Karim Gmir explores – Maintenance 4.0 has been a trending topic for several years now. Since the emergence of cloud computing, internet of things and artificial intelligence, people and consultants from the asset management community have been seeking to implement this new technology to enhance maintenance processes and optimise all kind of maintenance KPIs such as MTBF and MTTR. In this article, I will present a solution which enables the organisation to react quickly when failures occur. I will go through the definition of MTTR, the MCSA technique and its advantages and finally the application designed in IBM Maximo and Watson IoT Platform.
Maintenance 4.0 has been a trending topic for several years now. Since the emergence of cloud computing, internet of things and artificial intelligence, people and consultants from the asset management community have been seeking to implement this new technology to enhance maintenance processes and optimise all kind of maintenance KPIs such as MTBF and MTTR. In this article, I will present a solution which enables the organisation to react quickly when failures occur. I will go through the definition of MTTR, the MCSA technique and its advantages and finally the application designed in IBM Maximo and Watson Iot Platform.
MTTR short for mean time to repair is simply a duration from the start of fixing the failed equipment to the end. In this phase, the maintenance workforce performs the following tasks:
This time frame is considered finished as soon as the equipment starts running over according to the specifications. If the equipment is still failing during the testing phase, the maintenance agent starts over from step 1.
In this article, I chose to study different failures of an electrical induction motor. The following section explains why this asset is highly critical in a manufacturing process or any automated industrial process.
The induction motor represents 90% of all motors that are used in any industrial process or system. This makes the equipment one of the most used asset which is why I chose it for this article as a case study.
An induction motor has two principal components:
The next video explains with details how does the induction motor works.
The main topic of this article is to find how we can optimise the MTTR of a failing motor and more specifically the duration of failure diagnosis. In order to find answers to this purpose, we need to list all the possible failures. I found an interesting article which lists all the defaults that could impact the health of an induction motor.
Next, I would like to introduce a really interesting technique of diagnosis which enables the technician or the maintenance agent to determine the occurring failure remotely and without dismantling any components of the electrical motor. The technique is called MCSA. Let’s dive into more details in the next section.
MCSA short for Motor Current Signature Analysis is a tremendous technique which enables the maintenance and reliability engineer to monitor the motor health online without stopping any production process. The MCSA records the motor current in a time domain format. By applying a fast Fourier Transformation on the recorded signal, you obtain the current frequency spectrum.
Since every behaviour and every failure has its own frequency, nothing can escape from the MCSA. This technique is tremendously reliable. We can record broken bar default or failing bearing in the frequency spectrum of the motor current.
Moreover, the MCSA is very cost effective compared to the other techniques. In fact, you only need a handy device to monitor the motor current and develop a small application on Matlab to apply the FFT in order to display the frequency spectrum. The question is how can the reliability or maintenance engineer implement this kind of technique in order to decrease the MTTR ?
I developed from scratch an application which enables to diagnose the default without wasting any time. In fact, while the equipment remains running, the application monitors the motor current and displays it in a graph and uses it to determine the occurring default in real time.
As explained previously, the MCSA technique displays the frequency spectrum from which we can extract several parameters which I call health parameters. I used this dataset to train a classification model in order to determine the occurring defaults.
For this purpose, I recorded the current signal of 3 different induction motors. The first one is healthy and is running without observing any particular failure. The second one has broken bar inside the rotor and the third one has 2 broken bars.
I used this sensor to record the current signal. I designed this application in order to accomplish the following steps:
Basically, this report is for reliability engineers who understand the different behaviours and the different failures which could occur when the electrical motor is running.
As shown in the next graph, it displays a spectrum of frequency. One of which corresponds to the power supply frequency which in our case is 50 Hz. We can see next to it small peaks on the left and right side of both motors which have broken bars.
Before designing any application or any automated workflow using fancy tools or artificial intelligence programs, experts need to consult this kind of report when they are managing a failing induction motors. Peaks on both sides of the 50 Hz frequency represent the behaviour of the broken bars default. Since every default frequency is pre-calculated using a specific equation, experts can monitor the intensity of those peaks and react accordingly.
In this report, your customer could ask you to display further details like ongoing work orders or scheduled preventive maintenance. Please contact me if you need any assistance in defining or designing a BIRT report on IBM Maximo.
The following dashboard displays 3 different cards for each monitored induction motor:
This is a simple dashboard that displays relevant data which could require your customer. You can of course import more data by connecting your external system to your application using native APIs. More details could be added in another cards like your custom maintenance, inventory or even financial expenses KPIs.
More interesting, this dashboard can be shared to multiple users. You will have define security roles and grant access according to the business responsibilities within the organisation. In my case, I share access to prospects so that they can see the dashboard. This enables them to think about their business process and how they can implement this technology in their everyday operations.
Watson machine learning is a service in the IBM cloud. I used it to train a classification model which enables the maintenance workforce to determine the occurring failure in real time. Waston ML is basically a service which is connected to an application. It calls the trained model for scoring using APIs.
I used IBM SPSS which is an integrated module of IBM Watson studio in order to create and train the classification prediction model. In the next figure, you can see the designed model which is very simple design and doesn’t require specific and data extended knowledge.
After creating the model in IBM SPSS, I deployed it by creating a new deployment instance in the IBM Cloud Pak for Data platform. Since 2017, IBM has been changing its platform to improve performance and security. Please leave me some comments if you think that something needs to be changed.
This feature in IBM Maximo makes it a real differentiator among the EAM solutions on the market. In fact, you can define and automate business processes so that you won’t have to manually perform validation, notify people by emails and text messages or trigger different events under certain condition. The following video is an interesting workflow tutorial for beginners
The purpose of this application is to decrease the MTTR. In previous sections, we learnt how to quickly diagnose an occurring failure using artificial intelligence programs. The next task is to actively plan maintenance activities according to the failure and what a better feature to perform this other than Maximo workflows.
In fact, you can define a workflow which allows you generate a service request or work order if the scoring result exceed a predefined threshold. In those newly created records, you attach the right job plan in which material, labor crafts and services are already defined in the job plan application. You can go even further by performing a reorder of the needed materials and wait for the delivery.
As said previously, workflows on Maximo brings so much value to the business operations. If well defined and designed, workflows bring accuracy and integrity. Therefore, you will worry about a missing notification or action.
In this articles, you learnt how to optimise the mean time to repair an electrical induction motor which is present in more than 80% of the automated industrial processes. By monitoring the current signal, no failure would escape from your diagnosis. Moreover, you can actually detect the occurring failure even before you could see or sense any physical behaviour like vibrations or noises.
If you are interested to have more details about this use case or looking for ways to implement this technology with its functional workflow, please click on this link and leave your information.
Read more of Karim’s work here