Digital and wireless Operating Deflection Shape (ODS) for assets condition monitoring

Abstract—The maintenance of industrial assets is responsible for a large part of the costs of an industrial plant. In this context, low-cost sensors were developed to measure physical quantities and enable online monitoring of assets. Most of these sensors use vibration measurements as a method for detecting and diagnosing faults or behavior deviations. This paper shows a method for using the Operating Deflection Shape (ODS) vibration signal analysis technique in conjunction with low-cost wireless condition monitoring IoT sensors. The developed method was confronted with commercial devices that perform traditional ODS, thus, laboratory and field validations are presented.

Index Terms—Operating Deflection Shape, condition monitor- ing, maintenance, IoT, wireless sensors.

I.    Introduction

Plant maintenance represents a critical component for the success of the modern manufacturing industry. Studies show that depending on the industry between 15 and 70 percent of total production costs originate from maintenance activities [1]. Due to this criticality, maintenance teams use the strategy known as Predictive Maintenance or Condition Based Main- tenance (CBM), a condition-driven maintenance program that uses monitoring systems to determine the actual mean-time-to- failure, optimizing the availability of the process and reducing maintenance costs. The result is an overall improvement in product quality, plant productivity and ultimately profitability [2].

With the advances of the IoT and Industry 4.0, this kind of maintenance became much easier and cost-effective, lead- ing to the widespread of this technique. As a result, many systems were developed to monitor vibration, temperature, pressure, and other variables in industrial plants. Due to the advances in micro-electromechanical (MEMS) systems, it is possible to deploy thousands of low-cost sensors capable of sensing, computing, and communicating wirelessly to gather information for the environment and equipment monitoring [3] – [18]. They send data to the cloud for storage or further processing using IoT protocols and technologies [4]. Many of the public cloud service providers offer IoT services using standard protocols for real-time storage and extract analytics from the data [5]. Any potential problems are notified to the plant personnel as an advanced warning system. This enables plant personnel to repair or replace equipment, before their efficiency drops or they fail entirely. In this way, catastrophic equipment failures and the associated repair and replacement costs can be prevented, while complying with strict environ- mental regulations [6]. A typical architecture of the wireless sensor network is shown on Fig. 1.

The architecture showed in Fig. 1 in detail:

•  Assets: in the asset layer are the machines and equipment to be monitored and mainly involve electric motors, gearboxes, pumps, and any other devices that require attention from the maintenance teams.

•  Sensors: usually composed of sensor components ca- pable of measuring mechanical vibration, temperature, magnetic field, atmospheric pressure as well as other physical parameters. A microcontroller is responsible for controlling the scheduling of data collection from the sensors and sometimes doing some processing of these data. A wireless transmission unit transmits the collected data to the Gateway.

•  Gateway: it’s a wireless device that, through communi- cation with sensors, concentrates the collected data and sends it to a server, usually a Cloud solution. The gateway is capable of carrying data traffic from various sensors.

•  Analytics and Condition alerts: applications generally built to run on Cloud-type servers. Analytics solutions are designed to track changes in the behavior of data sent by sensors. Several techniques are used from deterministic methods to Machine Learning. Based on the results of the Analytics application, Condition Alerts are sent to users through an online platform or smartphone mobile applications.

Condition monitoring and fault detection at an early stage is an essential need, especially in critical application equipment. Non-invasive fault diagnosis helps save cost, time and effort by reducing maintenance. Accurate fault characterization is crucial to this process [7]. The basis of any reliable detection method is to understand the electrical, magnetic and mechanical behavior of the equipment in its standard condition and under fault conditions [8]. The most used indicators for monitoring electrical machines are based on current, temperature, voltage, chemical debris and vibration changes [9].

Vibration measurement is a well-established method for detecting and diagnosing any deviation from normal conditions in rotating mechanical structures. However, using conventional piezoelectric accelerometers to measure these vibrations is the predominantly used and accepted technology. Compared with traditional acceleration sensors, MEMS acceleration sen- sors have the advantages of small size, low cost, high data integration and accuracy, which have been widely used in natural disaster monitoring, tilt detection [10], displacement or velocity measurement, industrial vibration monitoring and other fields [11].

Many vibration signal analysis techniques are used in the diagnosis of industrial assets. Among them, the Operating Deflection Shape (ODS) stands out, a simple way to perform dynamic analysis and see how a machine or structure moves under its operating conditions. ODS tests have no external forces applied and only response vibration signals are mea- sured. It is very useful to identify causes of defects or to show the behavior of the machine to people who are new to vibration analysis [12]. The ODS technique has several applications in civil engineering [13], industrial rotating machinery [14] and transportation [15]. To obtain an ODS, simultaneous measurement of vibration responses at multiple locations is required. The greater the number of measurement response locations, the more informative the ODS.

Within this context, this work presents a technique for performing digital and wireless ODS (DW–ODS). In this way, the theory behind the ODS analysis technique, the method chosen for the synchronization of wireless IoT sensors, the user interface and field and laboratory validations will be presented.

II.      Digital and Wireless Operating Deflection Shape (DW–ODS)

A.    Operating Deflection Shape

ODS measurements are collected for specific frequencies using spectrum analysis. A measurement location is identified as a reference, and vibration and phase levels are measured for other locations in the structure using a roving sensor. The relative phase and amplitude values are sufficient for animation and the measurement methods commonly found are:

•  Order Spectra,

•  Phase Assigned Spectrum (PAS),

•  Ratio Based PAS,

•  Run-up / Down ODS,

•  Transmissibility (FRF between responses).

The most common technique is Transmissibility due to its simplicity in data acquisition and analysis. The ODS values are given directly by the relationship between the reference signal and the others signal measurements. The definition of transmissibility is as follows [16]:

where Xj is the response measurement in the reference location j, Xi is the response measurement in the i locations and the Tij is the ratio between the reference measurement and other positions measurement. This may be considered as the ratio of two linear combinations of the operating forces spectra. All measurement signals are complex and are in the frequency domain (Ω).

With the amplitude and phasis obtained by eq. 1, in the locations i measured in the operational frequencies Ω, is possible to verify the movement of the structure in each position.

The most common and minimum requirements necessary to perform an ODS are DAQ (Digital Acquisition) hardware with two or more channels, the equipment must have the function of ODS, accelerometers and cables, computer for the user interface in order to create the interface model and conduct the measurement. The interface model is only representative and does not require the actual dimensions or details that complicate its construction. It is enough to know the region where some special component is allocated. When properly modeled the ODS analysis can provide information about what is wrong with the structure. Some examples of checks that can be made are listed below:

•  Overall movement of the structure: Checks that the en- tire equipment is moving in solidarity, with no relative movement between components;

•  Relative motion between bearings and shaft: can provide an indication of misalignment;

•  Phase and relative movement between bolted and welded joints: may indicate loose fasteners. Backlash problems show similar movement for different frequencies;

•  Bending movements of structure components: may be an indication of resonance (Note: ODS does not prove resonance);

•  Localized movements in the machine’s feet or base: problems of lack of rigidity of the base.

The ODS is simple to apply but it requires knowledge about vibration analyses [17]. Fig. 2a presents an approximate model of the machine. The illustration of the structure’s motion is obtained through amplitude and phase measurements at various locations on the machine as shown in Fig 2b.

B. Measurement equipment

Considering the traditional tests demonstrated in this work, the following equipment was used, also shown in Fig. 3:

• A data acquisition system (DAQ), Dewesoft Sirius with 24 channels;

• Piezoelectric accelerometers for measuring vibration data in the laboratory, model 353B14, PCB Piezotronics;

• Piezoelectric accelerometers for measuring vibration data in the field, model 603C01, PCB Piezotronics;

• Software for analysis of results, Dewesoft X;

• Shaker used for laboratory validations, Type 4808, by Bruel & Kjær;

• Power amplifier, Type 2719, by Bruel & Kjær.

For wireless analyzes, were used: a wireless IoT sensor (with an embedded MEMS accelerometer and Bluetooth Low Energy communication protocol), and a cell phone with a specific application for analysis and condition monitoring. The wireless sensors are battery-powered and can be attached to the structure with glue, a screw or with a magnetic base.

A.    Wireless Sensor Synchronization

To perform an ODS analysis, the vibration measurements must be accurately time synchronized. Usually, this analysis is used to detect low-frequency deformations in structures, so the synchronization needs to have a low enough error not to distort the frequencies of interest.

There are multiple techniques to synchronize wireless nodes in a network. These wireless nodes can also have multiple communications protocols. In the case of this paper, the wire- less condition monitoring IoT sensor was already developed and it uses Bluetooth Low Energy (BLE) to communicate with the rest of the application. Thus, only the BLE synchronization methods are of interest.

Most synchronization techniques have the same formula, i.e. one sensor becomes the synchronization master so that it becomes the timing reference for all others. The other sensors in the network are ahead or behind the master with their time. All other sensors become synchronization slaves and must adjust their internal counters to have synchronous values to the master. There are essentially three ways of doing BLE time synchronization:

1)   Using the periodic advertising packets to communicate a time base from a master wireless IoT sensor to other slave wireless IoT sensors. This method needs to compensate for the various delays caused by the BLE protocol itself. With this technic at [19] an error of 10 µs was reached;

2)   Making use of BLE connection events to perform the time synchronization. This can happen because when the slave and the master are communicating to perform a connection, the connection event takes place almost at the same time in both nodes. In [20] an error of around 750 µs was achieved with this method;

3)   Using the radio directly when it is not busy by making BLE actions and transmitting synchronization packets from master to slave nodes. This method has a higher power consumption because it uses a 16 MHz timer instead of a 32 kHz Real Time Counter (RTC). In [21] an error of around 320 ηs was achieved;

Method 3 was utilized in this project because of the low error. Besides that, the high power consumption is not a prob- lem once the synchronization will occur sporadically. Other than that, the synchronization needs to happen just before the vibration measurements, and this method also has a proper way of scheduling a synchronous task between the nodes, that was used to start the measurement synchronously. The process starts with a sensor taking the position of synchronization master, sending packages containing time information, while the other surrounding sensors become slaves and receive these packages, updating their internal timers to be synchronous to the master. BLE is still functioning during the process of synchronization in a way that every other communication can take place normally.

The synchronization method adopted was implemented in the condition monitoring sensors and tested in the laboratory. Test setup and result are shown in Fig. 4. Two sensors were coupled to a shaker actuating with a 60 Hertz senoidal signal and 0.25 g of acceleration peak. Both sensors started the measurement at the same time by the synchronization and it is explicitly that the phase between the measurements is close to 0 degrees.

Multiple tests were performed in the same setup in order to measure the synchronization error. The tests showed that the error stays in a range of 20–500 ηs which is enough not to distort the low frequencies of interest.

The range of synchronization is directly associated with the radio performance of the device and the environment. In an open field, the tests showed that slaves in a radio 10 meters from the master were able to perform a synchronized vibra- tion measurement. Of course, in an industrial environment, these values will vary according to the machines and barriers between the master and slaves.

D.    User Interface

The complete digital and wireless ODS process is made via mobile application in a way that all the interface with the sensors and the signal processing becomes invisible to the user. Then, to perform the DW–ODS, the user needs to follow these steps on the smartphone:

Step 1: Select the sensors that are coupled to the structure of interest;

Step 2: Take a picture of the structure that shows where all the sensors are;

Step 3: Indicate the position of each sensor in the picture;

Step 4: Wait while the smartphone communicates with the sensors to perform the measurements and send it to the cloud to perform the signal processing;

Step 5: Check the results of the DW–ODS.

After item Step 3 the smartphone will connect to all the sen- sors and perform the synchronized measurement. In sequence, that will be processed in the cloud and only the amplitude and phase of each sensor to the 5-highest amplitude frequencies return back to the smartphone. With this information, the smartphone plots the screen animation on top of the indicated position of the sensors showing how the structure is deflecting in that specific frequency at that specific point of measurement.

A complete data flow is shown in Fig. 5. The whole process takes around 3 minutes for an DW–ODS performed with 7 sensors, the time may vary depending on the number of sensors and the distance between the smartphone and each sensor. To perform a quality ODS or DW–ODS, as many sensor elements as possible are required. In the DW–ODS, a maximum of 7 sensors were used, due to the measurement and collection time that can be extended depending on the wireless communication conditions.

III.        Results and discussion

A.    Laboratory validation case

In order to validate the whole process, an experiment was realized in a controlled ambient. The experiment shown in Fig. 6 consists of a beam with one of its tips attached to a solid structure. The beam is made of SAE 1020 steel (modulus of elasticity: 205 GPa; Density: 7870 kg/m³), with the following dimensions: length of 1000 mm - between the end of the beam and the fixed support, width of 25.4 mm and thickness of 6 mm. In the middle of the beam, a shaker (Type 4808, Bruel

& Kjær, with the Power Amplifier) is coupled to perform excitation on the system. Five sensors were equally distributed in the beam length to make the measurements.

In the case of traditional tests, piezoelectric accelerometers connected to a DAQ system (shown in section II-B) were used. For the wireless tests, five wireless IoT sensors and a cell phone with a specific application for the ODS analysis were used.

Prior to the experimental test, a numerical simulation was carried out (performed in the Ansys software, version 2022 R2) of the cantilever beam under study, to verify which are the mode shapes and their respective associated natural frequencies.

Obtaining this information, it was possible, through an excitation controlled by the shaker, to apply an excitation at a natural frequency of the beam (the 2nd mode shape was chosen, at the frequency of 31.376 Hz). In this way, it was possible to compare the results obtained by the numerical simulation, by the traditional ODS analysis and by the digital and wireless ODS.

It can be observed that the results in Fig. 6 of each type of analysis are very close to the numerical simulation model, both for the traditional ODS analysis (31.5 Hz, 0.4% variation to the model) and for the wireless ODS analysis (31.0 Hz, 1.2% variation to the model). This result was obtained by performing only one measurement with the set of sensors. Furthermore, the deformation indicated for both analyses is also very close. Thus, in a controlled environment, the potential of DW– ODS analysis is verified. In the later section, this potential is explored in a field case.

B.    Field validation case

A comparative verification between the results obtained by the traditional ODS method and the DW–ODS was also carried out in an industrial environment. An application was chosen in an industrial pump (WEG W22 motor 50hp, 60Hz, 200L frame / KSB pump, Meganorm 100.315) connected to a frequency inverter. The pump selected is localized in Shaft Machining Area, at WEG Factory located in Jaraguá do Sul, Brazil.

Measurements were defined for performing the ODS at seven points of the motor and pump assembly. To carry out the traditional ODS tests, seven piezoelectric accelerometers, a DAQ and the software for carrying out the ODS analysis were used (more information in the section II-B). As for the digital and wireless ODS, seven wireless IoT sensors were used, and a cell phone with a specific application for ODS analysis. The assembly of both systems is shown in Fig. 7.

For the comparative analysis, the three frequencies where there is a significant displacement were analyzed. The results of the traditional ODS and DW–ODS analyses are shown in Fig. 8.

Checking the first three ODS frequencies for both sys- tems, one can see great proximity in defining the relevant deformation frequencies. The variation between frequencies (traditional and wireless analyses) is about 3%, it is important to emphasize that the measurements were carried out in series, that is, they were not carried out at the same time witch can also lead to increase the variation. Furthermore, the deformations shown by the DW–ODS have the same shape and direction as those performed by the traditional method. This

demonstrates the full potential and excellent cost-effectiveness of using digital and wireless ODS.

IV.        Conclusion

Taking into account the criticality of maintenance within industrial plants, more and more predictive maintenance and asset monitoring solutions are emerging. In this context, digital and wireless ODS appears as a cheap way of using a technique already known and used in the industrial environment.

There are numerous advantages to using wireless IoT sen- sors to perform an ODS. Initially, it can be stated that there is an increase in the practicality of the process, reducing the time required for data collection. Furthermore, it is not necessary to have a specialist in the field to carry out this process, it is enough for the specialist to indicate the measurement points on the structure. Also, the post-processing of the data is carried out instantly at the time of collection, making the specialist focus only on the conclusion over the measurement. Finally, as wireless IoT sensors are much cheaper than common devices that perform ODS, the whole process becomes cheaper.

However, the solution still has room for improvement. At first, it would be possible to carry out the process using a gateway, and not only the mobile application, so that the whole process can occur automatically and periodically. Furthermore, it is possible to improve the user interface so that it indicates the orientation of each sensor so that the comparison axis in the signal analysis is always configurable. Finally, it would be possible to expand the structure deformation animation to 3D, making the whole system more complex and at the same time improving the analysis capacity.

It is also possible to take into account that the technique exposed in this article is taken as an addition to an industrial asset monitoring system. That is, in a system where the sensors already perform periodic measurements of vibration, tempera- ture, and other quantities, it is still possible to carry out DW– ODS, increasing the analysis capacity of technology users. This process empowers industrial maintenance specialists and increases the efficiency of overall manufacturing processes.

References

[1]    M. Y. You, F. Liu, W. Wang, G. Meng, Statistically planned and indi- vidually improved predictive maintenance management for continuously monitored degrading systems, IEEE Transactions on Reliability 59 (4) (2010) 744–753. doi:10.1109/TR. 2010.2085572

[2]    R.K. Mobley, “An Introduction to Predictive Maintenance,” Van Nos- trand Reinhold, New York, 1990.

[3]    Liu, Y., and Xu, X. (October 6, 2016). ”Industry 4.0 and Cloud Manufacturing: A Comparative Analysis.” ASME. J. Manuf. Sci. Eng. March 2017; 139(3): 034701.

[4]    L. D. Xu, W. He and S. Li, (Nov. 2014) ”Internet of Things in Industries: A Survey,” in IEEE Transactions on Industrial Informatics, vol. 10, no. 4, pp. 2233-2243, doi: 10.1109/TII.2014.2300753.

[5]    Industrial Wireless Sensor Networks: Challenges, Design Principles, and Technical Approaches

[6]    Magadan, Luis & Sua´rez, Francisco & Granda, Juan & Garcia, Frk Daniel. (2020). Low-cost real-time monitoring of electric mo- tors for the Industry 4.0. Procedia Manufacturing. 42. 393-398. 10.1016/j.promfg.2020.02.057.

[7]    Bindu S. and Thomas V. V., “Diagnoses of internal faults of three phase squirrel cage induction motor - a review”. (2014). 2014 International Conference on Advances in Energy Conversion Technologies (ICAECT).

[8]    R. Fiser, “Steady state analysis of induction motor with broken ro- tor bars”, In: Seventh International Conference on Electrical Ma- chines and Drives, Durham, UK, 1995, vol. 1995, p. 42–46. doi: 10.1049/cp:19950832.

[9]    P. V. Jover Rodr´ıguez, A. Belahcen, A. Arkkio, A. Laiho, and J. A. Antonino-Daviu, “Air-gap force distribution and vibration pattern of Induction motors under dynamic eccentricity,” Electr Eng, vol. 90, no. 3, pp. 209–218, Feb. 2008, doi: 10.1007/s00202-007-0066-2.

[10]    J. Balek and P. Klokocˇn´ık, “Development of low-cost inclination sen- sor based on MEMS accelerometers,” IOP Conf. Ser.: Earth Envi- ron. Sci., vol. 906, no. 1, p. 012057, Nov. 2021, doi: 10.1088/1755- 1315/906/1/012057.

[11]    J. Chen, Z. Qi, N. Yang, and W. Luo, “Development and Application of Vibration Comfort Detection Device for Escalator Based on MEMS Acceleration Sensor,” J. Phys.: Conf. Ser., vol. 2366, no. 1, p. 012004, Nov. 2022, doi: 10.1088/1742-6596/2366/1/012004.

[12]    M. H. Richardson, “Is It a Mode Shape, or an Operating Deflection Shape?,” Sound and Vibration, 1997.

[13]    E. J. OBrien, D. P. McCrum, M. A. Khan, and L. J. Prender- gast, “Wavelet-based operating deflection shapes for locating scour- related stiffness losses in multi-span bridges,” Structure and Infras- tructure Engineering, vol. 19, no. 2, pp. 238–253, Feb. 2023, doi: 10.1080/15732479.2021.1937235.

[14]    K. Mendrok, K. Dziedziech, and P. Kurowski, “Detection of structural abnormality of industrial rotary machine using DRS-aided operational modal analysis,” Measurement, vol. 164, p. 108098, Nov. 2020, doi: 10.1016/j.measurement.2020.108098.

[15]    A´ . J. Molina-Viedma, E. Lo´pez-Alba, L. Felipe-Sese´, and F. A. D´ıaz, “Operational Deflection Shape Extraction from Broadband Events of an Aircraft Component Using 3D-DIC in Magnified Images,” Shock and Vibration, vol. 2019, pp. 1–9, Apr. 2019, doi: 10.1155/2019/4039862.

[16]    O. Døssing, “Structural Stroboscopy – Measurement of Operational Deflection Shapes,” Sound and Vibration, vol. 22, no. 8, pp. 18–24, 1988.

[17]    V. S. Gonc¸alves, L. Herzog, and S. C. da S. Ribeiro, “Using ODS vibration techniques to solve problems in motor bearings to maximize compressor service time,” in Proceedings of the EEMODS 2013, Rio de Janeiro, Brazil, 2013.

[18]    V. S. Gonc¸alves, L. H. dos S. Tavares, R. R. H. Bubans, and T. G. C. da Silva, “Reliability of smart sensor in the diagnosis of unbalance and misalignment of electric motors,” in Proceedings of the EEMODS 2022, Stuttgart, Germany, 2022.

[19]    Sridhar, S., Misra, P., Gill, G. & Warrior, J. Cheepsync: a time synchro- nization service for resource constrained bluetooth le advertisers. IEEE Communications Magazine. 54, 136-143 (2016)

[20]    Dian, F., Yousefi, A. & Somaratne, K. A study in accuracy of time synchronization of BLE devices using connection-based event. 2017 8th IEEE Annual Information Technology, Electronics And Mobile Communication Conference (IEMCON). pp. 595-601 (2017)

[1]      Asgarian, F. & Najafi, K. BlueSync: Time Synchronization in Bluetooth Low Energy With Energy-Efficient Calculations. IEEE Internet Of Things Journal. 9, 8633-8645 (2022)