<p>Durations between events of interest such as intra-day transactions of assets can reflect the volatility of asset prices in financial markets. The diverse dynamics of these intervals, which we refer to as financial durations, offer valuable insights into market behavior for investors. Inspection of streaming price data for structural breaks and timely and accurate detection of transitions between different duration patterns within a trading day enables practitioners to update parameters of suitable duration models. In this article, an innovative Ensemble Penalized Estimating Function (E-PEF) approach is proposed to effectively detect change points in the logarithmic autoregressive conditional duration models for financial durations. As a quasi-score-based online detection approach, this methodology leverages Mahalanobis distances and the block bootstrap sampling method to compute critical values for finite sample time series. The online structural break detection rule is informed by comparing observed quasi-scores in the monitoring period with pre-calculated critical values from training data in an ensemble manner. Extensive simulations demonstrate that the E-PEF method has fast structural break detection performance, while effectively controlling the probability of false detection. In the real data application, we applied our method to identify structural breaks for four assets, explored their relationships with summarized changes in volatility patterns, and noted several considerations for practitioners in the financial market.</p>