A review of ultrasonic sensing and machine learning methods to monitor industrial processes

Response surface methodology and machine learning based tensile strength prediction in ultrasonic assisted coating of poly lactic acid bone plates manufactured using fused deposition modeling

Poly Lactic Acid (PLA) based bone plates fabricated using Fused Deposition Modeling have poor mechanical strength which can be improved by biocompatible polydopamine (PDM) coating. However, PDM particles, being heavy in nature, settle at the container bottom with increase in coating solution concentration at the time of bone plate coating using dip coating technique. Thus, the present work aims to witness the effect of ultrasonic assisted coating parameters on tensile strength of coated bone plates. The coating parameters involving power of ultrasonic vibrations, coating solution concentration and immersion time were varied. The standard Response Surface Methodology (RSM) was applied and experimental trials were performed for obtaining tensile strength of bone plates under varied coating parameters. The objective of the present study was to compare the values of tensile strength predicted using RSM and machine learning (ML) models. Based on the obtained experimental values, gradient boosting regression (GBReg), linear regression (LReg) and random forest regression (RFReg) were trained and tested for predicting tensile strength of bone plates. The accuracy and prediction errors corresponding to RSM and ML based models were compared with respect to R2, Mean Squared Error (MSE), Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The findings revealed that GBReg exhibited R2, MSE, RMSE and MAE values as 0.9312, 1.7142, 1.2877 and 1.0861 respectively, while RSM showed R2, MSE, RMSE and MAE values as 0.882, 2.13, 1.4595 and 1.258 respectively. RSM model has shown minimum accuracy with high prediction errors amongst the four models. GBReg has outperformed other ML models in terms of their accuracy and error metrics. The present study therefore suggests the application of GBReg based ML model for predicting tensile strength of PDM coated bone plates in response to its accurate and robust prediction performance.

Laser ultrasonics for nondestructive testing of composite materials and structures: A review

This paper presents a comprehensive overview of Laser Ultrasonic Testing (LUT) and its applications in composite materials. The working principles of LUT are thoroughly explained, and an assessment of its advantages and drawbacks is provided. The mechanisms of wave generation and detection are described, along with their influence on the capabilities and limitations of LUT. The paper includes an inclusive overview of each LUT application in composite materials, highlighting their potential, challenges, and research gaps. LUT is a noncontact and nondestructive technique that utilizes lasers to generate and detect ultrasonic waves, with the material itself acting as an emitting transducer. This unique noncontact approach offers an accurate, versatile, convenient, and rapid method for inspecting and characterizing materials. However, some challenges and research gaps have hindered its widespread adoption. One significant challenge in LUT is the low signal-to-noise ratio (SNR), which becomes more pronounced in composite materials due to their low ablation threshold and high wave attenuation. Furthermore, the characterization and inspection of composite materials are more intricate due to their anisotropy and complex damage patterns. Despite these challenges, the combination of ultrasonic waves capable of characterizing and inspecting materials, coupled with the capabilities of lasers and optics for noncontact and real-time operation, presents a promising outlook for the widespread implementation of LUT in Smart Industries and harsh industrial environments, including those with high temperatures, high pressures, or radioactive conditions. This paper contributes to the understanding of LUT's potential and limitations, paving the way for further advancements in its applications.

Study on ultrasonic quantitative evaluation technique based on BP neural network and D-S evidence theory

Ultrasonic detection technology is widely used because of its high sensitivity, strong penetrating ability, accurate defect location, simple operation, and harmlessness to the human body. However, it is still challenging to locate and quantify the defects whose shapes are complex based on ultrasonic testing. The amount of data required for ultrasonic imaging is relatively large, and the efficiency is relatively low. This paper proposes a new method that combines the BP neural network and D-S evidence theory fusion technology with the pulse reflection method. The circular and triangular defects are selected for numerical simulation and experimental testing. The diameter range of the circle is 1-5 mm. The base range of the isosceles triangle is 4-5 mm, and the height range is 3-5 mm. Finally, this paper researches the inversion imaging of single and multiple defects using neural networks, image processing technology and data fusion technology. The results show that after fusion, the similarity coefficient of defects can reach 0.96, and the minimum area error can reach 1.2 %. The maximum error of the average centroid x is only 9.25 %, and the minimum error is 6.36 %. The centroid y error is less than 12 %. The average centroid y error is only 8.73 % at the maximum and 5.09 % at the minimum, indicating that the defect-inversion is relatively accurate.

Self-compensation methodology for ultrasonic thickness gauges

There are many causes for the reduction of the thickness in pipelines, tanks and other mechanical structures. Corrosion, erosion, and abrasive wear cause degradation of mechanical structures and decrease their lifespan. These can be very slow processes that are difficult to track over time. Thickness gauging monitoring is commonly used as a way of preventive maintenance. The pulse-echo ultrasound can be a suitable technique to measure the thickness diminution in industrial facilities. Although ultrasound is considered a robust technique, in this particular application it presents two main difficulties: the mechanical stability of the assembly and the variation of the ultrasonic speed over time. Both mechanical assembly and acoustic propagation speed are strongly influenced by the temperature. In this paper, the implementation of a methodology that compensates for the temperature influences on the ultrasonic speed and the mechanical assembly is presented. The methodology can be applied in metallic structures to evaluate corrosion over long time periods. The temperature compensation data is obtained from the analysis of the ultrasonic signals. In this sense, the method can be called self-compensated. As initial data for the determination of thickness changes, the ultrasonic speed in the material at a reference temperature must be known. All results are evaluated at this temperature. An analysis of the uncertainty sources and limitations of the methodology is also included. To show the experimental application of the proposed technique, a rigid sample was designed in order to avoid mechanical instability. The results show that the methodology can compensate for the temperature, detecting a thickness reduction in the order of a few micrometers.

Ultrasonic Wave Mode-Based Application for Contactless Density Measurement of Highly Aerated Batters

An ultrasonic wave mode-based method for density measurement in highly foamed batters was developed. Therefore, a non-contact ultrasonic sensor system was designed to generate signals for batch-wise processes. An ultrasonic sensor, containing a piezoelectric ceramic at the fundamental longitudinal frequency of 2 MHz, was used to take impedance measurements in pulse-echo mode. The ultrasonic signals were processed and analysed wave-mode wise, using a feature-driven approach. The measurements were carried out for different mixing times within a container, with the attached ultrasonic sensor. Within the biscuit batter, the change to the ultrasonic signals caused by density changes during the batter-mixing process was monitored (R² = 0.96). The density range detected by the sensor ranges between 500 g/L and 1000 g/L. The ultrasonic sensor system developed also shows a reasonable level of accuracy for the measurements of biscuit batter variations (R² > 0.94). The main benefit of this novel technique, which comprises multiple wave modes for signal features and combines these features with the relevant process parameters, leads to a more robust system as regards to multiple interference factors.

Machine learning for ultrasonic nondestructive examination of welding defects: A systematic review

Recent years have seen a substantial increase in the application of machine learning (ML) for automated analysis of nondestructive examination (NDE) data. One of the applications of interest is the use of ML for the analysis of data from in-service inspection of welds in nuclear power and other industries. These types of inspections are performed in accordance with criteria described in the ASME Boiler and Pressure Vessel Code and require the use of reliable NDE techniques. The rapid growth in ML methods and the diversity of possible approaches indicate a need to assess the current capabilities of ML and automated data analysis for NDE and identify any gaps or shortcomings in current ML technologies as applied to the automated analysis of NDE data. In particular, there is a need to determine the impact of ML on the NDE reliability. This paper discusses the findings from a literature survey on the current state of ML for the automated analysis of data from ultrasonic NDE of weld flaws. It discusses an overview of ultrasonic NDE as used for weld inspections in nuclear power and other industries. Data sets and ML models used in the literature are summarized, along with a generally applicable workflow for ML. Findings on the capabilities, limitations and potential gaps in feature selection, data selection, and ML model optimization are discussed. The paper identified several needs for quantifying and validating the performance of ML methods for ultrasonic NDE, including the need for common data sets.

A novel ultrasonic inspection method of the heat exchangers based on circumferential waves and deep neural networks

The heat exchanger (HE) is an important component of almost every energy generation system. Periodic inspection of the HEs is particularly important to keep high efficiency of the entire system. In this paper, a novel ultrasonic water immersion inspection method is presented based on circumferential wave (CW) propagation to detect defective HE. Thin patch-type piezoelectric elements with multiple resonance frequencies were adopted for the ultrasonic inspection of narrow-spaced HE in an immersion test. Water-filled HE was used to simulate defective HE because water is the most reliable indicator of the defect. The HE will leak water no matter what the defect pattern is. Furthermore, continuous wavelet transform (CWT) was used to investigate the received CW, and inverse CWT was applied to separate frequency bands corresponding to the thickness and lateral resonance modes of the piezoelectric element. Different arrangements of intact and leaky HE were tested with several pairs of thin piezoelectric patch probes in various instrumental setups. Also, direct waveforms in the water without HE were used as reference signals, to indicate instrumental gain and probe sensitivity. Moreover, all filtered CW corresponding to resonance modes together with the direct waveforms in the water were used to train the deep neural networks (DNNs). As a result, an automatic HE state classification method was obtained, and the accuracy of the applied DNN was estimated as 99.99%.

AZ31B magnesium alloy matching layer for Lens-focused piezoelectric transducer application

Compared with planar transducers, focused transducers have higher ultrasound intensity and better lateral resolution in the focal zone. At present, the matching layer materials for focused transducers are mainly 0-3 composite materials, which have problems such as non-uniformity, difficulty to fabricate at high frequencies, and large sound attenuation. In this paper, finite element analysis is carried out to simulate lens-focused transducers with different matching layer structures and materials. It is found that the focused transducer with magnesium alloy matching layer has the best comprehensive performance. A lens-focused PZT-5H ultrasonic transducer was then fabricated with AZ31B magnesium alloy as the first matching layer. The measured results show that the center frequency of the transducer is 4.38 MHz, the -6-dB bandwidth is 68.35 % and the insertion loss is -13.88 dB. Benefiting from the high uniformity, high acoustic impedance and extremely low acoustic attenuation of magnesium alloy, the transducers in this research exhibit superior performances than other reported transducers with conventional matching layer. The current work suggests that AZ31B magnesium alloy is a promising matching layer material for ultrasonic transducers.

Domain Adaptation for In-Line Allergen Classification of Agri-Food Powders Using Near-Infrared Spectroscopy

The addition of incorrect agri-food powders to a production line due to human error is a large safety concern in food and drink manufacturing, owing to incorporation of allergens in the final product. This work combines near-infrared spectroscopy with machine-learning models for early detection of this problem. Specifically, domain adaptation is used to transfer models from spectra acquired under stationary conditions to moving samples, thereby minimizing the volume of labelled data required to collect on a production line. Two deep-learning domain-adaptation methodologies are used: domain-adversarial neural networks and semisupervised generative adversarial neural networks. Overall, accuracy of up to 96.0% was achieved using no labelled data from the target domain moving spectra, and up to 99.68% was achieved when incorporating a single labelled data instance for each material into model training. Using both domain-adaptation methodologies together achieved the highest prediction accuracies on average, as did combining measurements from two near-infrared spectroscopy sensors with different wavelength ranges. Ensemble methods were used to further increase model accuracy and provide quantification of model uncertainty, and a feature-permutation method was used for global interpretability of the models.