Risk Control Research Articles

Work accidents are one of the serious problems that have a direct impact on worker safety and company productivity. This study was conducted at PMKS PT. Sisirau located in Aceh Tamiang. This research aims to explore potential sources of danger, evaluate the degree of associated threats, and formulate strategic preventive measures through the structured framework of Hazard Identification, Risk Assessment, and Control Determination (HIRADC) method. The focus of this study was on three work stations, namely loading ramp, boiling (sterilizer station), and tipler. The data used were in the form of work accident reports from January 2021 to February 2025, as well as the results of observations and interviews in the field. The results of the study showed that there were a total of 18 work activities identified as hazardous, consisting of 8 activities at the loading ramp station, 5 activities at the boiling station (sterilizer), and 5 activities at the tipler station. The results of the risk assessment showed that at the loading ramp station there were 5 activities with One activity presented moderate risk; three others showed low-risk potential. At the boiling station (sterilizer) there were 4 activities with a moderate risk and 1 activity with a low risk. Meanwhile, at the tipler station, among the evaluated activities, one was determined to involve a significant risk factor, whereas two others presented comparatively lower levels of risk, and 2 activities with low risk. Recommended risk control includes 7 activities with engineering control, 9 activities with administrative control, and 10 activities with mandatory Application of individual safety gear.

This paper presents a new framework for portfolio management that incorporates sustainability considerations in the form of environmental, social, and governance risks (ESG) alongside the impact of historical and projected company performance, as well as a fuzzy environment to account for expected return uncertainty. To account for market and investor uncertainties, we integrate the conditional value-at-risk (CVaR) risk measure with credibility theory. Unlike traditional portfolio optimization methods, which rely heavily on probabilistic assumptions and may fail to capture real-world uncertainty, our model uses fuzzy logic principles to more realistically represent uncertainty when historical data is incomplete or expert opinions are inaccurate. To assess historical and projected company performance, we use data extracted from quarterly company reports and employ advanced text analytics tools such as FinBERT sentiment analysis and the NotebookLM platform. These tools enable us to extract subtle insights and sentiment trends that are critical for predicting future performance. Empirical validation of the proposed framework is conducted using historical stock return data from the DJIA. A diversified portfolio of assets is selected and the optimal stock allocation is obtained under the proposed credibilistic CVaR (CCVaR) validation approach. The results show that the portfolios optimized under the CCVaR framework offer superior adverse risk control and greater resilience to market volatility compared to traditional approaches. At the end of the paper, the proposed model is compared with the traditional equal weight strategy and the results are presented. This study provides valuable practical insights for risk-averse investors and portfolio managers seeking stronger and more stable investment strategies in uncertain financial environments.

ABSTRACT Balancing debt levels with capital needs is a key challenge in corporate financing decisions. The speed of capital structure adjustment (SOA) determines how efficiently firms return to their target leverage after deviations, thereby affecting risk control, the cost of capital, and strategic resilience. As artificial intelligence (AI) increasingly permeates corporate operations and governance, whether and how AI enhances SOA has become an important question. This study investigates the impact of AI on SOA using panel data on Chinese A-share listed firms from 2010 to 2023. We employ partial adjustment models and fixed effects estimation, supplemented by alternative AI proxies and leverage measures, system GMM, one- and two-year lag regressions, propensity score matching (PSM), and two-stage least squares (2SLS) methods to ensure robustness. The results show that AI significantly accelerates SOA, primarily by enhancing the information environment and mitigating operational risk. Heterogeneity analyses reveal that the effect is strongest at moderate board independence and among low-growth firms. These findings deepen our understanding of how emerging technologies influence corporate financing decisions and provide practical implications for managers and policymakers seeking to optimize capital structure in the era of AI.

Beaver dam–pond systems reshape the hydrology of lowland landscapes by slowing water flow and trapping sediments, thereby reducing the movement of pollutants. This study examined how such beaver-engineered wetlands can naturally filter and signal contamination risks associated with lead (Pb). We combined data from three matrices—bottom sediments, riparian vegetation, and non-invasively collected beaver fur—across three Lithuanian sites (2022–2024). Previously published datasets on plants and sediments were complemented with new information from beaver fur to explore seasonal and age-related effects as well as differences inside and outside dam zones. Lead levels were consistently higher in sediments than in plants, while beaver fur reflected variable, site-specific exposures. These results show that beaver activity contributes to the capture and redistribution of sediment-bound Pb in wetland buffers. The approach demonstrates how beaver habitats can serve as low-cost, nature-based sentinels for pollutant monitoring. Using beaver fur as a non-invasive bioindicator and managing dam stability can improve the ecological and policy relevance of buffer zones. Overall, the findings support the integration of beaver-engineered wetlands into environmental management and EU water policy, contributing to SDG 6 goals for clean water and sustainable wetland use.

A total of 18 polycyclic aromatic hydrocarbons (PAHs) were analyzed in 122 urban soil samples collected from Lanzhou (n = 54) and Urumqi (n = 68) in Northwest China for their sources and ecological and health risks based on positive matrix factorization and Monte Carlo simulation. The total concentration of 18 PAHs (∑18PAHs) in Lanzhou and Urumqi ranged from 39.1 to 6584 and 4.98 to 4690ng/g, averaged as 863 and 489ng/g, and dominated by 3-, 4-, and 5-ring PAHs. Higher ∑18PAHs mainly distributed in the east, west and middle parts of Lanzhou as well as northeast and west parts of Urumqi. PAHs in Lanzhou and Urumqi originated from vehicle emissions (69.2% and 60.0%), coal/biomass combustion (16.9% and 28.7%), and mixed sources (13.9% and 11.3%). Probabilistic risk assessment indicated that pyrene and benzo[b]fluoranthene exhibited higher ecological risks. PAHs had relatively high carcinogenic risk for adults and adolescents. Source-oriented risk assessment revealed vehicle emissions as primary contributors to ecological risk (53.3% and 53.8%) and carcinogenic risk (55.9% and 55.3%), necessitating priority control. The exposure of PAHs significantly impacted human respiratory system. The current research results provide certain scientific support for the prevention control of pollution and risk of PAHs.

Diet influences brain health through many connected metabolic and molecular pathways, and these effects are stronger in obesity. This review links diet quality with cognitive decline and dementia risk. Ultra-processed, high-fat, high-sugar diets drive weight gain, insulin resistance, and chronic inflammation. These changes trigger brain oxidative stress, reduce DNA repair, deplete NAD+, disturb sirtuin/PARP balance, and alter epigenetic marks. Gut dysbiosis and leaky gut add inflammatory signals, weaken the blood–brain barrier, and disrupt microglia. Mediterranean and MIND diets, rich in plants, fiber, polyphenols, and omega-3 fats, slow cognitive decline and lower dementia risk. Trials show extra benefit when diet improves alongside exercise and vascular risk control. Specific nutrients can help in certain settings. DHA and EPA support brain health in people with low omega-3 status or early disease. B-vitamins slow brain shrinkage in mild cognitive impairment when homocysteine is high. Vitamin D correction is beneficial when levels are low. A practical plan emphasizes healthy eating and good metabolic control. It includes screening for deficiencies and supporting the microbiome with fiber and fermented foods. Mechanism-based add-ons, such as NAD+ boosters, deserve testing in lifestyle-focused trials. Together, these measures may reduce diet-related brain risk across the life span. At the same time, artificial intelligence can integrate diet exposures, adiposity, metabolic markers, multi-omics, neuroimaging, and digital phenotyping. This can identify high-risk phenotypes, refine causal links along the diet–obesity–brain axis, and personalize nutrition-plus-lifestyle interventions. It can also highlight safety, equity, and privacy considerations. Translationally, a pattern-first strategy can support early screening and personalized risk reduction by integrating diet quality, adiposity, vascular risk, micronutrient status, and microbiome-responsive behaviors. AI can aid measurement and risk stratification when developed with privacy, equity, and interpretability safeguards, but clinical decisions should remain mechanism-aligned and trial-anchored.

In an increasingly competitive market environment, a differentiated marketing strategy has become a crucial path for Small and Medium-sized Enterprises (SMEs) to gain a competitive advantage and achieve sustainable development. Due to constraints in resources, talent, and brand influence, SMEs face numerous challenges and difficulties in formulating and implementing their differentiation strategies. This paper first elaborates on the necessity and theoretical foundations for SMEs to pursue differentiation. It then deeply analyzes key issues encountered during the strategy formulation phase, such as inaccurate market positioning, difficulty in selecting differentiation points, and resource constraints. Subsequently, it explores critical difficulties faced during the strategy implementation phase, including insufficient execution capability, poor innovation sustainability, organizational coordination challenges, and weak risk control. Finally, the paper proposes a series of countermeasures and suggestions, including strengthening core competencies, focusing on technological and model innovation, building a flexible organizational structure, and leveraging digital technology to enhance efficiency. The aim is to provide theoretical reference and practical guidance for SMEs to more effectively formulate and implement differentiated marketing strategies, helping them stand out in niche markets.

Green finance has emerged as a driving force in the restructuring of energy-intensive industries by integrating ecological objectives into financial decision-making. This paper explores the significance and strategic pathways of financial transformation for high-energy-consumption enterprises under the guidance of green finance. The discussion emphasizes the necessity of industrial upgrading, diversification of financing channels, improvement of environmental risk management, and reinforcement of sustainable competitiveness. Four major strategies are examined: enhancing green accounting information disclosure, innovating green investment instruments, strengthening environmental cost and performance management, and constructing green risk control systems. These approaches illustrate how enterprises can achieve financial resilience and ecological accountability simultaneously, thereby securing long-term growth within increasingly stringent environmental and financial frameworks.

Traditional Index Models have long been constrained by oversimplified assumptions limiting their accuracy in portfolio construction. This paper proposes a machine learning-enhanced Index Model framework, leveraging Random Forest algorithms whether this way can improve the accuracy of stock prediction or not. Using daily data of the OEX index (and its constituent stocks) from 2004 to 2024 as the empirical sample, this paper conducts the empirical research. The results show that the machine learning-enhanced Index Model can achieve a lower test-set MSE and higher R compared to the traditional Index Model, indicating superior return prediction accuracy. The ability of the random forest to optimize the index model portfolio depends on the compatibility of stock features with the model. Therefore, the original random forest has the characteristics of low interpretability and high bias failure, which leads to the model being prone to failure. Moreover, this prediction constructed based on the models predicted returns outperform the benchmark OEX index in risk-adjusted returns and risk control. The random forest model captures the complex situations overlooked by the traditional index model, and provide more reliable input for stock prediction.

In April 2020, Yuan You Bao, a retail financial derivative product launched by Bank of China, encountered unprecedented losses when the NYMEX WTI May futures contracts settled its price at -$37.63 per barrel. This incident reflects the cruel reality about Chinas retail finance ecosystem by revealing the structural weakness in product design, disclosure, and risk controls. This paper aims to apply a qualitative method, combining regulatory documents, media reports, and comparative international frameworks on investor protection and product governance, to analyze. Problems addressed in this research are purposed to conclude why a commodity derivative reached mass-market shares, and how institutional risk management malfunctioned, as well as what measures of reforms are worth considering to align retail-facing derivative products with long-term stability. Methods applied include timeline reconstruction, stakeholder analysis, and comparison to EU MiFID II product-governance and PRIIPs disclosure regimes and U.S. Title VII (Dodd-Frank) conduct standards. The research finds there is a mismatch between product structures and investor capacity, insufficient margining and position-limit safeguards, delayed rollover safety protocols, and limited updated risk communication. From the macro-perspective, this incident indicates that the financial market in China requires further regulatory methods, including robust product-approval committees, negative-price logistics, suitability gating, and pre-contractual information disclosures. The paper draws a conclusion that the financial crisis should catalyze governance and regulation upgrades across Chinese banks and accelerate rule-making that balances financial innovation for protecting investors stakes.

This article examines Beijing's real estate market risk controls, focusing on challenges in megacities. Findings show Beijing's real estate finance is unique: policy guidance is strong, and there are big differences between core and non-core areas in resources and housing value. It is also closely linked to the financial system, so price changes can greatly affect financial stability. Currently, multiple risks exist together, such as price bubbles in core areas, developers' cash flow problems, and chain-reaction risks. The market is very sensitive to policy changes, making risk prevention harder. Short-term downward pressure comes from policies, unbalanced supply and demand, and weak buyer confidence. Although support policies have been introduced, their effect is delayed, and the market decline remains clear. The article suggests creating a flexible prevention system, shifting from building new homes to improving existing ones, and balancing market stability with public needs. Specific ideas include different rules for core areas (to stabilize prices and prevent speculation) and non-core areas (to improve public facilities). It also recommends a multi-department evaluation system to adjust policies each quarter based on sales data. An intelligent monitoring platform using various data sources could spot abnormal changes, improve information transparency, guide better decisions, and support healthy market development.

ABSTRACT Reliable landslide hazard assessment in complex environments is essential for effective risk prevention and control. However, extracting and integrating relevant knowledge from extensive, unstructured geoscience literature remains a significant challenge. This paper proposes a framework to construct a landslide hazard assessment knowledge graph (LHAKG). First, the framework develops an ontology comprising four core elements: spatial information, causative factor, assessment model, and assessment data. These elements and their interrelations form the conceptual backbone of the LHAKG. Next, an XLNet-BiLSTM-CRF model is trained on a self-developed domain-specific corpus to extract entity information related to landslide hazard assessment elements from scientific literature. Finally, Sentence-Transformer techniques are applied for knowledge fusion, facilitating semantic alignment and refinement of entities across heterogeneous categories. A case study using a dataset of 1,612 geoscience papers demonstrates the effectiveness of the framework, yielding the LHAKG with 13,400 nodes and 62,975 relationships. The results confirm that LHAKG can efficiently extract, organize, and represent landslide hazard assessment knowledge. Furthermore, the LHAKG facilitates multi-dimensional knowledge recommendation across diverse methodologies and geographic contexts. This capability enhances intelligent analysis and decision support in geohazard research, especially with incomplete data, establishing a robust foundation for broad-scale landslide risk assessment and prediction.

Background: Lipoprotein(a) [Lp(a)] is a known risk factor for atherosclerotic cardiovascular disease (ASCVD). However, it is unclear whether the ASCVD risk associated with elevated Lp(a) is uniform or differs according to baseline risk profiles. Aims: To compare the association of elevated Lp(a) with incident ASCVD across levels of risk factor burden and 10-year predicted ASCVD risk. Methods: We performed a prospective analysis among participants at Visit 4 (1996-98) of the ARIC study without baseline ASCVD and with Visit 4 Lp(a) measurements (Denka Seiken assay). ASCVD risk factors were smoking, hypercholesterolemia, diabetes, hypertension, and chronic kidney disease (CKD), and ASCVD risk was estimated using the PREVENT-ASCVD calculator. Elevated Lp(a) was defined as ≥30 mg/dL. We assessed the association of elevated Lp(a) with incident ASCVD events (nonfatal myocardial infarction, fatal coronary heart disease, or ischemic stroke) after Visit 4 through 12/31/22 with multivariable adjusted Cox proportional hazards models. Lp(a) associations were assessed for those with and without individual risk factors, by number of risk factors and at different levels of predicted ASCVD risk. Results: Among 9,483 participants (mean age 63, 58% female, and 21% Black adults), there were 2,311 ASCVD events over a median 19.8 years of follow-up. Individuals with smoking, diabetes, hypertension, and CKD had nominally higher risk (HRs 1.3-1.4) associated with elevated Lp(a) than those without those conditions (HRs 1.1-1.2), with a significant interaction for diabetes (p=0.02) ( Table 1 ). Higher risk in association with elevated Lp(a) was seen for those with a greater number of risk factors and with higher predicted ASCVD risk ( Table 2 ). Elevated Lp(a) was associated with higher risk in those with 3-5 risk factors (HR 1.44 [95% 1.20-1.73]) than in those with 0-2 risk factors (HR 1.11 [95% CI: 1.00-1.23] p-interaction = 0.015). Similarly, elevated Lp(a) was associated with higher risk in those with ≥10% predicted risk (HR 1.42 [95% CI: 1.20-1.68]) than in those with <10% predicted risk (HR 1.11 [95% CI 1.00-1.24] p-interaction = 0.017). Conclusions: Elevated Lp(a) has stronger ASCVD risk associations in those with a higher burden of clinical risk factors, especially diabetes. These finding suggest the importance of risk factor prevention and control for reducing Lp(a)-related risk and may help to inform the optimal targeting of emerging Lp(a) lowering therapies.

Shellfish mariculture, particularly of oysters, poses a significant risk for the introduction of non-native species into marine ecosystems. This study investigates the diversity of invertebrate species colonizing live and discarded oyster shells originating from a farm and oyster bar in the Tylihul Estuary, a region with active oyster farming. Advanced molecular techniques identified several invasive species associated with the discarded shells, including Semibalanus balanoides , Austrominius modestus and Monocorophium insidiosum . These taxa have not been previously documented in the region. Our findings indicate that macrofaunal composition differs between live and discarded oyster shells, and that the richness of invasive invertebrates associated with oyster shells is higher than expected. The discarded shells not only act as a substrate for colonization but also serve as potential vectors for biological invasions. We performed a Species-related Risk Assessment to identify the potential ecological impacts on local biodiversity and ecosystems of the invasive species associated with oysters. Our study proposes management strategies aimed at mitigating the risks associated with shells discarded by oyster bars. Our recommendations include informing recreational travellers and retailers about the implications of discarding shells into the water and advocating for the control of risks related to the use of shells as a construction material.

The thermal coal supply chain serves as core infrastructure for ensuring the safe and stable supply of electricity in China. Effective risk management and control of this supply chain are therefore critical to national energy security and socio-economic development. However, the thermal coal supply chain involves multiple complex risk dimensions, including cross-regional multi-entity coordination, a complex network structure, and a dynamic policy environment. Traditional risk analysis methods often fall short in depicting the concurrent events and dynamic propagation characteristics inherent to such a system. This necessitates systematically investigating the thermal coal supply chain within the Coal–Electricity Joint Venture (CEJV) operational framework, which primarily involves equity-based consolidation and long-term contractual coordination between coal producers and power generators, to comprehensively analyze its critical risk factors and transmission mechanisms. Initially, based on the integration of coal-fired power joint operation policy evolution and industry characteristics, 28 risk factors were identified across three dimensions: internal enterprise, external environment, and overall structure. These encompassed production fluctuation risks, thermal coal transport process risks, and insufficient supply chain flexibility. A dynamic behavior model for the thermal coal supply chain was constructed by analyzing the causal relationships among these risk factors, based on the operational processes of each link. Utilizing Petri net simulation technology enables a quantitative analysis of supply chain risks, facilitating the identification of bottleneck links and potential risk points. Through model simulation, 18 key risk factors were determined, providing a theoretical basis for optimizing supply chain resilience within CEJV enterprises. The limitations of traditional methods in dynamic process modeling and industrial applicability were addressed through a Petri net-based methodology, thereby establishing a novel analytical paradigm for risk management in complex energy supply chains.

Purpose – The research examines how artificial intelligence (AI) affects industrial project outcomes in Kazakhstan by studying cost reduction and time efficiency alongside risk management effectiveness. Methodology – Based on the data collected from publicly accessible project databases, company reports as well as expert surveys, statistical methods including descriptive statistics, regression analysis, and comparative methods such as T-tests or ANOVA, were employed to evaluate the performance of 40 industrial projects in Kazakhstan, comprising 20 AI-enabled and 20 non-AI projects from the energy, manufacturing, and infrastructure sectors. Originality / value – This research addresses the insufficient statistical analysis of AI’s impact on industrial projects in Kazakhstan, by providing quantitative evidence of its effects on key project outcomes like cost, timeline, and risk control. This study offers valuable insights for industry executives and government leaders in Kazakhstan regarding the benefits and effectiveness of AI integration in industrial projects for enhanced performance and sustainable economic expansion. Findings – The study’s findings indicate that the adoption of AI in industrial projects in Kazakhstan leads to a reduction in project costs by 10% and an acceleration of project timeframes by 15%, when compared to traditional non-AI project efforts. Furthermore, AI implementation resulted in a 40% reduction in project risks and a higher project success rate of 92% for AI-enabled projects versus 85% for non-AI projects.

To explore the role of the traditional Chinese medicine (TCM) health management model in improving the quality of clinical nursing services. The respiratory department of a tertiary hospital began constructing and applying the TCM health management model in 2020. Taking this as the time node, the study was divided into two stages: between February and October 2019 (pre-implementation phase) and between April and December 2020 (post-implementation phase). A total of 300 patients were selected from those admitted during the post-implementation phase as the post-implementation group; another 300 patients were selected from those admitted during the pre-implementation phase as the pre-implementation group. Nursing management quality indicators, nursing service quality scores and patient satisfaction were compared and analysed between the two groups. Sensitivity analyses were conducted to adjust for potential time-related confounders. Compared with the pre-implementation group, the incidence of adverse events in the post-implementation group was significantly reduced (3.33% vs. 8.67%, p < 0.05), the qualified rate of nursing documents increased to 98.33% (vs. 89.33%, p < 0.05) and the service response time was shortened to 5.2 ± 1.3 min (vs. 9.8 ± 2.1, p < 0.05). Regarding nursing service quality, the scores for guidance service, health education, triage nursing, medication management and risk control increased to 90.55 ± 4.18, 91.48 ± 4.29, 90.82 ± 3.97, 90.17 ± 4.49 and 91.65 ± 4.50 points, respectively (all p < 0.05). For patient satisfaction scores, service situation, treatment environment and treatment experience increased to 90.51 ± 4.71, 90.69 ± 4.78 and 90.29 ± 4.12 points, respectively (all p < 0.05). Within the context of this single-centre quasi-experimental study, the TCM health management model showed potential in optimising nursing processes and improving selected quality indicators in the respiratory department. However, these findings require validation through multicentre randomised controlled trials before broader implementation can be recommended. This study offers preliminary insights into integrating traditional medicine approaches with modern nursing practices, although further research is needed to establish broader applicability, particularly in international settings.

Against the backdrop of policy-driven transformation in construction industrialization, the EPC general contracting model has emerged as a core pathway for the large-scale development of prefabricated buildings. However, the EPC mode integrates the links of design, procurement, production, and transportation, construction, resulting in a complex coupling correlation among the risk factors of prefabricated construction schedule, which is easy to induce the risk contagion effect and increase the difficulty of risk control of project schedule delay. To address this, this study constructs a hybrid model integrating the “Fuzzy Interpretive Structural Model (FISM)-Coupling Degree Model-Bayesian Network (BN)” to systematically analyze risk contagion mechanisms. Taking an EPC prefabricated building project as an example, FISM is used to reveal the hierarchical structure of risk factors, while the coupling degree model quantifies interaction strengths and maps them into the BN to optimize conditional probability parameters. Through comprehensive hazard analysis, seven key causal risk factors and two critical risk propagation paths are identified. Targeted control measures are designed for the key risk factors, and BN-based simulation is applied to locate critical risk nodes and implement break-chain interventions for the risk paths, resulting in a 23% reduction in the probability of schedule delay. Engineering applications demonstrate that this model can effectively achieve the dynamic identification and blocking of risk paths, providing valuable reference for similar projects and offering informed support for managers in formulating scientific response strategies.

Antibiotic accumulation in saline-alkaline soils is an escalating problem in agriculture and is receiving increasing attention. Biochar amendment for saline-alkaline land is an effective approach to improve soil quality. However, the effects of modified biochar on the antibiotic adsorption mechanism and occurrence forms in saline-alkaline land remain unclear. In this study, phosphate acid-modified biochar (PBC) was used as the research material, and its effects on the adsorption mechanism, occurrence forms, and bioavailability of oxytetracycline (OTC) in salinized-alkaline soil were investigated. The results showed that the adsorption of OTC by saline-alkali soil amended with 5% PBC reached equilibrium within 24h, with an adsorption capacity ranging from 2.05 to 4.14mg/g. The pseudo-second-order kinetic model (R2 = 0.9754-0.9422) was more suitable for describing this adsorption process, indicating that chemical adsorption dominated. Considering the complex compositions of saline-alkali soil and PBC provided multiple adsorption sites with uneven energy distribution for OTC adsorption (a characteristic of multi-layer adsorption), the Freundlich model (R2 = 0.9956-0.9634) better fitted the adsorption behavior. PBC effectively reduced the bioavailability of OTC by improving soil properties, including decreasing the pH value of saline-alkali soil, reducing soluble salt content, and increasing organic matter content (p < 0.01). The bioavailability of OTC decreased by 16.80-25.23% (p < 0.05) in Na2SO4 saline-alkali soil and by 11.56-15.49% (p < 0.05) in Na2CO3 saline-alkali soil. The application of PBC in the remediation of OTC-contaminated saline-alkali soil not only enhanced the adsorption capacity for OTC but also improved the physicochemical properties of saline-alkali soil, thereby achieving the synergistic effect of pollution control and soil improvement, as well as effective control of environmental risks.

Insights into the combined effect of ofloxacin and humic substances on sewage sludge anaerobic digestion.