ARDS is the most common acute respiratory failure in the ICU and the mortality rate is still as high as 40%. Mechanical ventilation(MV) is the major supportive treatment for ARDS, but inappropriate ventilator setting could lead to patients suffering from Ventilator-Induced Lung Injury(VILI).

VILI is an important factor in the aggravation of lung injury during MV. The main mechanism of VILI is the unreasonable pressure change (stress) causing excessive local stretch of the lung (strain), which eventually exceeds the capacity of the lung.

The protective strategies during MV (limited platform pressure, low tidal volume, suitable PEEP) are important means of avoiding VILI during MV. The essences of these strategies are to limit the stress and strain of the lung during MV. However, these lung protective ventilation strategies only start from a single indicator and have certain limitations. Considering the various shortcoming of the current strategies, Amato et al. combined two indicators and proposed the concept of driving pressure(driving pressure=tidal volume/respiratory compliance). Several studies also confirmed that limiting the driving pressure can significantly improve patients' outcomes. But the concept of driving pressure and its safety threshold have certain limitations.

Taking into the limitations of existing low tidal volume, limited platform pressure, and restricted driving pressure strategies in lung protection ventilation, Gattinoni et al. first integrated the all factors such as driving pressure, respiratory rate, airway resistance, respiratory rate and PEEP together and the concept of mechanical power was formally proposed.There is a good correlation between mechanical power and lung strain in a certain PEEP range. Cressoni et al. demonstrated through animal experiments that excessive mechanical power during MV caused significant VILI in animals; Guérin et al. also found that mechanical power was closely related to patient outcome in patients with ARDS. Not only that, but Gattinoni reanalyzed Güldner's experimental data and found that mechanical power is more valuable in reflecting lung damage than driving pressure. Mechanical power is a good indicator of response to patient VILI.

Therefore, the investigators hypothesized that only limiting the driving pressure during MV of patients could not achieve ideal lung protective ventilation. Mechanical power may be a better indicator of response VILI; and the safety threshold of driving pressure based on retrospective analysis may not be suitable for patients with severe ARDS, and a lower driving pressure can protect patients with severe ARDS. This study intends to use a single-center, self-controlled study design to reflect lung injury through stress and strain and mechanical work of the lungs, to verify the safety of different driving pressures for severe ARDS, and to further find a safer driving margin for patients with severe ARDS

Inclusion Criteria:

1. diagnosed as moderate to severe ARDS patients(P/F ratio<150mmHg)
2. ICU hospitalization time is expected to be more than 24 hours

Exclusion Criteria:

1. patients whose age<18 years old or >75 years old
2. Severe respiratory central depression, high paraplegia, neuromuscular disease;
3. esophageal obstruction, perforation, severe upper gastrointestinal bleeding / surgery, hernia, thoracic deformity;
4. patients with severe bullae; 5. Patients with severe coagulopathy;

6, severe heart, liver, kidney and other organ failure, hemodynamic instability; 7, the skull base fractures; 8. Other clinical trials are underway;

patients who are diagnosed ARDS in Southeast University Zhongda hospital and hospitalized in ICU from July 1,2018 to April 1,2018