Метод ультразвукового эхо-импульсного контроля для обнаружения и определения продольного и поперечного положения дефектов стальных трубопроводов с использованием интерферометров Фабри-Перо на основе двух чирпированных волоконных брэгговских решеток тема диссертации и автореферата по ВАК РФ 00.00.00, кандидат наук Сычева Софья Дмитриевна

  • Сычева Софья Дмитриевна
  • кандидат науккандидат наук
  • 2025, «Национальный исследовательский университет ИТМО»
  • Специальность ВАК РФ00.00.00
  • Количество страниц 290
Сычева Софья Дмитриевна. Метод ультразвукового эхо-импульсного контроля для обнаружения и определения продольного и поперечного положения дефектов стальных трубопроводов с использованием интерферометров Фабри-Перо на основе двух чирпированных волоконных брэгговских решеток: дис. кандидат наук: 00.00.00 - Другие cпециальности. «Национальный исследовательский университет ИТМО». 2025. 290 с.

Оглавление диссертации кандидат наук Сычева Софья Дмитриевна

РЕФЕРАТ

SYNOPSIS

ВВЕДЕНИЕ

ГЛАВА 1. ОБЗОР СОВРЕМЕННОЙ ЛИТЕРАТУРЫ

1.1 Промышленные стальные трубопроводы

1.1.1 Общая информация

1.1.2 Несплошности основного металла трубопроводов

1.2 Методы неразрушающего контроля для оценки состояния основного металла трубопроводов

1.2.1 Магнитные методы

1.2.2 Вихретоковые методы

1.2.3 Оптические (визуальные) методы

1.2.4 Акустические методы

1.3 Ультразвуковые дефектоскопические системы

1.3.1 Классические ультразвуковые дефектоскопы

1.3.2 Ультразвуковые дефектоскопические системы на основе фазированных решеток

1.3.3 Ультразвуковые дефектоскопические системы на основе направленной волны

1.3.4 Ультразвуковые дефектоскопические системы с применением волоконных датчиков

1.4 Волоконно-оптические акустические датчики

1.4.1 Акустические датчики на основе брэгговских решеток

1.4.2 Интерферометрические акустические датчики

1.4.3 Распределенное акустическое зондирование

1.5 Алгоритмы обработки данных с датчиков

1.5.1 Алгоритмы определения положения несплошностей

1.5.2 Алгоритмы акустической томографии

1.5.3 Алгоритмы классификации

Выводы по главе

ГЛАВА 2. ТЕОРЕТИЧЕСКОЕ ИССЛЕДОВАНИЕ КОНФИГУРАЦИИ УЛЬТРАЗВУКОВОЙ ДЕФЕКТОСКОПИЧЕСКОЙ СИСТЕМЫ НА ОСНОВЕ ЭХО-МЕТОДА

2.1 Влияние резонансной частоты излучателя на разрешение ультразвуковой дефектоскопической системы

2.2 Разработка программного инструмента для исследования конфигурации ультразвуковой дефектоскопической системы

2.3 Разработка алгоритма обнаружения и определения положения несплошностей

2.4 Разработка алгоритма классификации несплошностей

2.5 Результаты исследования алгоритма обнаружения и определения положения несплошностей металла в предложенной конфигурации ультразвуковой системы

2.6 Результаты исследования алгоритма классификации несплошностей

в предложенной конфигурации ультразвуковой системы

Выводы по главе

ГЛАВА 3. ЭКСПЕРИМЕНТАЛЬНОЕ ИССЛЕДОВАНИЕ ПРЕДЛОЖЕННОЙ КОНФИГУРАЦИИ УЛЬТРАЗВУКОВОЙ

ДЕФЕКТОСКОПИЧЕСКОЙ СИСТЕМЫ НА ОСНОВЕ ЭХО-МЕТОДА

3.1 Особенности измерения оптической разности хода лучей волоконным интерферометром Фабри-Перо

3.2 Описание предложенной конфигурации ультразвуковой системы на основе эхо-метода

3.3 Согласование параметров оптико-электронной системы и датчиков на основе интерферометров Фабри-Перо

3.4 Описание функциональных возможностей разработанной компьютерной программы акустического картографирования

3.5 Результаты экспериментального исследования по оценке несплошностей металла стальной пластины предложенным методом

3.6 Разработка и исследование способа крепления чувствительного элемента датчика к трубопроводу

3.7 Разработка структуры исследовательского образца ультразвуковой

системы измерения с предложенным вариантом крепления

3.8 Результаты экспериментального исследования несплошносей металла стальных труб предложенным методом

3.9 Пути совершенствования предложенной конфигурации системы... 224 Выводы по главе

ЗАКЛЮЧЕНИЕ

СОКРАЩЕНИЯ И ОБОЗНАЧЕНИЯ

СПИСОК ЛИТЕРАТУРЫ

ПРИЛОЖЕНИЯ

Приложение 1. Таблицы

Приложение 2. Охранные документы на результаты интеллектуальной деятельности

Приложение 3. Тексты публикаций

РЕФЕРАТ

Рекомендованный список диссертаций по специальности «Другие cпециальности», 00.00.00 шифр ВАК

Введение диссертации (часть автореферата) на тему «Метод ультразвукового эхо-импульсного контроля для обнаружения и определения продольного и поперечного положения дефектов стальных трубопроводов с использованием интерферометров Фабри-Перо на основе двух чирпированных волоконных брэгговских решеток»

Актуальность темы

В нефтяной и газовой промышленностях для обеспечения надежности и безопасности эксплуатации трубопроводных систем [1] применяются преимущественно неразрушающие методы контроля [21-27]. Их использование позволяет производить анализ уровня износа основного металла и оставшегося срока службы трубопровода за счет выявления и оценки характеристик обнаруженных дефектов (несплошностей). Одним из наиболее ценных преимуществ такого класса методов является возможность диагностики текущего состояния трубопроводных систем без нарушения их целостности [2,21]. Из широкого спектра методов неразрушающего контроля наибольшую известность для задачи осуществления дефектоскопии основного металла трубопроводов приобрели магнитные [29-38], вихретоковые [39-45], акустические [51-75] и оптические [46-50] системы. При этом системы на основе вихретоковых и оптических методов находят широкое применение в качестве дополнительных инструментов дефектоскопии основного металла трубопроводов. Использование таких систем в качестве дополнительных обусловлено тем, что дефектоскопические системы, опирающиеся на данные виды методов, обладают ограниченной глубиной проникновения в материал, что существенно затрудняет выявление внутренних дефектов трубопроводных систем. В то же время магнитные и акустические дефектоскопы обладают значительной глубиной проникновения, поэтому применяются как самостоятельные приборы для исследования и оценки текущего уровня износа эксплуатируемых трубопроводных систем.

Наиболее информативным прибором, построенным на основе магнитного индукционного метода и предназначенным для технической

диагностики состояния основного металла трубопроводов, считается магнитный дефектоскоп, осуществляющий внутритрубную инспекцию. Принцип работы такого метода, основанного на процессе намагничивания трубопровода с последующей регистрацией магнитных полей с помощью магнитной капсулы, последовательно продвигающейся внутри трубопровода. Таким способом диагностируют наиболее важные и протяженные участки магистральных стальных нефтегазопроводов, имеющих внешний диаметр больше 100 мм и общую протяженность в интервале 0,1-100 км. В то же время существуют участки трубопроводов, имеющие протяжённость менее 0,1 км, которые не могут быть диагностированы описанным способом. Общие ограничения, затрудняющие применение систем на основе магнитных методов для технической диагностики состояния основного металла трубопроводов, включают в себя: необходимость соответствия геометрических размеров диагностической капсулы и диаметра трубопровода, наличие достаточной длины исследуемого участка для измерения, требование к наличию камер для запуска и выпуска магнитного дефектоскопа в трубопровод. Преодолеть описанные ограничения позволяют системы на основе акустических методов [51-75].

В частности, базовый ультразвуковой метод отраженного излучения (эхо-метод) является одним из наиболее эффективных представителей методов ультразвукового контроля состояния стальных трубопроводов. Ультразвуковые системы на основе эхо-метода обладают высокой чувствительностью и точностью определения положения, а также способны выявлять различные виды несплошностей металла на эксплуатируемых трубопроводах, включая трещины и коррозионные повреждения [62]. Принцип его работы основан на генерации и приеме ультразвуковых волн, отраженных от несплошностей и границ раздела сред. Недостатками систем

на основе данного метода являются необходимость высокой квалификации оператора для корректной интерпретации результатов и малая область покрытия, что приводит к увеличенному времени проведения диагностики на эксплуатируемых трубопроводах.

В то же время системы, построенные на основе фазированных акустических решеток [65-66], отличаются возможностью восстанавливать цифровое изображение обнаруженных несплошностей за счет применения специализированных алгоритмов цифровой обработки сигналов датчиков, что повышает информативность результатов проведения диагностики состояния основного металла трубопровода. При этом для реализации данного класса систем наиболее часто используется массив пьезоэлектрических преобразователей (типичная конфигурация с 16-64 единицами) и сложный комплекс алгоритмов обработки сигналов, что приводит к значительному увеличению стоимости дефектоскопического оборудования. Кроме того, в системах используются высокочастотные пьезоэлектрические преобразователи, обладающие возможностью управления пучком ультразвуковых волн. Использование высокой частоты излучателя существенно ограничивает протяженность обследуемого участка трубопровода. При этом более низкая рабочая частота позволяет ультразвуковым волнам распространяться на расстояния порядка сотен метров. Таким образом, ультразвуковые системы на основе фазированных решеток обеспечивают повышенную информативность данных о состоянии металла трубопровода, но в то же время площадь области, контролируемой таким дефектоскопом зачастую, не превышает 0,02 м2, что сказывается на увеличении времени проведения дефектоскопии металла трубопровода.

Системы, основанные на технологии направленной ультразвуковой волны [68-71], позволяют проводить измерения с минимальным затрачиваемым временем и охватывать при этом участки трубопроводов

протяженностью до 200 метров. Основным недостатком таких систем является низкое пространственное разрешение, вызванное использованием ультразвуковых излучателей с частотой 50-200 кГц, которое ограничивает минимально различимый размер дефекта. Таким образом, снижение затрачиваемого оператором времени на проведение диагностики и увеличение максимальной протяженности диагностируемого участка трубопровода до нескольких сотен метров достигаются за счет снижения разрешающей способности по размеру несплошностей основного металла трубопровода.

Современные ультразвуковые дефектоскопические системы для оценки состояния металла трубопровода могут обеспечивать высокую точность определения положения дефектов, могут позволять оценивать их форму и размеры, а также могут обеспечивать минимальные временные затраты на проведение диагностики. Однако ни одна из них не способна обеспечить все преимущества одновременно. Также важно отметить, что на рынке ультразвуковых дефектоскопических систем отсутствуют решения, обеспечивающие область покрытия диагностируемого участка металла трубопровода в диапазоне от 0,02 до 1 м2 без необходимости перемещения излучателя.

Дополнительным ограничением является уязвимость ультразвуковых дефектоскопических систем к электромагнитным помехам, присутствующим в сложных технических условиях (вблизи трансформаторных подстанций или промышленного

электрооборудования). Одним из подходов уменьшению восприимчивости системы к электромагнитным помехам является применение волоконных датчиков [78, 80]. В частности, датчики на основе волоконно-оптических интерферометров Фабри-Перо, используемые для регистрации ультразвуковых волн, характеризуются широким спектральным

диапазоном измеряемых частот (1 Гц - 1 МГц), широким диапазоном измеряемых давлений (от 500 Па и выше), обеспечивают близкую к линейной зависимость оптического сигнала от акустического давления [93]. Волоконно-оптические датчики на основе интерферометров Фабри-Перо за счет имеющихся преимуществ могут расширить область применения ультразвуковых дефектоскопических систем, повысить вероятность обнаружения и точность определения положения несплошностей основного металла трубопровода. Таким образом, в настоящей диссертационной работе предложено применение волоконно-оптических датчиков на основе интерферометров Фабри-Перо в составе ультразвуковой дефектоскопической системы для преодоления ограничений и усовершенствовании следующих параметров: повышение устойчивости к внешнему электромагнитному излучению, расширение области покрытия системы и минимизации временных затрат на проведение дефектоскопии, увеличение информативности ее результата.

Целью настоящей работы является разработка и исследование системы на основе ультразвукового эхо-метода и волоконно-оптических интерферометров Фабри-Перо, обладающей повышенной информативностью при обнаружении и определении положения несплошностей металла и сокращенной длительностью процесса проведения технической диагностики состояния металла в стальных трубопроводах.

Для достижения данной цели в рамках диссертационного исследования были поставлены и решены следующие задачи:

1. Выполнить аналитический обзор по направлениям: характеристики промышленных трубопроводов; типы несплошностей основного металла трубопровода; методы и системы ультразвукового контроля, предназначенные для осуществления дефектоскопии основного металла

трубопроводов; анализ существующих вариантов волоконных акустических датчиков; анализ существующих алгоритмов для задач обнаружения, определения положения и классификации выявленных несплошностей металла трубопровода.

2. Разработать и исследовать метод моделирования ультразвуковой дефектоскопической системы трубопроводов в конфигурации с одним излучателем и несколькими датчиками.

3. Разработать и исследовать алгоритм обработки данных с датчиков в конфигурации с одним излучателем и несколькими датчиками.

4. Разработать структурную схему и осуществить реализацию ультразвуковой системы с одним излучателем и несколькими датчиками.

5. Разработать и исследовать конструкцию крепления чувствительного элемента датчика к трубопроводу.

6. Экспериментально исследовать применимость предложенной конфигурации ультразвуковой дефектоскопической системы для выявления и определения пространственного положения несплошностей на участках стальных труб.

7. Разработать и исследовать алгоритм классификации несплошностей основного металла трубопровода, выявленных с помощью предложенного метода.

Методы исследования

Глубина анализа современных источников литературы в рамках исследуемой тематики составила 10 лет с целью достижения достаточного уровня актуальности рассматриваемых данных. Моделирование распространения ультразвуковых волн, алгоритма обработки данных с датчиков, а также вспомогательных алгоритмов обработки цифровых сигналов, анализ и визуализация результатов обработки данных были выполнены в программной среде МаНаЬ. Предложенные решения в области

обработки сигналов с датчиков базируются на основе проверенных и широко применяемых методов (времени пролета сигнала, синтезируемой фокусируемой апертуры и обработки изображений). Экспериментальные исследования выполнены с применением измерительного оборудования 2-го класса точности. Полученные экспериментальные результаты и результаты моделирования имеют близкий порядок величин.

Научные положения, выносимые на защиту

1. Разработанный программный инструмент на основе предложенного метода моделирования работы ультразвуковой дефектоскопической системы, содержащей один излучатель и несколько датчиков, включающего этап расчета распространения ультразвуковых волн в металле с заданными параметрами несплошностей металла (количество, плотность, геометрические форма и размеры, пространственное положение) с учетом частоты и расположения излучателя, выполняемый с использованием программного пакета «k-wave toolbox» (Matlab), этап расчета сигналов датчиков с учетом их пространственного расположения и этап обработки сигналов датчиков посредством используемого алгоритма, позволяет моделировать восстанавливаемое системой двумерное изображение несплошностей металла и рассчитывать ключевые характеристики системы (вероятность обнаружения и погрешность определения пространственного положения несплошностей), что предоставляет возможность обоснованно подбирать параметры ультразвуковой дефектоскопической системы и алгоритма обработки сигналов датчиков с учетом предъявляемых требований к системе.

2. Разработан алгоритм пространственно-временной обработки сигналов датчиков ультразвуковой дефектоскопической системы, содержащей один излучатель и несколько локальных датчиков,

включающий этап построения двумерного изображения поверхностного и подповерхностного слоев металла за счет предложенной версии метода синтезируемой апертуры на основе технологии «ЬеатАэгт^» с добавлением вспомогательных алгоритмов гамма-коррекции и морфологического замыкания и этап обработки сформированного изображения путем применения алгоритмов бинаризации, сегментации и поиска центра масс, позволяющий определить наличие несплошностей металла и рассчитать их пространственные положения, что подтверждается результатами моделирования и экспериментального исследования для случая ультразвуковой системы с четырьмя локальными датчиками на пластине размером 0,16 м2.

3. Предложенная конфигурация ультразвуковой дефектоскопической системы, содержащая излучатель частотой 156 кГц, оптико-электронную систему с четырьмя датчиками на основе интерферометров Фабри-Перо, образованных парой чирпированных волоконных решеток Брэгга и разработанного варианта крепления чувствительного элемента, в сочетании с предложенным алгоритмом обработки сигналов датчиков, позволяет восстанавливать двумерное изображение поверхностного и подповерхностного слоев металла в области размером 0,06 м2 на участке стальной трубы длиной 500 мм, диаметром 108 мм и толщиной стенки 3,5 мм, а также определять наличие несплошностей металла с вероятностью обнаружения 80% и средней погрешностью определения пространственного положения несплошностей 27 мм, что демонстрирует большую область покрытия, чем у известных ультразвуковых систем на основе фазированных решеток.

4. Предложенный алгоритм классификации двумерных изображений несплошностей металла, основанный на архитектуре сверточной нейронной сети с модулем параллельных вычислений и примененный для

анализа восстановленных изображений, рассчитанных с помощью разработанного программного инструмента, позволяет определить соответствие обнаруженных несплошностей металла в стальных трубах одному из пяти классов: каверна, коррозия, трещина, смешанный дефект и шум (отсутствие дефекта), при этом достигается точность классификации 98%, что превосходит значения точности классификации, полученные с помощью наиболее распространенных сверточных нейронных сетей: «GoogleNet», <^е№^5», «ResNet-18», «ResNet-101», «MobileNet у2», - не менее чем на 1%.

Научная новизна диссертационного исследования представлена следующими пунктами:

1. Впервые предложен и продемонстрирован метод моделирования работы ультразвуковой дефектоскопической системы, содержащей один излучатель и несколько датчиков, основанный на расчете распространения ультразвуковых волн в среде с заданными параметрами несплошностей (количество, плотность, геометрические форма и размеры, пространственное положение) с учетом частоты и расположения излучателя, сигналов датчиков с учетом их пространственного расположения и обработке этих сигналов посредством используемого алгоритма, позволяющий моделировать восстанавливаемое системой двумерное изображение несплошностей в металле и рассчитывать теоретические значения вероятности обнаружения и погрешности определения пространственного положения несплошностей.

2. Впервые предложен и продемонстрирован алгоритм пространственно-временной обработки сигналов датчиков ультразвуковой дефектоскопической системы, содержащей один излучатель и несколько датчиков, основанный на предложенной версии метода синтезируемой апертуры на основе технологии «Ьеатйэгт^» с добавлением

вспомогательных алгоритмов гамма-коррекции и морфологического замыкания для построения двумерного изображения поверхностного и подповерхностного слоев металла и добавлении алгоритмов бинаризации, сегментации и поиска центра масс для определения наличия несплошностей металла и расчета их пространственных положений.

3. Впервые предложена и продемонстрирована конфигурация ультразвуковой дефектоскопической системы, содержащая излучатель частотой 156 кГц, оптико-электронную систему с четырьмя датчиками на основе интерферометров Фабри-Перо, образованных парой чирпированных волоконных решеток Брэгга и разработанного варианта крепления чувствительного элемента, позволяющая восстанавливать двумерное изображение поверхностного и подповерхностного слоев металла, а также производить на основе сформированного системой изображение выявление и определение положения несплошностей металла.

4. Впервые предложен и продемонстрирован алгоритм классификации двумерных изображений несплошностей металла, основанный на архитектуре сверточной нейронной сети с одним модулем параллельных вычислений, позволяющий определять соответствие обнаруженных несплошностей металла в стальных трубах одному из пяти классов: каверна, коррозия, трещина, смешанный дефект и шум (отсутствие дефекта), достигая точности классификации, превосходящей значения известных сверточных нейронных сетей не менее чем на 1%.

Научно-техническая задача, решаемая в диссертации, заключается в исследовании возможности создания ультразвуковой дефектоскопической системы с одним излучателем и несколькими волоконно-оптическими датчиками на основе интерферометров Фабри-Перо, способной осуществлять обнаружение и определение пространственного положения несплошностей в поверхностном и

подповерхностном слоях металла с областью покрытия участка стального трубопровода более 0,02 м2 и дополнительной возможностью определения вероятности соответствия выявленной несплошности одному из пяти классов дефектов: каверна, коррозия, трещина, смешанный дефект и шум (отсутствие дефекта).

Объектом исследования являются ультразвуковые методы и системы на их основе, направленные на задачу дефектоскопии основного металла трубопровода.

Предметом исследования является конфигурация ультразвуковой дефектоскопической системы на основе метода ультразвукового эхо-импульсного контроля с одним излучателем и несколькими волоконно-оптическими датчиками на основе интерферометров Фабри-Перо.

Практическая значимость работы состоит в том, что в ней:

1. Разработанный программный инструмент может быть применен для обоснованного выбора оптимальных характеристик ультразвуковой дефектоскопической системы с учетом предъявляемых к ней требований, а также оценки применяемых алгоритмических решений и формировании объемного набора данных для обучения нейронных сетей для классификации образов несплошностей.

2. Разработанный алгоритм обработки сигналов датчиков может быть применен для определения наличия несплошностей поверхностного и подповерхностного слоев металла и расчета их пространственных положений в конфигурации ультразвуковой дефектоскопической системы, содержащей один излучатель и несколько локальных датчиков.

3. Предложенная конфигурация ультразвуковой дефектоскопической системы, содержащая излучатель частотой 156 кГц, оптико-электронную систему с четырьмя датчиками на основе интерферометров Фабри-Перо, образованных парой чирпированных волоконных решеток Брэгга и

разработанного варианта крепления чувствительного элемента, в сочетании с предложенным алгоритмом обработки сигналов датчиков, способна обеспечивать большую область покрытия, чем у известных ультразвуковых систем на основе фазированных решеток.

4. Предложенный алгоритм классификации восстановленных изображений несплошностей металла, может быть применен для определения соответствия выявленных предложенным методом несплошностей одному из пяти классов: каверна, коррозия, трещина, смешанный дефект и шум (отсутствие дефекта) и может служить вспомогательным программным инструментом оператора, осуществляющего дефектоскопию основного металла трубопровода, повышая информативность результат технической диагностики.

Достоверность полученных в рамках выполнения диссертационного исследования результатов обусловлена представленным объемом теоретических исследований; использованием современных версий программного обеспечения для моделирования и обработки данных (программная среда Matlab версии V.2024A); применением оборудования высокого класса точности; представленным сравнением результатов вычислений, полученных в ходе моделирования с результатами экспериментального исследования; апробацией достигнутых результатов на международных и всероссийских конференциях; публикацией достигнутых результатов в отечественных научных журналах, входящих в перечень ВАК и зарубежных журналах, входящих в базы цитирований SCOPUS и WEB OF SCIENCE.

Внедрение

Получено свидетельство о государственной регистрации программы для ЭВМ №2024691994, осуществляющей обработку данных с предложенной конфигурации ультразвуковой дефектоскопической

системы. Кроме того, в рамках выполнения НИОКТР .№423055 «Разработка автоматизированной акусто-оптической системы диагностики энергетического оборудования с применением технологий искусственного интеллекта», выполненного университетом ИТМО под руководством соискателя при консультативной поддержке представителей компании ПАО «Газпромнефть», был разработан исследовательский образец системы на основе настоящего диссертационного исследования.

Апробация

Основные результаты диссертационной работы докладывались и обсуждались на 7 научных конференциях, в том числе 3 международного уровня. Соискатель принял участие в следующих конференциях: LI, LII, LIII и LIV Учебная и научно-методическая конференция университета ИТМО (февраль 2022 г., 2023 г., 2024 г., 2025; г. Санкт-Петербург, Россия); Sino-Russian Advanced Photonics and Optics Seminar (27-28 ноября 2022; г. Харбин, Китай); XIII Международная конференция "Фотоника и информационная оптика" (24-26 января 2024 г.; г. Москва, Россия); XXI Международная конференция по голографии и прикладным оптическим технологиям ХОЛОЭКСПО (9-12 сентября 2024 г.; г. Казань, Россия).

Личный вклад автора

Результаты, изложенные в тексте настоящей диссертационной работы, были получены автором лично или при его непосредственном участии. Соискатель активно участвовал во всех этапах исследования, выполняя работы самостоятельно при консультативной поддержке научного руководителя и сотрудников университета ИТМО.

Благодарности

Автор выражает отдельную благодарность научному руководителю работы Дейнеке Ивану Геннадьевичу за организационную, методическую и консультативную поддержку в процессе выполнения диссертационного

исследования. Официальным оппонентам настоящей работы, членам диссертационного совета 06.22.00 и коллегам из подразделений «Высшая Инженерно-Техническая Школа» и «Факультета систем управления и робототехники» университета ИТМО за обсуждение результатов и консультативную поддержку в процессе выполнения настоящей диссертационной работы. Научным сотрудникам «кафедры электроакустики и ультразвуковой техники» университета СПбГЭТУ ЛЭТИ за предоставление дополнительного измерительного оборудования. За осуществление технической и консультативной поддержки Ахмерову Артему Харисовичу по части моделирования распространения ультразвуковых волн и анализа результатов исследования известных алгоритмов нейронных сверточных сетей, Кузнецову Владимиру Николаевичу по части разработки проекта управления системой в ПЛИС, Волковскому Сергею Александровичу по части разработки оптико-электронной системы опроса датчиков и Ходжимухамедову Озару Исломовичу по части разработки компьютерной программы «Vision» на основе предложенной автором обработки данных с датчиков, а также научным сотрудникам, профессорам и инженерам «Научно-Исследовательского Центра Световодной Фотоники» университета ИТМО.

Работа выполнена в рамках государственного задания Министерства науки и высшего образования Российской Федерации (проект № FSER-2024-0006).

Публикации

Основное содержание диссертационного исследования опубликовано в 4 научных изданиях (2 - индексируются в Scopus, 1 - входит в перечень ВАК, 1 - охранный документ на результат интеллектуальной деятельности).

Публикации в международных изданиях, индексируемых в базах данных Scopus/Web of Science:

1 - Akhmerov A.K., Sycheva S.D., Khodzhimukhamedov O.I., Deyneka I.G., Vasilev A.S., Kulikov A.V. Acoustic energy mapping method for acousto-optical non-destructive testing system with automated positioning // Studies in Systems, Decision and Control. 2024. V. 544. pp. 111-125.

2 - Kuncak J., Forinova M., Pilipenco A., Prochazka V., Horak P., Sycheva S.D., Deineka I.G., Vaisocherova-Lisalova H. Automating data classification for label-free point-of-care biosensing in real complex samples // Sensors and Actuators A: Physical. 2024. Vol. 374. pp. 115501.

Публикации в изданиях из перечня ВАК РФ:

1 - Сычева С.Д., Ахмеров А.Х., Дейнека И.Г., Кузнецов В.Н., Пиха Д.В., Волковский С.А., Яндыбаева Ю.И., Дмитриев А.А. Методика определения оптимального рабочего спектрального диапазона интерферометра Фабри-Перо на основе чирпированных волоконных решеток Брэгга // Оптический журнал. 2025. Т. 92. № 7. С. 62-72.

Охранный документ на результат интеллектуальной деятельности:

1 - Ходжимухамедов О.И., Сычева С.Д., Ахмеров А.Х., Дейнека И.Г., Кузнецов В.Н., Пиха Д.В. Программа для управления акусто-оптической системой диагностики трубопроводов Acoustic Vision // Свидетельство о государственной регистрации программы для ЭВМ № 2024691994 от 25.12.2024.

Похожие диссертационные работы по специальности «Другие cпециальности», 00.00.00 шифр ВАК

Список литературы диссертационного исследования кандидат наук Сычева Софья Дмитриевна, 2025 год

СПИСОК источников

1. Habel W.R. Optical fiber methods in nondestructive evaluation. Handbook of Advanced Non-Destructive Evaluation. Springer, 2018. 1-49 p. https://doi.org/ 10.1007/978-3-319-30050-4_39-l

2. Zhang X., Jin J., Jiang J., et al. Acoustic emission sensing in fiber Bragg grating based on phase demodulation for material health detection // Opt. and Lasers Eng. 2023. V. 171. P. 107823. https://doi.org/10.1016/ j .optlaseng. 2023.107823

3. Zhe G., Che J., Wei H., et al. Large dynamic-range fiber Bragg grating sensor system for acoustic emission detection // Appl. Opt. 2021. V. 60. P. 547-5552. http:// doi.org/10.1364/AO.427185

4. Reimer V., Abdalwareth A., Flachenecker G., et al. Non-destructive technique for characterization fiber Bragg gratings via diffraction profile // Photonic Instrumentation Engineering XI. United States, California. March 11, 2024. https://doi.org/10.1117/ 12.3002504

5. Chanu-Rigaldies S., Lecomte P., Ollivier S., et al. Self-mixing interferometer for acoustic measurements through vibrometric calibration // Sensors. 2024. V. 24. P. 1777. https://doi.org/10.3390/s24061777

6. Moradi H., Parvin P., Ojaghloo Al., et al. Ultrasensitive fiber optic Fabry-Perot acoustic sensor using phase detection // Measurement. 2021. V. 172. P. 108953. https://doi.Org/10.1016/j.measurement.2020.108953

7. Ромашко P.B., Башков O.B., Ефимов T.A. и др. Особенности применения адаптивных интерферо-метрических волоконно-оптических датчиков акустической эмиссии для контроля состояния полимерных композиционных материалов // Дефекто-

скопия. 2024. Т. 1. С. 21-27. http://doi.org/10.31857/ S0130308224010023

Romashko R.V., Bashkov O.V., Efimov Т.А., et al. Features of application of adaptive interferometric fiber sensors of acoustic emission to monitor the condition of polymer composite materials [in Russian] // Defek-toskopia. 2024. № 1. P. 21-27. http://doi.org/10.31857/ S0130308224010023

8. Залетин В.В., Савитский О.А., Силников М.В. и др. Акустико-эмиссионная диагностика корпусных конструкций системой интегрированных волоконно-оптических датчиков для обеспечения безопасной эксплуатации самолетов и космических аппаратов [на англ. языке] // Акта Астронавтика. 2024. Т. 229. С. 547-551. https://doi.Org/10.1016/j.actaastro.2024. 02.001

Zaletin V.V., Savitsky О .A., Silnikov M.V., et al. Acoustic emission diagnostics of a hull structures by a system of integrating fiber-optic sensors for the aircraft and spacecraft safe operation // Acta Astronáutica. 2024. V. 229. P. 547-551. https://doi.org/10.1016/ j.actaastro.2024.02.001

9. Bakhtiari G.S., Masoudi A., Brambilla G. Long range Raman-amplified distributed acoustic sensor based on spontaneous Brillouin scattering for large strain sensing // Sensors. 2022. V. 22. P. 2047. https://doi.org/ 10.3390/S22052047

10. Zhang Q., Zhu Yup., Luo X., et al. Acoustic emission sensor system using a chirped fiber-Bragg-grating Fabry-Perot interferometer and smart feedback control // Opt. Lett. 2017. V. 42. P. 631-634. https://doi. org/10.1364/OL .42.000631

11. Tosi D. Review of chirped fiber Bragg grating (CFBG) fiber-optic sensors and their applications // Sensors. 2018. V. 18. P. 2147. https://doi.org/10.3390/sl8072147

12. Rose J.L., Philtron J., LiuG., et al. A hybrid ultrasonic guided wave-fiber optic system for flaw detection in pipe // Appl. Sci. 2018. V. 8. P. 727. https://doi.org/ 10.3390/app8050727

13.Ахмеров A.X., Сычева С.Д., Ходжимухамедов О.И. и др. Метод картирования акустической энергии для акустооптической системы неразрушающего контроля с автоматическим позиционированием [на англ. языке] // Умные электромеханические системы. 2024. Т. 544. С. 111-125. https://doi.org/10.1007/ 978-3-031-64277-7_9

Akhmerov А.К., Sycheva S.D., Khodzhimukhame-dov O.I., et al. Acoustic energy mapping method for acousto-optical non-destructive testing system with automated positioning // Studies in Systems. Decision and Control. 2024. V. 544. P. 111-125. https://doi.org/ 10.1007/978-3-031-64277-7_9

14.Бочкова С.Д., Волковский С.А., Ефимов М.Е. и др. Метод локализации воздействия в композитном материале с использованием волоконно-оптических датчиков акустической эмиссии // Приборы и техника эксперимента. 2020. № 4. С. 73-77. https://doi. org/10.31857/S0032816220040230

Bochkova S.D., Volkovsky S.A., Efimov M.E., et al. A method for determining the locations of impacts in

a composite material using fiber optical acoustic emission sensors // Instruments and Experimental Techniques. 2020. V. 63. P. 507-510. https://doi.org/ 10.1134/S0020441220040235

15. Ефимов М.Е. Метод и аппаратура для регистрации акустической эмиссии и деформаций композитного графит-эпоксидного материала на основе анализа амплитудно-фазовых характеристик сигнала волоконно-оптического интерферометра Фабри-Перо // Дисс. канд. техн. наук. 05.11.01. СПб.: Университет ИТМО, 2018.

Efimov М.Е. Method and equipment for recording acoustic emission and deformations of composite graphite-epoxy material based on the analysis of amplitude-phase characteristics of the signal of a fiberoptic Fabry-Perot interferometer [in Russian] // PhD (Engineering). Saint Petersburg: ITMO University Press, 2018.

16. Дмитриев A.A., Грибаев А.И., Варжель С.В. и др. Метод описания высокопроизводительных волоконных брэгговских решеток [на англ. языке] // Оптиковолоконная технология. 2021. Т. 63. С. 102508. https://doi.Org/10.1016/J.YOFTE.2021.102508 Dmitriev A.A., Gribaev A.I., Varzhel S.V., et al. Highperformance fiber Bragg gratings arrays inscription method // Opt. Fiber Technol. 2021. V. 63. P. 102508. https://doi.Org/10.1016/J.YOFTE.2021.102508

авторы

Софья Дмитриевна Сычева — младший научный сотрудник, инженер, аспирант, Университет ИТМО, Санкт-Петербург, 197101, Россия; Scopus ID; 57209362819; https://orcid.org/ 0000-0001-6614-3790; 286114@edu.itmo.ru Артем Харисович Ахмеров — кандидат технических наук, научный сотрудник, инженер, Университет ИТМО, Санкт-Петербург, 197101, Россия; Scopus ID: 57211095787; https:// orcid.org/0000-0003-4388-622X; akakhmerov@itmo.ru Сергей Александрович Волковский — кандидат технических наук, научный сотрудник, ассистент, Университет ИТМО, Санкт-Петербург, 197101, Россия; Scopus ID: 57195476252; https://orcid.org/0000-0001-8462-5208; savolkovskii@itmo.ru Иван Геннадьевич Дейнека — кандидат технических наук, доцент, заведующий лабораторией, Университет ИТМО, Санкт-Петербург, 197101, Россия; Scopus ID: 6603305268; https://orcid.org/0000-0003-1557-8035; igdeyneka@itmo.ru Владимир Николаевич Кузнецов — инженер, программист, аспирант, Университет ИТМО, Санкт-Петербург, 197101, Россия; Scopus ID: 57673212700; https://orcid.Org/0000-0001-7565-2466; viikuziietcov@itmo.ru

Юлия Игоревна Явдыбаева — инженер, Университет ИТМО, Санкт-Петербург, 197101, Россия; Scopus ID: 57426281000; https://orcid.org/0000-0002-1451-7102; iuiandybaeva@itmo.ru Даниил Витальевич Пиха — инженер, Университет ИТМО, Санкт-Петербург, 197101, Россия; https://orcid.org/0009-0000-0940-1693; dvpikha@itmo.ru

Андреи Анатольевич Дмитриев — кандидат технических наук, начальник группы, Университет ИТМО, Санкт-Петербург, 197101, Россия; Scopus ID: 57215003358; https:// orcid.org/0000-0002-0674-8381; aadmitriev@itmo.ru

authors

Sofya D. Sycheva — Junior Researcher, Engineer, PhD Student, ITMO University, St. Petersburg, 197101, Russia; Scopus ID: 57209362819; https://orcid.org/0000-0001-6614-3790; 286114@edu.itmo.ru

Artem Kh. Akhmerov — PhD (Engineering), Researcher, Engineer, ITMO University, St. Petersburg, 197101, Russia; Scopus ID: 57211095787; https://orcid.org/0000-0003-4388-622X; akakhmerov@itmo.ru

Sergey A. Volkovsky — PhD (Engineering), Researcher, Assistant, ITMO University, St. Petersburg, 197101, Russia; Scopus ID: 57195476252; https://orcid.org/0000-0001-8462-5208; savolkovskii@itmo.ru

Ivan G. Deineka — PhD (Engineering), Associate Professor, Head of laboratory, ITMO University, St. Petersburg, 197101, Russia; Scopus ID: 6603305268; https://orcid.org/0000-0003-1557-8035; igdeyneka@itmo.ru

Vladimir N. Kuznetsov — Engineer, Programmer, PhD Student, ITMO University, St. Petersburg, 197101, Russia; Scopus ID: 57673212700; https://orcid.org/0000-0001-7565-2466; vnkuznetcov@itmo.ru

Yulia I. Yandybaeva — Engineer, ITMO University, St. Petersburg, 197101, Russia; Scopus ID: 57426281000; https://orcid. org/0000-0002-1451-7102; iuiandybaeva@itmo.ru Daniii V. Pikha — Engineer, ITMO University, St. Petersburg, 197101, Russia; https://orcid.org/0009-0000-0940-1693; dvpikha@itmo.ru

Andrey A. Dmitriev — PhD (Engineering), Group Leader, ITMO University, St. Petersburg, 197101, Russia; Scopus ID: 57215003358; https://orcid.org/0000-0002-0674-8381; aadmitriev@itmo.ru

Статья поступила в редакцию 01.11.2024 Одобрена после рецензирования 20.03.2025 Принята к печати 30.05.2025

The article was submitted to the edtorial office 01.11.2024 Approved after review 20.03.2025 Accepted for publication 30.05.2025

Sensors & Actuators: A. Physical 374 (2024) 115501

Contents lists available at ScienceDireet

Sensors and Actuators: A. Physical

journal homepage: www.journals.elsevier.com/sensors-and-actuators-a-physical

ELSEVIER

Automating data classification for label-free point-of-care biosensing in real complex samples

Jakub Kuncâk Michala Forinovâ Alina Pilipenco", Viktor Prochâzka", Peu Hoiâka, Sofya Dmitrievna Sychevac, Ivan Gennadievich Deynekac, Hana Vaisocherovâ-Lisalovâ"'

1 FZU - Institute of Physics of the Czech Academy of Sciences, Na Slovance 2, Prague 182 00, Czech Republic

Institute of Polymer Processing and Digitalization, Johannes Kepler Universität Linz, Altenbergerstraße 69. Linz 4040, Austria c ITMO - St Petersburg National Research University of Information Technology University, Kronvcrksky av. 49, Saint Petersburg 197101, Russia

n>

ARTICLE INFO

ABSTRACT

Keywords:

Quartz crystal microbalance Detection of pathogens Automatic data classification Point-of-core biosensors E. coli 0157:H7 SARS-CoV-2

Surface-based affinity biosensois present a promising avenue for point-of-care (POC) pathogen detection in real-world samples. While laboratory-based devices commonly employ various techniques to mitigate noise, signal drifts, fluidic artifacts, and other system imperfections, their simple cost-effective POC counterparts designed for field use frequendy lack such capabilities. This paper addresses this gap by introducing a procedure for automatically classifying pathogen presence in unprocessed liquids from direct detection data measured by a simple POC quartz crystal microbalance sensor device. The procedure integrates classical analytical tools such as filtering, data selection, baseline de-drifting, and result calculation in tailored successive steps, considering the nature of the sensor signal and the challenges posed by real-world media. We show that the developed procedure exhibits exceptional robustness across different biosensing assays and complex real-world media. Through optimizing parameters using diverse datasets encompassing Escherichia coli 0157:H7 (E coli) and SARS-CoV-2 detection in various media including food-derived matrices and cell culture media, we achieved rates of successful detection as high as 80.8 % and 90.9% for E coli and SARS-CoV-2, respectively, without extensive machine learning. Furthermore, we analyse the sensitivity of the procedure to variations of input parameters and with examples discuss key factors influencing overall procedure accuracy. Our results suggest that this exceptionally robust method holds potential as a straightforward tool for automating sample classification in point-of-care diagnostics, underpinning its promising broader applicability.

1. Introduction

In selecting diagnostic tools, medical testing systems frequently must balance the advantages of centralized laboratory networks with those of point-of-care (POC) diagnostics. While centralized laboratories offer reliable services, they often entail lengthy analysis times, and their' service can suffer from suboptimal distribution [11, as well as funding, regulations, and quality standards for effective disease management [2]. Conver sely, POC testing devices may provide a cost-effective, portable, and rapid alternative, suitable for non-experts and requiring minimal equipment [3]. Despite their" increasing prevalence, particularly dur ing the COVID-19 pandemic, many current POC devices, primarily relying on lateral flow assays, face challenges such as subjective result interpretation and are limited to qualitative outcomes [4,5].

The advancement of the next generation of POC devices hinges on

addressing three key areas: simplifying device and assay complexity, and automating data analysis [6]. Label-free surface-based affinity biosensors offer a potential solution by integr ating these appr oaches into a single POC biodetection system. However, challenges related to limited sensor r obustness per sist, par ticular ly in real-world applications such as pathogen detection in complex source media, where minimizing sample preparation is critical. Non-specific biofouling remains a significant obstacle, often limiting biosensor utilization to basic buffer solutions or well-defined model media [7-9]. This limitation necessitates either unwanted extensive sample pre-treatment [10,11] or the development of tailored antifouling surface chemistries [12,13].

The quartz crystal microbalance (QCM) method holds significant promise for POC applications due to its compact size, simple electronics, high resolution, and label-free operation [14,15]. Recently, we have demonstrated the potential of in-lab QCM devices for direct, rapid, and

* Corresponding author. E-mail address: lisalova@fzu.cz (H. Vaisocherova-Lisalova).

https://doi. org/10.1016/j.sna. 2024.115501

Received 5 March 2024; Received in revised form 18 April 2024; Accepted 18 May 2024 Available online 19 May 2024

0924-4247/© 2024 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Sensors and Actuators: A. Physical 374 (2024) 115501

J. Kuncak et aL

sensitive detection of viral and bacterial pathogens including SARS-CoV-2 and Staphylococcus aureus in complex fluids through the integration of tailor-made antifouling coatings [16-18]. However, compared to their laboratory counterparts, QCM sensors tailored for POC use usually exhibit significantly degraded performance characteristics. This is primarily attributed to the challenges associated with miniaturizing the device for portability and potential environmental exposures, leading to increased thermal drifts, electronic noise, and fluidic ar tifacts [19]. Despite numerous engineering str ategies aimed at mitigating these effects, portable QCM devices often necessitate operation with relatively noisy and drifting signals, contributing to the potential false classification of results.

To enhance the reliability of POC data and sample classification for QCM and other label-free POC biosensors, alternative approaches involving machine learning tools have emerged. However, these methods typically require extensive datasets, which are not readily available for POC devices designed to respond rapidly to emerging public health threats, such as pathogen outbreaks. Additionally, challenges such as model development, sensitivity to outliers, scalability, and data privacy further complicate the implementation of machine learning tools in this context [20,21]. Consequently, a manual or semi-automated evaluation of POC sensor data utilizing common analytical toolkits involving baseline de-drifting or filtering remains common practice, limiting the full exploitation of QCM devices in routine settings. While the integrated analytical tools are well-studied individually (e.g., signal de-drifting and filtering are ubiquitous in the analysis of biosensing data) [22,23], their combined use in real-world biosensing data remains largely unexplored. These results underpin the need for more robust and straightforward automated analysis methods.

In this article, we propose and evaluate a robust straightforward procedure for automated reliable analysis of real-world dir ect detection data obtained by a simple portable QCM device without relying on machine learning tools. This approach combines common analytical methods like signal filtering and de-drifting in tailor-made successive steps to achieve a high accuracy of automated complex sample classification. Moreover, we analyse how changes to the input parameters of these methods affect the overall procedure's accuracy. By optimizing and validating this procedure against datasets encompassing E coli 0157:H7 (E coli) and SARS-CoV-2 detection in various media, including liquids derived from food products (hamburger, pasta, milk, dumplings) and cell culture media, we highlight its consistency and universality across detection assays, environments, and devices. Our' results demonstrate the exceptional robustness of the developed procedure across diverse complex fluids and pathogen types, applicable to any surface-based label-free biosensors, promise to streamline data analysis regardless of sensor data quality. Thus, to note, this paper r efrains from suggesting improvements in microbiology, assay development, sensor transduction, interfacing and electronics, or fluidics.

2. Experimental methods

QCM device. For experiments regarding the detection of E. coli we combined the data measured using two different devices. In-lab device consisted of individual OpenQCM Q-l packages (Novaetech, Italy; with a frequency standard deviation for the fundamental frequency of 0.5 Hz in air, and 2.8 Hz for liquids) with both the original commercially available fluidic system (SARS-CoV-2 detection) as well as a custom-build fluidic package (E. coli detection), laboratory scale peristaltic pump, original OpenQCM electronics package (which is a small vector network analyser circuit) and data collection software with an irregular sampling rate.

POC QCM device is a custom-built 4-channel briefcase-sized POC instrument developed in the FZU - Institute of Physics of the Czech Academy of Sciences with a built-in per istaltic pump and a touchscreen control for data collection and analysis was employed in this study. The

channels are also inter facable with OpenQCM Q-l electronics. Her e for each time point a resonant frequency estimate was obtained by scanning the range 10 MHz ± 32 kHz for a crude maximum of the r esonant curve. In the proximity of the maximum, 400 additional measurements within 1 kHz were measured and fitted to a parabola using the least-squares method [24], where the par abola peak is assumed to be the resonant frequency. We used 10 MHz AT-cut crystals (Figure SI) with 100 nnr thick gold electrodes, a nominal sensitivity of 6.8ng/Hz and thermal stability of ± 5ppm (Krystaly, Czech Republic). The working detection principle of the QCM method is described in detail in Supplementary Material.

2.1. Biochip and sample preparation

The synthesis of functional antifouling terpolymer brush nano-coating, along with its functionalization with antibodies, followed procedures described elsewhere [16,17]. The terpolymer brush nanocoating was chosen for its high antifouling and antibody-binding capacity, as well as exceptional stability in various environmental conditions, making it highly suitable for use with POC biosensors [16-18]. Further details regar ding the synthesis of the terpolymer- brush, buffer and sample preparation, and functionalization of the QCM crystal with antibodies can be found in the Supplementary Material. In brief, "positive" samples were prepared by spiking E coli or SARS-CoV-2, ranging in concentr ation fr om 104 to 109 CFU/mL and approximately 104 PFU/nrL for E. coli and SARS-CoV-2, respectively, all within clinically relevant concentration ranges [25], into respective food-derived (E coli), or cell culture media (SARS-CoV-2) samples.

Detection assay. A dir ect detection assay format was utilized for all detection experiments conducted in this study. While detection assays for E coli and SARS-CoV-2 had been previously developed [17], new datasets were collected for the pur pose of this study using var ied QCM devices (as described above). E. coli was bound to the immobilized respective antibody through specific affinity interactions between the antibody and the cell sur face-bound protein. Detection of SARS-CoV-2 was achieved through specific affinity interactions between the immobilized antibody and N-protein/vRNA complexes [17]. The specific detection procedure was as follows: PBS buffer was injected into the flow cells at a flow rate of 35 fiL/min to establish a stable baseline (B1). Sample media (PBS, burger, dumpling, pasta, milk, or cell culture media) was then injected at 35pL/min for approximately 10 minutes (D), following which the flow cell was rinsed with PBS for 10 minutes (B2). A full binding calibration curve for E coli detection in milk was measured to determine the operating range, resulting in a limit of detection of 8 x 102 CFU/mL (Figur e S2).

3. Data processing procedure

Fig. 1 illustrates the detection scheme of an affinity QCM biosensor operated in a direct detection format, along with a sample classification approach utilizing the sensor resonant fr equency shifts. Initially, a buffer solution (PBS) is injected over the crystal surface functionalized with biorecognition elements (i.e., antibodies) to establish a baseline (Bl). Subsequently, during the detection phase (D), the injection of sample media induces a drop in the resonant frequency, primarily caused by significant differences in mass and viscosity compared to the control liquid. In this phase, a tar get analyte (i.e., E coli or SARS-CoV-2) pr esent in the analyzed complex sample selectively binds to the biorecognition elements on the sensing surface, r esulting in an asymptotic decr ease in the resonant frequency. Finally, the sample medium is replaced with the buffer control solution (PBS) to establish a second baseline (B2). If the analyte (i.e., E coli or SARS-CoV-2 with concentration r elevant to sensor operating range) is captured by the antibody-functionalized sensor surface during the detection phase, B2 should shift to a lower resonant frequency than Bl, indicating an increase in mass on the crystal surface. Fig. 1 b depicts a typical real sensor response for ' positive" sample (in a

J. Kuncak et aL

segments, we ensure the integrity of the data for the subsequent analysis, minimizing the impact of transient fluctuations. After extracting the data, the procedure can be split into four successive steps:

a. Filtering.

The extracted signal is first filtered using a moving median filter of window length M (i.e. replacing each data point with the median value of the M neighbouring data points within respective window), which rids the signal of occasional spikes stemming from a variety of sources. To smooth the signal, we used a Savitzky-Golay filter of length S and order 2 to replace data points with their estimates based on small-window polynomial fitting [26]. This step enhances the signal's clarity by applying a polynomial smoothing process, further refining the data for accurate analysis. Because our sensor devices have an uneven sampling rate, the filter lengths ar e initially designated as Mlength and S[cngr/,, both with a domain between 0 and 1. These values denote the fractions of data points existing between the markers x2 and x3, which ar e then used to calculate values for both M and S. This adaptive approach accounts for variations in sampling rates, ensuring consistent signal processing across different datasets. Here we note that via this method, sensors with uneven sampling rates will lead to slight differences between M and S for two different sensorgranrs; however , these differences are minor and are not expected to change the results of the analysis procedur e. The result of this step is a smoothed signal of the same length as the original extracted data that contains no outliers. An example result of filtering is presented in Fig. 2a.

b. Identification of the start of each baseline.

Due to the fluidic setup used herein, there is a delay in the sensor response after switching between analysed samples due to the dead volume of liquid upstream of the sensing surface (i.e., effects related to the axial dispersion between samples). Because x3 indicates the time that switching of D to B2 was initiated (initially outside the B2 phase), a shift is requir ed to place the marker within B2 phase (where only buffer is flowing through the sensor). Rather than treat both B1 and B2 separ ately, it is logical to shift the start of each phase forwar d in time (xl, x3) by N^ seconds (essentially a measure of the dispersion time). Adjusting for this delay improves the accur acy of phase identification, critical for the subsequent analysis steps. This allows to adapt the method for different fluidic setups or flow rates, where the optimal value of N^ (discussed below) will need to be increased for systems with lower flow rates, longer fluidic channels, or tubing with lar ger inner diameter. We deliber ately do not shift the ends of the segments, as the data after them are already considered transient.

c. De-drifting.

Many sensorgranrs typically suffer from long-term (and often constant) signal drifts that need to be removed before the baselines can be compared. Addressing this drift is paramount for ensuring the comparability of data across different time points. This drift is inherent to QCM applications, which for POC devices becomes further amplified due to the volume constraints that allow device portability [27]. Due to the short time window of the detection assay (10 minutes per section in our case), we assume the drift to be linear', and compensate it with a common technique of baseline curve fitting, as per [28]. Here a straight line is fitted to the isolated initial baseline Bl, bound by [xl, x2], where and the entire extracted signal is shifted such that the slope gradient of the signal in Bl is zero (as per Fig. 2c).

d. Baseline selection and comparison.

To further" minimise the effects related to the punrp star ting and stopping or any other transition-induced artifacts, only a portion from both baselines is selected for comparison. Selecting the most stable sections of the baselines ensures the reliability of the comparative analysis. It makes sense to select the par ts of Bl and B2

Sensors and Actuators: A. Physical 374 (2024) 115501

closer to the detection phase D to lessen the effects related to signal drift. In this regard, we define two parameters, B\p and B2p, which respectively signify what portion of each of the baselines (shortened by Nshift) are subject to comparison These parameters are bound between 0 and 1, which signifies what fraction of each baseline is used relative to the baseline length after marker shifting (e.g., a value of Blp = 0.5 would indicate that only Vi of the 1st baseline would be considered).

After steps (a)-(d) we calculate the mean of both the Bl and B2 windows (Fig. 2d). For this, we define a threshold value 7H, which is related to the signal change required to ensur e a specific sample was positive for the pathogen. Setting this threshold is a critical step in defining the criteria for positive detection, balancing sensitivity, and specificity. Hence, if the difference between the average value of Bl vs. B2 exceeds such threshold (i.e., if mean(Bl) - mean(B2) > TH), the sample is considered positive.

4. Experimental results and parameter optimisation

The developed results processing pr ocedur e was applied to two case studies, i.e., for the detection of the bacterial pathogen (£ coli) in epidemic outbreaks-relevant food-derived media and a vir al pathogen (SARS-CoV-2) in cell culture media. Fig. 3 illustrates the schematic r epr esentation of the biochip and its surface modification applied in this study. The functional surface on the QCM crystal was formed by a tailor-made terpolymer brush nanocoating functionalized with antibodies to E coli or SARS-CoV-2 (see also Supporting Mateiial for more details). This type of functional coating was chosen for the pur pose of this study to i) ensure an antifouling layer which enables the analysis of complex samples, and ii) enable the possibility to immobilize the biorecognition elements (antibodies in this study). In both E coli and SARS-CoV-2 cases, a dir ect detection assay format was employed. E. coli was bound to the immobilized respective antibody through specific affinity interactions between the antibody and the cell surface-bound protein. Detection of SARS-CoV-2 was achieved thr ough specific affinity interactions between the immobilized antibody and N-protein/vRNA complexes [17].

Optimizing analysis parameter for E. coli detection. The previous section defines the device- and analyte-specific parameters that define the data analysis procedure: S/engrf,, Nshift, Blp.B2p and TH. We first opti-

mized these parameters r egar ding the success of the analysis procedure using data taken regarding the detection of E. coli using 125 total samples (positive and negative) representing real-world detection conditions.

We analysed 62 samples using the POC device, 33 positive for E coli (with varying concentrations) and 29 control samples (E coli free). The control samples were derived from a variety of sources: bur ger (8x), PBS (llx), nrilk (4x) and pasta (6x). Likewise, we analysed 63 samples using the In-lab device, 29 positive and 34 negative, with the latter taken using samples derived from bur ger (21 x) and Czech dumplings (13x), some of which were subjected to a slight pre-treatment procedure (heat treatment or treatment via sonification). The positive samples for both devices are summarized in Table 1; also see Fig. 4 and Figure S3 with successfully evaluated sensorgranrs. To contextualize our findings, comparing these success rates with those reported in existing literature underscores the effectiveness of our parameter optimization strategy, especially in the face of sample diversity and varying concentrations of E coli.

Our goal was to obtain optimal values of each input parameter that result in the highest success rate for correct sample classification across all samples. Due to the relatively small data and parameter sets we opted for a more direct approach (randomized brute-force) rather than machine learning. The results obtained thus show that the method is sufficiently successful to the point where attempting alternative approaches was deemed unnecessary. Random parameter combinations

Sensors and Actuators: A. Physical 374 (2024) 115501

J. Kuncak et aL

there is not any appropriate analysis method for positive/negative classification. From our observation, the most typical reason for failure was improperly selected TH and non-linear drift in the signal.

These individual failure cases stem from the use of constant analysis par ameters for the entire data set; some of these data would be correctly classified under a different set of parameters. To investigate the importance of the individual parameters, we subjected the E. coli data to a sensitivity analysis, varying each parameter within its range while keeping the others constant. The results of this sensitivity analysis are shown in Fig. 5. The length of Savitzky-Golay filter Sfe^j, and the first baseline portion Blp have only a limited effect on the procedure's success. There is a significant dr op in accuracy with small B2p values, as the analysis considers a small portion of the B2 signal close to region D and the transition dominates the selected sample. Both lar ge N^ or small B2p value increases the likelihood of non-linear" drifts or long-term instabilities affecting the signal. A large median filter length M better removes outliers but also snroothens desir able signal pr oper ties, such as the B1 - D - B2 tr ansitions.

Interestingly, our analysis identified mislabelling in a small subset of samples, with 6 negative samples and 18 positive samples for E coli, and all four incorrect SARS-CoV-2 samples being positive (Table 1). This pattern suggests a predilection towards false negatives within our current parameter settings. The sensitivity analysis shows that higher TH leads to the correct classification of more negative samples (at the cost of more false negatives): a large TH is more likely to label every sample negative, and a few samples have such large frequency change A/ that they would be recognised as positive. Thus, manual tweaking of the TH parameter may be of significance for epidemic monitoring, as during mass screening, a false negative diagnosis may have a lar ger impact than a false positive.

There is a difference in data quality across collection devices (with the In-lab devices having a higher quality, see also Fig. 4 and Figur e S3). However, re-running the optimiser with just the data from the In-lab device correctly recognises 53/63 (84 %) of measurements, and repeating with the POC device gives 51/62 (82 %). This versatility suggests our method's potential for broad implementation in POC diagnostics, moving us closer to a goal of technician-fr ee operation. It highlights the practical aspect of generating a confidence level for each measur ement, allowing for automated processing of high-confidence detections and manual r eview of low-confidence cases.

Admittedly, there potentially exist more efficient algorithms to optimize the parameters used in the analysis procedure considered herein, such as genetic algorithms or simulated annealing [34]; however-, those methods applied with the analysis procedure here would likely lead to similar results, as several of the parameters have limited effect on the classification success rate (Mfe,^. Sfengr/,.B1P). Fur ther increases in the success rate would necessitate the development of a different analysis pr ocedure, or alternatively, through entirely differ ent approaches based on machine learning, which, however, typically require significantly larger data sets for training. Regardless, the success rate achieved here for £ coli detection (80.8 %) and SARS-CoV-2 (91 %) is close to that achieved by automated machine learning procedures used towar ds the analysis of QCM data: Tatarko et aL reported accur acy of —90 % to distinguish between trypsin and plasnrin [35] via mechanisms similar to that reported here; where Rao et al. reported 98 % and above in the use of QCM to classify between different alcohols [36], albeit using an entirely different detection format than that used here. Our" work not only advances the field of POC diagnostics by enhancing detection reliability but also sets a foundation for future exploration in automated biosensing technologies.

6. Conclusion

In conclusion, our study introduces a straightforward yet highly effective procedur e for automating data analysis in surface-based label-

free biosensors aiming at point-of-care diagnostics. By integrating commonly employed methods such as signal filter ing and de-dr ifting in a sequence and optimizing parametric settings for unprocessed biological liquids in real-world environmental conditions, which often introduce significant signal imperfections, we demonstrate the excellent robustness of the developed procedure and its potential to streamline data analysis across diverse biosensing applications. Through optimization against datasets encompassing pathogen detection (E. coli 0157:H7 and SARS-CoV-2) in various real-world media, including complex biological and food-derived samples, we validate the reliability and consistency of the procedure. We achieved successful data classification rates as high as 80.8% and 90.9% for E coli and SARS-CoV-2, respectively. These values are compar able to the results reported for data analysis methods involving extensive machine learning procedures. Our findings demonstrate the r obustness and universality of the proposed procedure, applicable to a wide range of surface-based affinity biosensors regardless of their transduction mechanisms. Consequently, these results hold significant potential for both laboratory and point-of-care (POC) diagnostic devices, promising enhanced oper ational efficiency and expanded application areas. The simplicity and effectiveness of this automated data analysis approach pave the way for widespread adoption in various fields, facilitating rapid and reliable detection of target analytes in diverse sample matrices.

Funding

This work was suppor ted by Praemium Lunrina quaer untur of the Czech Academy of Sciences [LQ100101902], by the Czech Science Foundation [24-10671S], and Operational Programme Johannes Amos Comenius financed by European Structural and Investment Funds and the Czech Ministry of Education, Youth and Sports [Project SENDISO CZ.02.01.01/00/22.008/0004596].

CRediT authorship contribution statement

Jalcub Kuncak: Writing - original draft, Formal analysis, Data curation. Michala Forinova: Writing - review & editing, Validation, Data curation, Conceptualization. Alina Pilipenco: Wr iting - r eview & editing, Validation, Data cur ation. Viktor Prochazka: Formal analysis. Petr Horak: Software, Methodology. Sofya Dmitrievna Sycheva: Formal analysis. Ivan Gennadievich Deyneka: Formal analysis. Hana Vaisocherova-Lisalova: Writing - review & editing, Supervision, Resources, Conceptualization.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal r elationships that could have appear ed to influence the work reported in this paper.

Data Availability

Data will be made available on request. Acknowledgements

The author's would like to thank Nicholas Scott Lynn Jr., the Institute of Physics of the Czech Academy of Sciences for his critical assessment and advice on the work, Vaclav Honig, Institute of Parasitology, Biology Centre of the Czech Academy of Sciences, Ceske Budejovice, Czech Republic, Jan Sterba, Faculty of Science, University of South Bohemia, Czech Republic, Czech Republic, Zdenek Farka and Petr Skladal, Masaryk University, Brno, Czech Republic, for providing bacterial and viral cultures and supporting the presented research with BSL-2 and BSL-3 laboratory infrastructures.

Sensors and Actuators: A. Physical 374 (2024) 11SS01

[31] A. Rahtu. M. Ritala, Compensation of temperature effects in quartz crystal microbalance measurements. Appl. Phys. Lett. 80 (3) (2002) 521-523.

[32] J.H. Seo. et al.. Temperature compensation method for resonant microsensors based on a controlled stiffness modulation. J. Appl. Phys. 104 ( 1 ) (2008) 014911.

[33] BP. Sagmeister. et al.. User-friend h', miniature biosensor flow cell for fragile high fundamental frequency quartz crystal resonators. Biosens. Bioelectron. 24 (8) (2009)2643-2648.

[34] O, Castillo. P. Melin. Soft computing for control of non-linear dynamical systems. Phvsica Vol. 63 (2012).

[35] M. Tatarko. et al.. Machine learning enabled acoustic detection of sub-nanomolar concentration of trypsin and plasmin in solution. Sens. Actuators B Chem. 272 (2018)282-288.

[36] B.K. Rao. et al.. QCM Sensor-Based Alcohol Classification by Advance Machine Learning Approach, Intell. Comput. Control Commun. (2021) 305-320.

J. Kuncak et aL

Appendix A. Supporting information

Supplementary data associated with this article can be found in the

online version at doi:10.1016/j.sna.2024.115501.

References

[1] D. Mabey, et al.. Diagnostics for the developing world. Nat. Rev. Microbiol 2 (3) (2004) 231-240.

[2] P. Yager. G.J. Domingo, J. Gerdes. Point-of-care diagnostics for global health. Annu Rev. Biomed. Eng. 10 (2008) 107-144.

[3] D. Liu. et al.. Trends in miniaturized biosensors for point-of-care testing, TrAC Trends Anal. Chem. 122 (2020) 115701.

[4] G.A. Posthuma-Trumpie. J. Korf. A. van Amerongen. Lateral flow (immuno)assay: its strengths, weaknesses, opportunities and threats. A literature survey, Anal. Bioanal. Chem. 393 (2) (2009) 569-582.

[5] F. Di Nardo. et al.. Ten years of lateral flow immunoassay technique applications: trends, challenges and future perspectives. Sensors 21 (15) (2021).

[6] K. Markandan. et al.. Emergence of infectious diseases and role of advanced nano mate rials in point-of-care diagnostics: a review, Biotechnol. Genet Eng. Rev. (2022) 1-89.

[7] F. Ma, et al.. Glycosvlation of quinone-fused polvthiophene for reagentless and label-free detection of E. coli. Anal. Chem. 87 (3) (2015) 1560-1568.

[8] N.A. Masdor. Z. Altintas. I.E. Tothill. Sensitive detection of Campylobacter jejuni using nanoparticles enhanced QCM sensor, Biosens. Bioelectron. 78 (2016) 328-336.

[9] M. Pohanka. QCM immunosensor for the determination of Staphylococcus aureus antigen, Chem. Pap. 74 (2) (2020 ) 451-458.

[10] H. Wang, et al.. Rapid and sensitive detection of Campylobacter jejuni in poultry products using a nanoparticle-based piezoelectric immunosensor integrated with magnetic immunoseparation. J. Food Prot. 81 (8) (2018) 1321-1330.

[11] G. Bavramoglu. et al., Rapid and label-free detection of Brucella melitensis in milk and milk products using an aptasensor, Talanta 200 (2019) 263-271.

[12] M J. Russo, et al.. Antifouling strategies for electrochemical biosensing: mechanisms and performance toward point of care based diagnostic applications, ACS Sens. 6 (4) (2021) 1482-1507.

[13] 1. Visova, et al.. Surface preconditioning influences the antifouling capabilities of zwitterionic and nonionic polymer brushes. Langmuir 36 (29) (2020) 8485-8493.

[14] H.J. Lim. et al.. Quartz crystal microbalance-based biosensors as rapid diagnostic devices for infectious diseases. Biosens. Bioelectron. 168 (2020) 1 12513.

[15] P. Prakrankamanant. Quartz crystal microbalance biosensors: prospects for point-of-care diagnostics, J. Med. Assoc. Thai 97 (Suppl 4) (2014) S56-S64.

[16] M. Forinova. et al.. A comparative assessment of a piezoelectric biosensor based on a new antifouling nanolayer and cultivation methods: enhancing detection in fresh dairy products, Curr. Res. Biotechnol. 6 (2023).

[17] M. Forinova. etal., Functionalized terpolymer-brush-based biointerface with improved antifouling properties for ultra-sensitive direct detection of virus in crude clinical samples, ACS Appl. Mater. Interfaces (2021).

[18] A. Pilipenco. et al.. Negligible risk of surface transmission of SARS-CoV-2 in public transportation, J. Travel Med. 30 (5) (2023).

[19] M.L. Sin. et al.. Advances and challenges in biosensor-based diagnosis of infectious diseases. Expert Rev. Mol. Diagn. 14 (2) (2014) 22S-244.

[20] F. Cui. etal.. Advancing biosensors withmachine learning. ACS Sens. 5(11) (2020) 3346-3364.

[21] S. Kadian. et al.. Recent advancements in machine learning enabled portable and %vearable biosensors. Talanta Open 8 (2023) 100267.

[22] S. Paul, et al.. On-line drift compensation for continuous monitoring with arrays of cross-sensitive chemical sensors, Sens. Actuators B Chem. 368 (2022) 132080.

[23] Chinrungrueng, C. Combining Savitzky-Golay filters and median filters for reducing speckle noise in SAR images, SMC'03 Conference Proceedings. 2003 IEEE International Conference on Systems, Man and Cybernetics. Conference Theme -System Security and Assurance (Cat. No.03CH37483). Washington, DC, USA (1) (2003) 690-696.

[24] S. Na Songkhla. T. Nakamoto. Overview of quartz crystal microbalance behavior analysis and measurement, Chemosensors 9(12) (2021) 350.

[25] Y.C. Lin. et al.. Detection and quantification of infectious severe acute respiratory coronavirus-2 in diverse clinical and environmental samples, Sci. Rep. 12 (1) (2022) 5418.

[26] A. Savitzkv, M.J.E. Golay. Smoothing and differentiation of data by simplified least squares procedures. Anal. Chem. 36 (8) (1964) 1627-1639.

[27] A.D. Easley. et al., A practical guide to quartz crystal microbalance with dissipation monitoring of thin polymer films. J. Polym. Sci. 60 (7) (2022) 1090-1107.

[28] J.E. Haugen. O. Tomic. K. Kvaal. A calibration method for handling the temporal drift of solid state gas-sensors. Anal. Chim. Acta 407 (1) (2000) 23-39.

[29] J. Bergstra. Y. Bengio. Random search for hyper-parameter optimization. J. Mach. Learn Res. 13 (2) (2012).

[30] R Fernandez, et al.. A real-time method for improving stability of monolithic quartz crystal microbalance operating undo" harsh environmental conditions, Sensors 21 (12) (2021) 4166.

Jakub Kuncak received a title of Master of Engineering and Electrical and Electronics Engineering from University of Aberdeen, Scotland in June 2022. In 2022 - 2023, he worked as a Graduate Project Assistant at Hoerbiger Institute for Smart Actuators at Johannes Kepler Universitaet Linz. Since 2023, he is a researcher and recently a PhD student at Institute of Polymer Processing and Digitalization. also at Johannes Kepler Universitaet Linz, where his research interest is sensing, control, and digitalization of polymer extrusion machines.

Michala Forinova earned her master's degree in 2020 from Charles University's Faculty of Science, specializing in processing 3D craniofacial scans with landmark and curve-based methods, rigorous statistical analysis, and advanced data visualization techniques. She is currently pursuing a PhD at Charles University's Faculty of Mathematics and Physics in Prague, focusing on Biophysics, Chemical, and Macromolecular Physics. She conducts research at the Institute of Physics of the Czech Academy of Sciences, with a particular focus on developing advanced antifouling polymer brushes. Her research explores their application in biosensors based on quartz crystal microbalance or surface plasmon resonance, as well as studying molecular interactions at the interface of complex samples and surfaces.

Alina Pilipenco, a Ph.D. student in Biophysics at Charles University in Prague, excels in antifouling polymer research at the Institute of Physics of the Czech Academy of Sciences. Her research focuses on developing coatings that resist microbial adhesion, critical in medical and environmental applications. Her work extends to virus diagnosis, contributing significandy during the SARS-CoV-2 pandemic. Alina's expertise in biointerface science allows her to conduct complex in vivo and in vitro experiments. Her contributions to the field are highlighted in several high-impact journals.

Viktor Prochazka received a BSc degree in Mathematics from Charles University, Prague in 2022. Currendy pursues the MSc degree in Mathematics in Data Science at Technical University of Munich. His research interests are mathematical statistics and machine learning.

Petr Horak received the M.Sc. and Ph.D. degree at the Czech Technical University in Prague, Faculty of Electrical Engineering. He worked as a researcher at the Institute of Photonics and Electronics of the Academy of Sciences of the Czech Republic focusing on text-to-speech synthesis and SPR biosensors. Since 2018, he has been working at the FZU -Institute of Physics of the Academy of Sciences of the Czech Republic on distributed computing at CERN, and QCM biosensors. His research interests mainly concern the development of new biosensory devices.

Sofya Dmitrievna Sycheva is a third-year postgraduate student in the speciality of Instruments and methods of measurement (Engineering Sciences) and holds the position of a junior researcher at ITMO University. The main areas of work and interests are processing and analysis of digital data from fiber systems and sensors, FPGA design.

Ivan Gennadievich Deynelca, Ph.D. currendy works as an associate professor at the Institute of Higher School of Engineering and Technology, and is also the head of the Laboratory of "Modeling and Programming" of the Research Center of Light-guided Photonics at ITMO University. His professional interests include research and development of interferometric fiber optic sensors, nature-inspired algorithms, and FPGA and SoC programming.

Hana Vaischerova-Lisalova is actively pursuing research on biosensors, functional biomaterials, and (nano)coatings, with a focus on understanding the intricate phenomena of biomolecular interactions at interfaces in their native environments. Holding a Ph.D. in Biophysics from Charles University in Prague, Hana furthered her expertise in chemical engineering during a postdoctoral fellowship at the University of Washington in Seatde under the supervision of Shaoyi Jiang. In 2019, she initiated an interdisciplinary biointerface research program at the FZU - Institute of Physics of the Czech Academy of Sciences. Over the past five years, she has assembled a research team, secured prestigious grants, and earned recognition, including the Werner von Siemens and global Tech Innovation Excellence (TIE) Awards.

Обратите внимание, представленные выше научные тексты размещены для ознакомления и получены посредством распознавания оригинальных текстов диссертаций (OCR). В связи с чем, в них могут содержаться ошибки, связанные с несовершенством алгоритмов распознавания. В PDF файлах диссертаций и авторефератов, которые мы доставляем, подобных ошибок нет.