Condition Monitoring of Metal-Cutting Machine

FH Joanneum, University of Applied Sciences, has six departments at three sites across the eastern Austrian province of Styria: 

• Graz,

• Kapfenberg, and 

• Bad Gleichenberg. 

Within the Department of Management, the Institute of Industrial Management operates the Smart Production Lab in Kapfenberg. 

This laboratory is a research facility for accelerating digital transformation processes. Step by step, the laboratory utilizes state-of-the-art technologies and implements use cases for digitalizing a classical production process. 

Companies, research institutions, and students are trained hands-on in this environment on the possibilities and tools of digitalization and how these can enhance the workflow of industrial production. This includes practical experience with tool condition monitoring systems, which play a key role in improving machine efficiency and predictive maintenance.

Machinery state - the problem

Modern machinery connects to the Industrial Internet of Things (IIoT). This results in an interconnected network of sensors, instruments, and devices integrated with industrial applications. It is prevalent within manufacturing and production processes. 

These connected devices collect, exchange, and analyze data to monitor, control, and optimize industrial machining processes in real time.

In the metal industry, heavy, old, and non-digitized machines are still widely spread. These machine tools can still manufacture high-quality parts, and they are not yet obsolete. But to use the machines more efficiently and plan their utilization effectively, information about the machines' state is necessary. 

Examples of such evidence include the current condition of used tools or the wear of critical machine tool parts. This long-term data can prevent tool failure and prolong tool life. 

Additionally, it is essential to know the actual state of the machine, specifically, whether it is in use or not. This information is crucial for accurately analyzing Overall Equipment Effectiveness (OEE). Relying solely on data entered into the production control terminals may not reflect the machine’s true usage. However, to retrieve this information from a machine, engineers must collect data about the machinery's state.

To retrieve relevant information from a non-digitized machine, engineers must retrofit it with specific sensors, and visualize and store the data for further analysis and interpretation. This task involves a digital retrofit, integrating additional hardware into the machine to generate the data for process improvement.

State analysis - the concept

The PDCA cycle—short for Plan-Do-Check-Act—is a structured, iterative method used to guide the digital retrofit of metal-cutting machines. It’s a practical roadmap for transforming traditional machines into smart, data-driven assets through sensor integration and digital monitoring, which ensures systematic implementation and continuous improvement of tool condition-based monitoring systems. 

To conduct a digital retrofit, specific steps based on the PDCA cycle are defined to ensure a functional solution and support the goal of making the production process more efficient. The phases of the PDCA cycle are:

Plan

1) Use case definition
Defining the final situation regarding the machine and its production process.    
What conditions/parameters/processes should be improved?

2) State analysis
It must be elaborated on what information and data are currently available about the machine, as well as which functionalities.      
What is the current workflow like, and what data is already accessible?

Do

3) Action planning     
Defining the measures needed to implement the defined use case.
What sensors and measurement systems are required, and how are they integrated into the machines?

4) Implementation     
The hardware components must be set up and attached to the machine.
Where and how are the sensors and measurement system mounted on the machine?

Check

5) Industrial Internet of Things (IIoT) Integration      
Setting up the communication between the hardware components and the machine, as well as the publication of the collected and preprocessed data. 

How are the sensors and measurement systems connected? How is the data made accessible, and what preprocessing is needed to retrieve information?

6) Verification
Testing the technical functionality.
Are the sensors delivering data, and is it processed and visualized properly?

Act

7) Validation   
Demonstrating that the implemented use case aligns with the one defined initially.
Can the necessary information be extracted from the collected data to improve the machine's conditions, parameters, or processes?

8) Standardization      
The proof of concept needs to be improved and ultimately implemented on other machines as well.
What improvements are needed, and how can the setup be applied to other machines?

Tool condition monitoring (TCM) - the solution

In this case, a band saw is the metal-cutting machine where a lack of information existed. However, to determine the conditions of the tool (monitor tool wear) and its efficiency, further data is needed. 

First, we can conclude the machine's production state over time. Specifically, define the three states: Idle, Running, and Cutting. From the machine, already on the panel, we can retrieve the configured cutting velocity and feed, but no further information or data is available.

Based on this, we selected an acceleration sensor (Hansford HS100S010M12M6x1), a microphone (PCB 130A24), and a temperature sensor (thermoelement type K) to capture the machine's condition and state. To access, preprocess, and visualize the data, we chose a DEWE-43A. To measure the vibration of the saw blade, we mounted the acceleration sensor with a magnet on the cover of the saw blade's guidance. 

To minimize environmental noise, we positioned the microphone within a 3D-printed adapter adjacent to the cutting position and directed it towards the workpiece. The temperature sensor was fixed in place with duct tape on a cooling rib of the primary motor as it warms up during cutting. Additionally, the workpiece heats up, but the sensor would have to be attached and removed for each cut.

List of equipment used

• DEWE-43A: Data acquisition instrument with USB interface. It includes eight analog input channels. 24-bit / 200 kHz analog-to-digital converter (ADC). 8 super counter channels for encoders, tacho sensors, and digital IO with 100 MHz digital inputs. Two CAN bus ports.

• DSI-TH-K: Connection adapter for Type K thermocouples.

• DSI-ACC: Connection adapters for IEPE sensors (vibration sensor and microphone).

• DewesoftX Professional: Data acquisition and signal processing software package (bundled with DEWE-43A).

• Hansford HS100S010M12M6x - IEPE vibration sensor.

• PCB 130A24 - IEPE microphone.

• Thermocouple type K temperature sensor.

After positioning the sensors, we connected them to the DEWE-43A with the appropriate cables. Furthermore, we connected the measurement system to a laptop, where we configured it within DewesoftX. After the configuration, we set up a dashboard to visualize the data acquired from the three sensors.

With the completed setup test, cuts have been made using the band saw to verify that the sensors deliver reliable data. We derived further limit values to determine if the saw is Idle, Running, or Cutting. Based on those limits, alarms have been configured and are displayed within the dashboard to inform the production worker.

Furthermore, we need to forward the information to the production controlling system and test the solution on other machines, such as a turning or milling machine. Additionally, the collected data needs to be compared over time to conclude the wear condition of the saw blade.

State determination - the results

Based on numerous test cuts and measurements, we demonstrated that, using only vibration data, we can derive information about the current state of a band saw from the limits that the average acceleration exceeds. 

Even clearer is a Fast Fourier Transformation (FFT) of the vibration data, which transforms the signal from the time to the frequency domain, revealing the frequency components of the signal. The FFT shows that the saw plate is cutting within the material, with predominantly high frequencies and very few low ones. Where else if just the saw plate is running, the frequencies of the vibrations are homogeneously distributed. 

Further, in an industrial environment, the sound pressure alone cannot determine the state of a band saw, as environmental noise heavily influences it. Already talking causes errors within the signal. But the first analysis has also shown that, for acoustics, an FFT interpretation will enable a conclusion about the wear conditions of the saw blade. 

Regarding the use of the cool rip temperature in this context, it increases at a higher rate when the saw blade is cutting, rather than just running in air. This phenomenon only occurs when there is no air circulation around the motor. Already opening one window of the 600m² laboratory causes air movements that cool the motor, but the temperature reading is no longer reliable. 

These conditions lead to the conclusion that the machine's production states — Idle, Running, and Cutting — can be determined solely by monitoring the vibration; however, for interpreting the wear conditions of the tool, analyzing the sound's frequency distribution is necessary.

Conclusion - the value added

In the case of metal-cutting machines, such as the band saw, knowing the real-time machine state (e.g., Idle vs. Cutting) through vibration and sound analysis helps calculate the actual availability component of OEE. Such analysis avoids relying solely on what the operator enters into the control terminal, which may not reflect actual usage.

Companies, students, and other visitors to the Smart Production Lab can now see the potential of digital retrofitting. For this, the lab has adapted the solution within a hands-on retrofitting exercise, where participants digitalize the band saw entirely on their own. During the exercise, not only is knowledge gained about the potentials of retrofitting and its methods, but also crucial correlations between physical values and machine behavior are understood. 

The solution can be applied broadly, as the exercise demonstrates the ease of fitting sensors on an existing machine. Furthermore, participants learn tools to easily acquire, process, and visualize various types of data. The setup demonstrates that Dewesoft is a leading partner for measurement systems, as well as for setting up comprehensive solutions to address specific industrial issues.