The paper discusses the experimental optimisation of both chemical and mechanical cleaning procedures for a flat-sheet submerged membrane bioreactor fed with municipal wastewater. Fouling was evaluated by means of the critical flux concept, which was experimentally measured by short-term flux-stepping tests. By keeping constant most important parameters of the biological process (MLSS, sludge age), two different chemical cleaning protocols (2,000 mg L−1 NaOCl and 200 mg L−1 NaOCl) were applied with different frequency and, after approximately 9 months of operation, the criticality threshold was determined under different values of SADm (specific aeration demand per unit of membrane surface area). The weaker and more frequent chemical cleaning regime (200 mg L−1, monthly) proved much more effective than the stronger and less frequent strategy (2,000 mg L−1, once every three months). The improvement of performances was quantified by two TMP-based parameters, the fouling rate and the ΔTMP (difference between TMP values during the increasing and decreasing phase of hysteresis). The best performing configuration was then checked over a longer period by running four long-term trials showing an exponential trend of the sub-critical fouling rate with the imposed flux.