra expenses.

2.0 Approaches to Maintenance

There are mainly two approaches towards maintenance:

  1. Reactive Approach
  2. Proactive Approach

2.1 Reactive Approach

In Reactive approach, we have corrective maintenance which is nothing but running for failure management. It simply means to fix the equipment once it breaks down. This is the oldest approach to maintenance and the most expensive one. Essentially it means waiting for the machine to stop functioning before any maintenance activity is performed. This results in higher costs for spare parts, labor cost, machine downtime, and low production.

Maintenance activities performed after a jet aircraft engine has stopped working is an example of corrective maintenance.

2.2 Proactive Approach

Next comes the proactive approach which is a process of periodically performing technical assessments and inspections on the equipment and machinery to minimize the likelihood of failure. For example, performing assessments on a jet aircraft engine which is functioning properly after every two weeks is an example of preventive maintenance, also known as scheduled maintenance.

2.3 Predictive Maintenance

Predictive maintenance is a condition-based preventive maintenance approach. Instead of periodically performing assessments and checks on the equipment or waiting for the equipment to fail, predictive maintenance makes use of actual data of the devices and equipment to predict when maintenance should be done. This approach has great economic value for the organization because it enables maintaining machines proactively to avoid business disruption.

With IoT (Internet of Things) coming to mainstream, sensors can be attached with industry equipment and machinery. These sensors can be used to monitor the operating conditions of the equipment. The information provided by these sensors can be used to analyze equipment health, calculate MTTF (mean time to failure) and RUL (Remaining useful life). We can build solutions on top of that to see whether the machine is operating in normal mode or if there is an interruption in the normal working flow and schedule maintenance accordingly. The solution will let you know when an equipment needs maintenance instead of you wasting a lot of time and money on unnecessary scheduled maintenance or waiting for the equipment to fail.

3.0 Approaches to Predictive Maintenance

Predictive maintenance can be implemented through the following techniques:

  1. Anomaly detection 
  2. Vibration analysis  
  3. Machine Learning

3.1 Anomaly Detection

Anomaly means something different or strange. Anomaly detection is the process of detecting data values that deviate from the regular pattern. The idea behind this approach is simply that when an equipment is functioning properly, operational values such as temperature, pressure, and current of that equipment behave in a certain way. An anomaly detection solution can be used to detect values that deviate from the standard pattern and generate alerts based on that. Many algorithms are available that use different statistical techniques for anomaly detection each with its own pros and cons.

3.2 Vibration Analysis

Vibration analysis is one of the most used techniques for predictive maintenance. The input values obtained from the machinery are analyzed with respect to Fast Fourier Transform (FFT) and Inverse Fast Fourier Transform (IFFT); based on that future spectrogram is obtained which can lead to an accurate assessment of the failure of machinery or equipment.

3.3 Machine Learning

Machine learning model can be trained to predict that an equipment is malfunctioning and requires maintenance. Furthermore, the model can be used to calculate MTTF (mean time to failure) which can be used to predict failure within a given time. In Machine learning, we have two approaches; one is supervised ML that can be used when we have enough historical data that represents all possible scenarios. The other approach is unsupervised ML which is used when we don’t have historical data to train our model.

These approaches can be used either separately or combined to build an intelligent predictive maintenance solution for accurately predicting when an equipment or machinery is likely to fail or when it needs maintenance. This can result in lower maintenance costs, higher productivity and better product reliability.

We have series of blogs coming up to discuss these approaches in detail. This blog was to give you an overview of different maintenance approaches that are used and how predictive maintenance provides an efficient solution to maintenance problem.

We at AlphaBOLD are dedicated in providing predictive maintenance solutions to help your business grow.

If you have any question or queries, do not hesitate to reach out to us